WO2008074249A1 - Frame loss concealment method, system and apparatuses - Google Patents

Abstract

A frame loss concealment method includes: history time-domain waveform is stored in a decoder, and a coder transmits a present modulated lapped transform MLT frame and the obtained frame correlation information to the decoder (501); the decoder performs error/loss frame detection to the present MLT frame and generates error/loss frame information (502); according to the error/loss frame information, the frame correlation information and the history time-domain waveform, the decoder recovered the error/loss frame (503). The invention also discloses a frame loss concealment system, a coder and a decoder. Using the invention, the frame loss concealment ability of MLT coding-decoding scheme is improved, and error/loss frames can be recovered preferably.

Description

丢帧隐藏方法、 系统和装置 Frame dropping method, system and device

技术领域Technical field

本发明涉及音频数据传输技术，具体涉及针对基于 MLT的音频编解 码方案的丢帧隐藏方法、 系统、 编码装置和解码装置。 发明背景 The present invention relates to audio data transmission techniques, and in particular, to a frame loss concealment method, system, encoding apparatus and decoding apparatus for an MLT-based audio codec scheme. Background of the invention

当通过分组网络传输实时音频数据时， 编码后的音频帧被打包并通 过分组网络逐帧传送至解码端。 如果出现丢帧或错帧情况， 将引起解码 端音频播放质量的明显下降， 给听者造成明显的不适感。 因此， 丢帧隐 藏技术应运而生。 丢帧隐藏技术是将丢帧 /错帧隐藏甚至恢复出来的技 术， 它能够緩解因丢帧或错帧给听者造成的不适感。 When real-time audio data is transmitted over a packet network, the encoded audio frames are packed and transmitted frame by frame to the decoder via a packet network. If there is a frame loss or a wrong frame condition, it will cause a significant drop in the audio playback quality of the decoder, which will cause obvious discomfort to the listener. Therefore, frame dropping hidden technology came into being. The frame loss concealment technique is a technique for concealing or even recovering lost/missed frames, which can alleviate the discomfort caused to the listener due to dropped frames or wrong frames.

重叠调制变换（ MLT, Modulated Lapped Transform )是一种编码方 法， 避免了变换区间边界的不连续性， 对量化误差的敏感程度低， 因此 越来越多的应用在音频、视频的编码方案中。 MLT是一个线性的变换过 程， 每一个 MLT变换的输入是时域上两个相邻帧的 2N个采样点， 输出 为 N个 MLT系数。 MLT变换的过程可描述如下： MLT (Modulated Lapped Transform) is a coding method that avoids the discontinuity of the boundary of the transform interval and is less sensitive to quantization error. Therefore, more and more applications are applied in audio and video coding schemes. MLT is a linear transform process. The input of each MLT transform is 2N sample points of two adjacent frames in the time domain, and the output is N MLT coefficients. The process of MLT transformation can be described as follows:

步骤 1 , 对时域波形信号进行重叠相加运算， 并以窗口函数为加权 系数。 Step 1: Perform an overlap and add operation on the time domain waveform signal, and use a window function as the weighting coefficient.

v(n) = w[(N / 2— 1)— n]x[(N /2 - l) - n] + w(N 12 + n)x(N 12 + n) 0<n<N/2 v(n + N/2) = w[(N - 1) - n]x(N + n) - w(n)x[(2N - l) - n] 0<n<N/2 其中，^w(ⁿ) = ( (η +⁰·⁵)^/2Ν)是窗口函数， 0≤n<N。 χ(η)是输入的音频 信号。 v(n)为某个时域帧经重叠相加运算后得到的重叠相加时域帧的前 半帧， v(n + N/2)为同一时域帧经重叠相加运算后得到的重叠相加时域帧 的后半帧。 步骤 2: 对 ")再进行 IV型离散余弦变换（DCT, Discrete Cosine Transform ), 得到最终的 MLT帧。 IV型 DCT变换公式如下：

v(n) = w[(N / 2— 1)— n]x[(N /2 - l) - n] + w(N 12 + n)x(N 12 + n) 0<n<N/ 2 v(n + N/2) = w[(N - 1) - n]x(N + n) - w(n)x[(2N - l) - n] 0<n<N/2 where^w (ⁿ ) = ( (η +⁰ ·⁵ )^/2Ν ) is a window function, 0 ≤ n < N. χ(η) is the input audio signal. v(n) is the first half of the overlap-added time-domain frame obtained by the overlap-add operation of a certain time-domain frame, and v(n + N/2) is the overlap obtained by the overlap-add operation of the same time-domain frame. Add the last half of the time domain frame. Step 2: Perform a type IV discrete cosine transform (DCT, Discrete Cosine Transform) on ") to obtain the final MLT frame. The IV type DCT transform formula is as follows:

其中， m为 MLT帧中各数据的序数， n为参加 DCT变换的重叠相 加时域帧中各数据的序数。 Where m is the ordinal number of each data in the MLT frame, and n is the ordinal number of each data in the overlap-added time domain frame participating in the DCT transform.

图 1为现有技术 MLT重叠相加运算机制的原理示意图，即上述步骤 1的计算。 如图 1所示， 1 , 2 , 3 , 4 , 5 , 6代表输入时域音频信号。 其中 1和 2, 3和 4, 5和 6分别代表的是连续三帧中的第 1帧、 第 2帧 和第 3帧，每帧的帧长都为 N。每帧又进一步分为长度相等的两个子帧， 子帧长为 N/2。 a, b , c , d , e , f 分别代表时域音频信号经上述步 骤 1重叠相加之后得到的重叠相加时域帧。 其中， a, c, e分别为各重 叠相加时域帧的前半帧， b, d , f分别为各重叠相加时域帧的后半帧。 FIG. 1 is a schematic diagram of the principle of the prior art MLT overlap addition operation mechanism, that is, the calculation of the above step 1. As shown in Figure 1, 1 , 2 , 3 , 4 , 5 , 6 represent input time domain audio signals. Where 1 and 2, 3 and 4, 5 and 6 respectively represent the first frame, the second frame and the third frame of three consecutive frames, each frame has a frame length of N. Each frame is further divided into two subframes of equal length, and the subframe length is N/2. a, b, c, d, e, f respectively represent overlapping and added time domain frames obtained by overlapping and adding the time domain audio signals through the above step 1. Where a, c, and e are the first half of each overlap-added time domain frame, and b, d, and f are the second half of each overlap-added time-domain frame.

以图 1所示第 2帧为例， 结合以上公式可以看出， c是 3和 4分别 通过窗口函数 w(n)的加权后再重叠相加计算得到的， d是 5和 6分别通 过窗口函数 w(n)的加权后再重叠相加计算得到的， 如图中箭头所示。 对 c d再进行步骤 2的 IV型 DCT变换， 得到最终的 MLT系数， 即完成 了基于 MLT的音频编码。 以此类推， 对其它帧的处理与对第 2帧的处 理相同。 可见， 在 MLT变换中， 为得到一帧的 MLT帧， 需要对时域上 连续两帧的时域信号 3, 4, 5, 6进行计算。 Taking the second frame shown in Figure 1 as an example, it can be seen from the above formula that c is 3 and 4 respectively calculated by weighting the window function w(n) and then superimposing and adding, d is 5 and 6 respectively through the window The weighting of the function w(n) is then added to the overlap and added, as indicated by the arrows in the figure. For the c d, the IV type DCT transform of step 2 is performed to obtain the final MLT coefficient, that is, the MLT-based audio coding is completed. By analogy, the processing of other frames is the same as the processing for the second frame. It can be seen that in the MLT transform, in order to obtain an MLT frame of one frame, it is necessary to calculate the time domain signals 3, 4, 5, 6 of two consecutive frames in the time domain.

发送端将上述基于 MLT音频编码的 MLT帧经分组通信网络传送给 接收端， 接收端采用逆向 MLT变换将接收到的 MLT帧还原成时域帧， 便得到了发送端当前传送的一帧实际音频数据。 The transmitting end transmits the MLT frame based on the MLT audio encoding to the receiving end via the packet communication network, and the receiving end uses the reverse MLT transform to restore the received MLT frame to the time domain frame, thereby obtaining an actual audio frame currently transmitted by the transmitting end. data.

逆向重叠调制变换（ IMLT, Inverse Modulated Lapped Transform )是 MLT变换的逆过程。 IMLT变换的具体步骤如下： 步骤 1，： 将当前 MLT 帧进行逆向离散余弦变换（IDCT, Inverse Discrete Cosine Transform ), IDCT公式为：

Inverse Modulated Lapped Transform (IMLT) is the inverse of the MLT transform. The specific steps of the IMLT transformation are as follows: Step 1:: Perform the inverse discrete cosine transform (IDCT, Inverse Discrete Cosine Transform) on the current MLT frame, and the IDCT formula is:

u(n)为经 IDCT得到的重叠相加时域帧。 u(n) is the overlap-added time domain frame obtained by IDCT.

步骤 2，：用当前 MLT帧 IDCT变换结果的前一半与前一 MLT帧 IDCT 变换结果的后一半进行重叠相加运算， 并以窗口函数为加权系数。 公式 如下： Step 2: The first half of the current MLT frame IDCT transform result is overlapped and added with the last half of the previous MLT frame IDCT transform result, and the window function is used as the weighting coefficient. The formula is as follows:

y(n) = w(n)u[(N /2-l)-n] + w[(N

old (n)y(n) = w(n)u[(N /2-l)-n] + w[(N

Old (n)

0<n<N/20<n<N/2

y(n + N/2) = w(N/2 + n)u(n)-w[(N/2-l)-n\i_old[(N/2-l)-n] y(n + N/2) = w(N/2 + n)u(n)-w[(N/2-l)-n\i_old[(N/2-l)-n]

0<n<N/2 0<n<N/2

其中， w(n)是与 MLT变换相同的窗口函数， y(n)是输出的时域信号。 u(n)为当前帧 IDCT变换后的重叠相加时域帧， u_old(n)为前一帧 IDCT 变换后的重叠相加时域帧。 Where w(n) is the same window function as the MLT transform, and y(n) is the output time domain signal. u(n) is the overlap-added time domain frame of the current frame IDCT transform, and u_old(n) is the overlap-added time domain frame of the previous frame IDCT transform.

图 2为现有技术 IMLT的原理示意图。 如图 2所示， a, b, c, d, e, f为经 IDCT变换得到的重叠相加时域帧， 对应步骤 2，中的 u(n)。 1, 2, 3, 4, 5, 6为输出的时域帧， 对应步骤 2，中的 y(n)。 当已知 a, b, c, d, e, f欲求 1, 2, 3, 4, 5, 6时， 3是由 b, c分别通过窗口函数的加权 后再重叠相加计算得到， 如图 2中实线箭头所示。 同时， 4也是由 b, c 分别通过窗口函数的加权后再重叠相加计算得到， 如图 2中虚线箭头所 示。 以此类推， 在 IMLT变换中， 为得到一帧的时域信号， 如 3, 4, 需 要用当前帧经 IDCT变换的前半部分和前一帧经 IDCT变换的后半部分 进行计算得到的， 如 c, b。 2 is a schematic diagram of the principle of the prior art IMLT. As shown in FIG. 2, a, b, c, d, e, and f are overlapping and added time domain frames obtained by IDCT transform, corresponding to u(n) in step 2. 1, 2, 3, 4, 5, 6 are the output time domain frames, corresponding to y(n) in step 2. When a, b, c, d, e, f is known to be 1, 2, 3, 4, 5, 6, 3 is calculated by weighting the b and c respectively through the window function and then adding and subtracting, as shown in the figure. 2 solid arrows indicate. At the same time, 4 is also calculated by b, c respectively weighted by the window function and then overlapped and added, as shown by the dotted arrow in Figure 2. By analogy, in the IMLT transform, in order to obtain a time domain signal of one frame, such as 3, 4, it is required to calculate the first half of the current frame and the second half of the previous frame by the IDCT transform, such as c, b.

国际电信联盟远程通信标准化组（ITU-T)的 G.722.1音频编码标准 定义了基于 MLT 的音频编码方案的丢帧隐藏策略。 该标准定义的丢帧 隐藏策略十分筒单， 可描述如下：The G.722.1 audio coding standard of the International Telecommunication Union Telecommunication Standardization Group (ITU-T) defines a frame loss concealment strategy for MLT-based audio coding schemes. Frame loss defined by the standard The hidden strategy is very simple and can be described as follows:

1、 对于单帧的丢失， 重复前一帧的 MLT帧即可。 1. For the loss of a single frame, repeat the MLT frame of the previous frame.

2、 对于出现连续丢帧的情况， 全部丢帧的 MLT帧均被置 0。 2. For the case of continuous frame loss, all MLT frames that are dropped are set to 0.

由于采用了直接重复前一帧的 MLT帧的隐藏策略，采用该丢帧隐藏 策略恢复出的丢帧有很大部分是对前一帧的重复。 图 3为 MLT音频编 码方案中， 现有技术对丢失单帧的丢帧隐藏效果图。 图 3中， 上方的波 形是原始音频信号的波形， 下方是经丢帧隐藏后的波形。 丢失的帧段位 于波形中央能量较低处。 可以看出， 恢复出的丢失帧的波形与该帧前一 帧的波形十分相似， 而与实际波形相去甚远。 Due to the hidden strategy of directly repeating the MLT frame of the previous frame, a large part of the lost frame recovered by the lost frame hiding strategy is a repetition of the previous frame. FIG. 3 is a diagram showing the effect of the prior art on the lost frame loss of a single frame in the MLT audio coding scheme. In Figure 3, the upper waveform is the waveform of the original audio signal, and the lower is the waveform hidden by the lost frame. The missing frame segment is located at a lower energy center of the waveform. It can be seen that the waveform of the recovered lost frame is very similar to the waveform of the previous frame of the frame, and is far from the actual waveform.

图 4为 MLT音频编码方案中，现有技术对连续丢帧的丢帧隐藏效果 图。 图 4 ( a )说明出现 2个连续丢帧时， 由 G.722.1隐藏策略得到的恢 复信号及其原始音频信号波形； 图 4 ( b )说明出现 3个连续丢帧时， 由 G722.1隐藏策略得到的恢复信号及其原始音频信号波形；其中原始音频 信号波形在各图的上方， 下方为丢帧隐藏后的恢复波形。 可以看出， 将 丢帧的 MLT帧置 0, 使得恢复的信号出现 0信号点或 0信号段。 FIG. 4 is a diagram showing the effect of the prior art on the frame loss of continuous frame loss in the MLT audio coding scheme. Figure 4 (a) shows the recovered signal and its original audio signal waveform obtained by the G.722.1 hiding strategy when two consecutive frames are dropped. Figure 4 (b) shows that when three consecutive frames are dropped, it is hidden by G722.1. The recovery signal obtained by the strategy and its original audio signal waveform; wherein the original audio signal waveform is above the respective graphs, and below is the recovered waveform after the lost frame is hidden. It can be seen that the MLT frame of the dropped frame is set to 0, so that the recovered signal has a 0 signal point or a 0 signal segment.

因此， G.722.1提供的丢帧隐藏策略的一个缺点是， 在处理单帧丢失 时， 只在较稳定的信号段中有较好的隐藏效果， 在变化较大的信号段， 如信号的过渡段， 其隐藏效果较差。 该丢帧隐藏策略的另一个缺点是， 在处理连续丢帧情况时， 将全部丢帧的 MLT系数置 0, 则恢复信号出现 0信号， 导致在播放该恢复音频时会感到明显中断， 给听着造成不适感。 Therefore, one of the disadvantages of the frame loss concealment strategy provided by G.722.1 is that it has better concealment effect in the more stable signal segment when dealing with single frame loss, and in the signal segment with large changes, such as signal transition. Segment, its hidden effect is poor. Another disadvantage of the frame loss concealment strategy is that when the continuous frame loss situation is processed, the MLT coefficient of all frame drops is set to 0, and the recovery signal shows a 0 signal, which causes a significant interruption when playing the restored audio. Causes discomfort.

从整个编解码系统的高度看， G.722.1所提供的丢帧隐藏策略属于基 于解码端的隐藏技术。 基于解码端的隐藏技术是一种被动隐藏， 所有用 于恢复丢帧的信息均是由解码端自己得到的。 相应的， 基于编码端的隐 藏技术是一种主动隐藏， 一些包含有帧内重要信息的比特在编码端生成 并传送给解码端， 解码端依据这些辅助信息能够较基于解码端的隐藏技 术更好的将丢帧恢复出来， 代价则是一定程度上增加了带宽。From the height of the entire codec system, the frame loss hiding strategy provided by G.722.1 belongs to the hidden technology based on the decoder. The decoding technology based on the decoding side is a passive hiding, and all the information used to recover the lost frames is obtained by the decoding end itself. Correspondingly, the hiding technology based on the encoding end is an active hiding. Some bits containing important information in the frame are generated and transmitted to the decoding end at the encoding end, and the decoding end can be based on the hidden technology based on the decoding end. The better way to recover lost frames is to increase the bandwidth to some extent.

音频编码标准如 G.729.1中涉及了主动丢帧隐藏技术， 即由 G.729.1 的编码端产生并传送用于辅助解码端进行丢帧恢复的辅助信息， 该辅助 信息包括帧类型信息， 帧能量信息和帧相位信息。 其中， 帧类型信息是 由 2比特表示的 5种音频帧的类型， 分别为： 非发声类， 非发声过渡类， 发声过渡类， 发声类和起始类。 在 G.729.1 的解码端， 根据丢失帧的不 同类型， 隐藏算法采用相应的隐藏策略进行隐藏。 The audio coding standard, such as G.729.1, involves the active frame dropping technique, that is, the auxiliary information generated by the encoding end of G.729.1 and used for assisting the decoding end to perform frame loss recovery, the auxiliary information includes frame type information, frame energy. Information and frame phase information. The frame type information is a type of five types of audio frames represented by 2 bits, namely: a non-sounding class, a non-sounding transition class, a vocal transition class, a utterance class, and a start class. On the decoding side of G.729.1, according to the different types of lost frames, the hiding algorithm uses the corresponding hidden strategy to hide.

G.729.1提供的丢帧隐藏技术， 对长度较长的帧有较好的恢复能力， 却完全没有对短音符帧的恢复能力。 5种分类中没有一种可以标识出短 音符帧的存在。 在实际传输过程中， 音频信号中的短音符帧可能就此丢 失， 因而对接收到的音频质量有较大影响。 另外， G729.1的丢帧隐藏算 法并非基于 MLT的音频编码方案专门设计的， 因而没有充分利用 MLT 变换本身的特征， 在算法效率和隐藏质量上并非最优。 The frame dropping technology provided by G.729.1 has better recovery ability for frames with longer lengths, but has no ability to recover short note frames. None of the five categories can identify the presence of a short note frame. In the actual transmission process, short note frames in the audio signal may be lost, and thus have a large influence on the received audio quality. In addition, the G729.1 frame loss concealment algorithm is not specifically designed based on the MLT audio coding scheme, so the characteristics of the MLT transform itself are not fully utilized, and the algorithm efficiency and hidden quality are not optimal.

可见，现有的多种丢帧隐藏技术， 没有能够在基于 MLT的音频编码 方案中实现较好的丢帧隐藏， 较为准确的恢复出丢失帧的时域波形。 发明内容 It can be seen that the existing multiple frame dropping techniques have not been able to achieve better frame loss hiding in the MLT-based audio coding scheme, and more accurately recover the time domain waveform of the lost frame. Summary of the invention

有鉴于此，本发明实施例的第一个目的在于提供一种丢帧隐藏方法， 能够提高恢复错 /丢帧的准确性。 In view of this, the first object of the embodiments of the present invention is to provide a frame dropping method, which can improve the accuracy of recovering errors/drop frames.

本发明实施例的第二个目的在于提供一种丢帧隐藏编码装置， 能够 提高恢复错 /丢帧的准确性。 A second object of the embodiments of the present invention is to provide a frame loss concealment coding apparatus, which can improve the accuracy of recovering a wrong/lost frame.

本发明实施例的第三个目的在于提供一种丢帧隐藏解码装置， 能够 提高恢复错 /丢帧的准确性。 A third object of the embodiments of the present invention is to provide a frame loss concealment decoding apparatus, which can improve the accuracy of recovering a wrong/lost frame.

本发明实施例的第四个目的在于提供一种丢帧隐藏系统， 能够提高 恢复错 /丢帧的准确性。 为达到上述发明目的的第一个方面， 本发明提供了一种丢帧隐藏方 法， 在解码端存储历史时域波形； 该方法包括：A fourth object of the embodiments of the present invention is to provide a frame loss concealment system, which can improve the accuracy of recovering errors/drop frames. In order to achieve the first aspect of the above object, the present invention provides a frame dropping method for storing a historical time domain waveform at a decoding end; the method includes:

编码端将获得的相关帧信息以及当前 MLT帧发送给解码端； 所述解码端对接收的所述当前 MLT帧进行错 /丢帧检测， 生成错 /丢 帧信息； 根据所述错 /丢帧信息， 所述相关帧信息和所述历史时域波形恢 复出错 /丢帧。 The encoding end sends the obtained related frame information and the current MLT frame to the decoding end; the decoding end performs error/drop frame detection on the received current MLT frame, and generates error/lost frame information; according to the error/drop frame Information, the related frame information and the historical time domain waveform recover error/drop frame.

为达到上述发明目的的第二个方面， 本发明实施例提供了一种丢帧 隐藏编码装置， 该编码装置包括 MLT编码单元、 历史緩沖单元和相关 帧信息生成单元； In order to achieve the second aspect of the present invention, an embodiment of the present invention provides a frame loss concealment coding apparatus, where the coding apparatus includes an MLT coding unit, a history buffer unit, and an associated frame information generation unit;

所述 MLT编码单元，用于根据该编码装置获取的后续帧和从所述历 史緩沖单元获取的当前帧进行 MLT变换， 生成 MLT帧， 并发送出去； 所述历史緩沖单元， 用于緩沖所述当前帧和所述前置帧； The MLT encoding unit is configured to perform MLT transformation according to a subsequent frame acquired by the encoding device and a current frame acquired from the history buffer unit, generate an MLT frame, and send the MLT frame, and send the data out to the history buffer unit. a current frame and the preamble frame;

所述相关帧信息生成单元， 用于根据该编码装置获取的后续帧， 以 及从所述历史单元緩沖获取的所述当前帧和所述前置帧， 生成相关帧信 息， 并发送出去。 The related frame information generating unit is configured to generate related frame information according to a subsequent frame acquired by the encoding device, and the current frame and the pre-frame obtained from the history unit buffer, and send the related frame information.

为达到上述发明目的的第三个方面， 本发明实施例提供了一种丢帧 隐藏解码装置， 该解码装置包括帧错检测器、 丢帧隐藏单元和解码波形 历史緩沖； In order to achieve the third aspect of the present invention, an embodiment of the present invention provides a frame loss concealment decoding apparatus, where the decoding apparatus includes a frame error detector, a frame loss concealing unit, and a decoding waveform history buffer.

所述解码波形历史緩沖， 用于存储历史时域波形； The decoded waveform history buffer is used to store a historical time domain waveform;

所述帧错检测器， 用于对当前 MLT帧进行错 /丢帧检测， 生成错 /丢 帧信息； 并将所述错 /丢帧信息发送给所述丢帧隐藏单元； The frame error detector is configured to perform error/drop frame detection on the current MLT frame, generate error/lost frame information, and send the error/drop frame information to the frame loss concealment unit;

所述丢帧隐藏单元， 用于根据接收自所述帧错检测器的所述错 /丢帧 信息、 接收自编码装置的相关帧信息和所述解码波形历史緩沖中的历史 时域波形恢复出错 /丢帧。 The frame loss concealing unit is configured to recover an error according to the error/lost frame information received from the frame error detector, related frame information received from an encoding device, and a historical time domain waveform in the decoded waveform history buffer. / Drop frame.

为达到上述发明目的的第四个方面， 本发明实施例提供了一种丢帧 隐藏系统， 该系统包括编码装置和解码装置：In order to achieve the fourth aspect of the above object, the embodiment of the present invention provides a frame dropping frame. Hidden system, the system comprising an encoding device and a decoding device:

所述编码装置，用于将获得的相关帧信息以及当前 MLT帧发送给解 码装置； The encoding device is configured to send the obtained related frame information and the current MLT frame to the decoding device;

所述解码装置， 用于对接收的所述当前 MLT帧进行错 /丢帧检测， 生成错 /丢帧信息； 根据所述错 /丢帧信息， 所述相关帧信息和历史时域 波形恢复出错 /丢帧。 The decoding device is configured to perform error/drop frame detection on the received current MLT frame, and generate error/drop frame information; according to the error/drop frame information, the related frame information and the historical time domain waveform recovery error / Drop frame.

与现有技术相比， 本发明实施例所提供的丢帧隐藏方法、 系统、 编 码装置和解码装置， 提供了一种主动丢帧隐藏技术， 在编码端产生相关 帧信息并随经 MLT编码的 MLT帧发送至解码端。 解码端如果接收到的 是错 /丢帧， 则可以根据相关帧信息比较好的恢复出错 /丢帧的原始波形。 附图简要说明 Compared with the prior art, the frame dropping method, system, encoding device and decoding device provided by the embodiments of the present invention provide an active frame dropping technique, which generates relevant frame information at the encoding end and is encoded by the MLT. The MLT frame is sent to the decoder. If the decoder receives the error/drop frame, the original waveform of the error/drop frame can be recovered according to the relevant frame information. BRIEF DESCRIPTION OF THE DRAWINGS

图 1为现有技术 MLT重叠相加运算机制的原理示意图。 FIG. 1 is a schematic diagram showing the principle of a prior art MLT overlap addition operation mechanism.

图 2为现有技术 IMLT重叠相加运算的原理示意图。 FIG. 2 is a schematic diagram of the principle of the prior art IMLT overlap addition operation.

图 3为 MLT音频编码方案中，现有技术对丢失单帧的丢帧隐藏效果 图。 FIG. 3 is a diagram showing the effect of the prior art on the lost frame loss of a single frame in the MLT audio coding scheme.

图 4 ( a ) 为 MLT音频编码方案中， 现有技术对 2个连续丢帧的丢 帧隐藏效果图。 Figure 4 (a) shows the effect of the prior art on the frame loss of two consecutive dropped frames in the MLT audio coding scheme.

图 4 ( b )为 MLT音频编码方案中， 现有技术对 3个连续丢帧的丢 帧隐藏效果图。 Figure 4 (b) shows the effect of the prior art on the frame loss of three consecutive dropped frames in the MLT audio coding scheme.

图 5为本发明实施例丢帧隐藏方法的方法流程图。 FIG. 5 is a flowchart of a method for dropping a frame hiding method according to an embodiment of the present invention.

图 6为本发明实施例丢帧隐藏方法第一较佳实施例的方法流程图。 图 7为本发明实施例解码端选择隐藏策略的方法流程图。 FIG. 6 is a flowchart of a method for deleting a frame hiding method according to a first preferred embodiment of the present invention. FIG. 7 is a flowchart of a method for selecting a hiding policy by a decoding end according to an embodiment of the present invention.

图 8为本发明实施例隐藏策略原理示意图。 FIG. 8 is a schematic diagram of a principle of a hidden policy according to an embodiment of the present invention

图 9为本发明实施例连续丢帧 /错帧的隐藏策略选择方法流程图。 图 10为本发明实施例丢帧隐藏系统组成框图。FIG. 9 is a flowchart of a method for selecting a hidden frame of a continuous frame loss/error frame according to an embodiment of the present invention. FIG. 10 is a block diagram of a frame dropping system according to an embodiment of the present invention.

图 11 为本发明实施例丢帧隐藏编码装置第一较佳实施例的组成框 图。 FIG. 11 is a block diagram showing a first preferred embodiment of a frame dropping concealment coding apparatus according to an embodiment of the present invention.

图 12 为本发明实施例丢帧隐藏解码装置第一较佳实施例的组成框 图。 FIG. 12 is a block diagram showing a first preferred embodiment of a frame loss concealment decoding apparatus according to an embodiment of the present invention.

图 13为本发明实施例丢帧隐藏方法第二较佳实施例的方法流程图。 图 14 为本发明实施例丢帧隐藏编码装置第二较佳实施例的组成框 图。 FIG. 13 is a flowchart of a method for deleting a frame hiding method according to a second preferred embodiment of the present invention. FIG. 14 is a block diagram showing a second preferred embodiment of a frame dropping concealment coding apparatus according to an embodiment of the present invention.

图 15为本发明实施例丢帧隐藏方法第三较佳实施例的方法流程图。 实施本发明的方式 FIG. 15 is a flowchart of a method for deleting a frame hiding method according to a third preferred embodiment of the present invention. Mode for carrying out the invention

为使本发明实施例的目的、 技术方案和优点更加清楚明白， 下面结 合附图， 对本发明实施例进一步详细说明。 In order to make the objects, the technical solutions and the advantages of the embodiments of the present invention more clearly, the embodiments of the present invention are further described in detail below with reference to the accompanying drawings.

本发明实施例的核心思想是， 提供一种主动丢帧隐藏技术， 编码端 将获得的相关帧信息以及当前 MLT 帧发送给解码端。 解码端对接收的 当前 MLT帧进行错 /丢帧检测， 生成错 /丢帧信息； 在进行丢帧隐藏处理 时， 解码端根据错 /丢帧信息， 相关帧信息和历史时域波形恢复出错 /丢 帧， 从而完成了丢帧隐藏。 The core idea of the embodiment of the present invention is to provide an active frame dropping prevention technology, where the encoding end sends the obtained related frame information and the current MLT frame to the decoding end. The decoding end performs error/drop frame detection on the received current MLT frame, and generates error/lost frame information. When performing frame loss concealment processing, the decoding end recovers errors according to the error/lost frame information, the related frame information, and the historical time domain waveform/ Drop frames, thus completing frame loss hiding.

本发明实施例是针对基于 MLT的编解码方案的丢帧隐藏方案。 基于以上本发明实施例的核心思想， 图 5示出了本发明实施例丢帧 隐藏方法的方法流程图， 该方法的具体步骤如下： Embodiments of the present invention are directed to a frame loss concealment scheme for an MLT-based codec scheme. Based on the core idea of the embodiment of the present invention, FIG. 5 is a flowchart of a method for dropping a frame hiding method according to an embodiment of the present invention. The specific steps of the method are as follows:

步骤 501 , 编码端将获得的相关帧信息以及当前 MLT帧发送给解码 端。 Step 501: The encoding end sends the obtained related frame information and the current MLT frame to the decoding end.

步骤 502, 解码端对当前 MLT帧进行错 /丢帧检测， 生成错 /丢帧信 步骤 503, 在进行丢帧隐藏处理时， 解码端根据错 /丢帧信息， 结合 相关帧信息和历史时域波形恢复出错 /丢帧。Step 502: The decoding end performs error/drop frame detection on the current MLT frame, and generates a wrong/lost frame signal. Step 503: When performing frame loss concealment processing, the decoding end recovers the error/drop frame according to the error/lost frame information, combined with the related frame information and the historical time domain waveform.

MLT帧及其相关帧信息可以分包发送， 以免出现丢帧时， 用于进行 丢帧恢复的相关帧信息也随之丢失； 或者， 相关帧信息可以被嵌在与当 前帧紧邻的前置帧或者后续帧中传送， 在当前帧丢失， 相关帧信息不丢 失； 也可以在一个可分层的编码体系中在增强层中传输相关帧信息。 The MLT frame and its associated frame information can be sent in packets to avoid loss of related frame information for frame loss recovery when the frame is lost. Alternatively, the related frame information can be embedded in the preceding frame immediately adjacent to the current frame. Or transmitting in subsequent frames, the relevant frame information is not lost when the current frame is lost; the related frame information may also be transmitted in the enhancement layer in a layered coding system.

在实际应用中， 可以将解码端根据错 /丢帧信息， 结合相关帧信息和 历史时域波形恢复错 /丢帧的具体方法作为隐藏策略以表格或者逻辑推 断的方式预先存储于解码端，不同的错 /丢帧情况和相关帧信息对应不同 的隐藏策略。 在进行丢帧隐藏处理时， 解码端首先根据错 /丢帧信息， 结 合相关帧信息选择出对应的隐藏策略， 再根据该隐藏策略和历史时域波 形恢复出错 /丢帧。 In practical applications, the specific method of recovering the wrong/lost frame according to the error/drop frame information combined with the relevant frame information and the historical time domain waveform may be pre-stored in the decoding end in a table or logical inference manner, differently. The error/drop frame condition and related frame information correspond to different hidden strategies. When performing frame loss concealment processing, the decoding end first selects a corresponding concealment strategy according to the error/lost frame information, combines the relevant frame information, and recovers the error/drop frame according to the concealment strategy and the historical time domain waveform.

图 6为本发明实施例丢帧隐藏方法第一较佳实施例的方法流程图。 该方法采用分包的形式， 将 MLT帧及其相关帧信息分别打包发送给解 码端。 该方法的具体步骤为： FIG. 6 is a flowchart of a method for deleting a frame hiding method according to a first preferred embodiment of the present invention. The method adopts the form of packetization, and separately packages and sends the MLT frame and its related frame information to the decoding end. The specific steps of the method are:

步骤 601 ,编码端根据接收的后续帧和存储的当前帧进行 MLT变换， 生成 MLT帧。 Step 601: The encoding end performs MLT transformation according to the received subsequent frame and the stored current frame to generate an MLT frame.

本步骤中， 从前述实现 MLT变换的步骤 1中可以看出， 由于 MLT 变换的需求， 对一帧音频数据进行编码时， 需要将当前处理帧与一后续 帧进行重叠相加。 因此， 当前处理帧被保存在编码端的历史緩沖中， 当 前处理帧筒称为当前帧。 后续帧实际上才是当前时刻接收的帧， 将其称 为后续帧。 当前帧与后续帧一起进行 MLT变换， 生成当前帧的 MLT编 码数据， 即当前 MLT帧。 MLT变换与现有技术的 MLT变换相同。 In this step, it can be seen from the foregoing step 1 of implementing the MLT transform that, due to the requirement of the MLT transform, when encoding one frame of audio data, the current processed frame and a subsequent frame need to be overlapped and added. Therefore, the current processing frame is saved in the history buffer of the encoding side, and the current processing frame is called the current frame. The subsequent frame is actually the frame received at the current time, which is called the subsequent frame. The current frame is MLT transformed with the subsequent frames to generate MLT encoded data of the current frame, that is, the current MLT frame. The MLT transform is the same as the prior art MLT transform.

步骤 602, 编码端生成当前 MLT帧的相关帧信息。 Step 602: The encoding end generates related frame information of the current MLT frame.

本步骤中， 编码端根据当前帧进行帧内相关性检测， 生成帧内相关 性信息； 并根据当前帧和前置帧进行帧间相关性检测， 生成帧间相关性 信息。 前置帧是当前帧的前一帧音频数据时域波形。 该前置帧也保存在 编码端的历史緩沖中。 因此历史緩沖中保存了两帧的时域波形。In this step, the encoding end performs intra-frame correlation detection according to the current frame to generate intra-frame correlation. Sex information; and inter-frame correlation detection according to the current frame and the pre-frame, and generating inter-frame correlation information. The preamble frame is the time domain waveform of the previous frame of the current frame. The preamble frame is also saved in the history buffer of the encoding side. Therefore, the time domain waveform of two frames is saved in the history buffer.

可见， 本实施例中， 相关帧信息包括帧内相关性信息和帧间相关性 信息。 其中， 帧内相关性是指帧内信号的稳定性， 基音越稳定， 基音周 期性越强， 相关性就越强。 帧间相关性包含了两部分内容， 一为本帧前 半帧与其前置帧或前置帧末尾之间的相关性， 另一部分为本帧前半帧内 的帧内相关性。 当本帧前半帧的基音与前置帧或至少前置帧的末尾处的 基音很相似， 且本帧前半帧的帧内相关性也为强时， 才能认为帧间相关 性为强。 在判断时， 先判断本帧前半帧的帧内相关性， 当本帧前半帧的 帧内相关性为强， 才需要判断本帧前半帧与其前置帧的帧间相关性。 如 果本帧前半帧与其前置帧的帧间相关性为强， 则本帧的帧间相关性为 强， 否则为弱。 如果本帧前半帧的帧内相关性已经为弱， 则不用进行后 续的判断， 直接认为本帧的帧间相关性为弱。 帧内相关性及帧间相关性 强弱的判断可以依靠判决门限来完成， 当相关性高于门限值就判定为相 关性强， 否则为弱。 作为相关帧信息的帧内相关性是指本帧的帧内相关 性。 It can be seen that, in this embodiment, the related frame information includes intra-frame correlation information and inter-frame correlation information. Among them, the intra-frame correlation refers to the stability of the intra-frame signal. The more stable the pitch is, the stronger the pitch period is, and the stronger the correlation is. Inter-frame correlation consists of two parts, one is the correlation between the first half of the frame and the end of its preamble or preamble, and the other is the intraframe correlation in the first half of the frame. When the pitch of the first half of the frame is very similar to the pitch of the preamble or at least the end of the preamble, and the intraframe correlation of the first half of the frame is also strong, the interframe correlation is considered strong. When judging, first determine the intra-frame correlation of the first half of the frame. When the intra-frame correlation of the first half of the frame is strong, it is necessary to determine the inter-frame correlation between the first half of the frame and its pre-frame. If the inter-frame correlation between the first half of the frame and its pre-frame is strong, the inter-frame correlation of the frame is strong, otherwise it is weak. If the intra-frame correlation of the first half of the frame is already weak, no subsequent judgment is made, and the inter-frame correlation of the frame is directly considered to be weak. The intra-frame correlation and the inter-frame correlation can be judged by the decision threshold. When the correlation is higher than the threshold, it is judged to be strong, otherwise it is weak. The intra-correlation as the relevant frame information refers to the intra-frame correlation of the present frame.

帧内相关性和帧间相关性的生成方法， 最直接的一种方法是依靠自 相关， 互相关函数来判断帧内， 帧间相关性。 这里的帧内相关性计算适 用于本帧的帧内相关性计算和本帧前半帧的帧内相关性计算。 在计算本 帧的帧内相关性时， 对本帧求其自相关。 自相关函数定义为： The most straightforward method for generating intra-frame correlation and inter-frame correlation is to rely on autocorrelation and cross-correlation functions to determine intra-frame and inter-frame correlation. The intra-correlation calculation here is applied to the intra-correlation calculation of the current frame and the intra-correlation calculation of the first half of the frame. When calculating the intra-frame correlation of this frame, the auto-correlation of this frame is sought. The autocorrelation function is defined as:

^ χ(η) · x(n - k) ^ χ(η) · x(n - k)

^ x² (n - k)^ x² (n - k)

o k=l ...N-l o k=l ...N-l

在这里， x(n)就是本帧的时域信号， N是信号的长度， k是时移。 帧间相关性可利用互相关函数求得。 首先利用上述公式求得本帧前 半帧的帧内相关性， 其中 x(n)为本帧前半帧的时域信号。 如果为强， 则 在将本帧前半帧的部分信号与前置帧末尾处的相同长度的信号进行互 相关计算， 得到帧间相关性。 归一化互相关函数的定义为：

Here, x(n) is the time domain signal of this frame, N is the length of the signal, and k is the time shift. Inter-frame correlation can be obtained using a cross-correlation function. Firstly, the intra-frame correlation of the first half of the frame is obtained by using the above formula, where x(n) is the time domain signal of the first half of the frame. If it is strong, the partial correlation signal of the first half of the frame is cross-correlated with the signal of the same length at the end of the preamble to obtain inter-frame correlation. The definition of the normalized cross-correlation function is:

在这里， x(n),y(n)分别代表本帧前半帧的部分信号与前置帧末尾处的 相同长度的信号。 当自 /互相关的计算结果高于一个事先定义的门限时 就认为相关性是强的， 反之则弱。 Here, x(n), y(n) respectively represent the partial signal of the first half of the frame and the signal of the same length at the end of the preceding frame. Correlation is considered strong when the self/cross-correlation calculation is above a predefined threshold, and vice versa.

帧间相关性也可以这样计算： 计算本帧前半帧开头和本帧前半帧末 尾处的基音， 如果一致则认为前半帧帧内相关性强， 否则为弱。 当前半 帧帧内相关性为强时， 再与前置帧末尾处的基音相比较， 如果一致就认 为本帧的帧间相关性强， 反之则弱。 The inter-frame correlation can also be calculated as follows: Calculate the pitch at the beginning of the first half of the frame and at the end of the first half of the frame. If they match, the correlation in the first half of the frame is considered strong, otherwise it is weak. When the intra-frame intra-frame correlation is strong, it is compared with the pitch at the end of the pre-frame. If it is consistent, the inter-frame correlation of the frame is strong, and vice versa.

本步骤生成的相关帧信息， 可以采用帧内相关性标识和帧间相关性 标识表示， 各标志由 1比特二进制数表示。 则相关帧信息由帧内和帧间 相关性标识的 2比特组成， 这 2比特代表了 4种不同的组合， 这 4种组 合如表 1所示： The relevant frame information generated in this step may be represented by an intra-frame correlation identifier and an inter-frame correlation identifier, and each flag is represented by a 1-bit binary number. The relevant frame information is composed of 2 bits identified by the intra-frame and inter-frame correlation. These 2 bits represent 4 different combinations. The four combinations are shown in Table 1:

表 1 Table 1

步骤 603 , 编码端将 MLT帧及其相关帧信息发送给解码端。 本步骤中，发给解码端的信息为一 MLT帧以及该帧相应的帧内相关 性信息和帧间相关性信息。 发送时， MLT帧和相关帧信息分包发送。Step 603: The encoding end sends the MLT frame and its related frame information to the decoding end. In this step, the information sent to the decoding end is an MLT frame and corresponding intra-frame correlation information and inter-frame correlation information of the frame. When transmitting, the MLT frame and related frame information are sent in packets.

步骤 604, 解码端对接收的 MLT帧进行错 /丢帧检测， 生成错 /丢帧 信息， 判断当前 MLT帧或前置 MLT 帧是否为错帧或丢帧。 如果当前 MLT帧为错 /丢帧， 或前置 MLT帧为错 /丢帧， 或当前 MLT帧和前置 MLT帧均为错丢帧， 则执行步骤 605; 否则， 执行步骤 610和 611 , 对 当前 MLT帧进行正常解码； 将当前 MLT帧经 IDCT变换后得到的重叠 相加时域帧 （如图 2中的 a, b, c... )和经 IMLT变换后得到的最终时 域帧 （如图 2中的 1 , 2, 3... )保存。 Step 604: The decoding end performs error/drop frame detection on the received MLT frame, generates error/drop frame information, and determines whether the current MLT frame or the pre-MLT frame is a wrong frame or a dropped frame. If the current MLT frame is an error/drop frame, or the pre-MLT frame is a wrong/drop frame, or the current MLT frame and the pre-MLT frame are both lost frames, step 605 is performed; otherwise, steps 610 and 611 are performed, The current MLT frame is normally decoded; the overlap-added time domain frame obtained by IDCT transformation of the current MLT frame (such as a, b, c... in FIG. 2) and the final time domain frame obtained by IMLT transformation ( Save as 1 , 2, 3... in Figure 2.

本步骤中， 错 /丢帧检测的目的是检查当前 MLT帧是否出现丢帧、 错帧的情况， 该检测结果用于确定后续对当前 MLT帧的操作。 该检测 结果还要被保存到下一 MLT帧到来，作为前置 MLT帧是否为错 /丢帧的 判断依据。 In this step, the purpose of the error/drop frame detection is to check whether the current MLT frame has a frame loss or a wrong frame. The detection result is used to determine the subsequent operation on the current MLT frame. The detection result is also saved to the next MLT frame, as a basis for determining whether the pre-MLT frame is a wrong/lost frame.

在实时传输中，丢帧检测功能是由实时传输协议（RTP, Real - Time Transport Protocol )完成的。 语音 /音频帧被打包传送， 每个 RTP包的包 头中都包含有一个序列号， 连续发送出去的包的序列号是连续的。 当解 码端收到一个包时会检查包头内的序列号是否与前一个包的序列号连 续， 如果中间有间隔则说明有包丢失了， 此时判定该帧为丢帧。 In real-time transmission, the frame loss detection function is implemented by Real-Time Transport Protocol (RTP). The voice/audio frames are packed and transmitted. Each RTP packet contains a serial number in the header, and the serial number of the continuously transmitted packets is continuous. When the decoder receives a packet, it checks whether the serial number in the packet header is continuous with the sequence number of the previous packet. If there is an interval in the middle, it indicates that the packet is lost. At this time, it is determined that the frame is a dropped frame.

错帧检测是检测帧内是否出现错误， 该功能可以采用前向纠错 ( FEC , Forward Error Correction )技术实现。 FEC可以应用在传输协议 中， 也可以集成在编码算法中。 在编码端向解码端传送的 MLT帧及其 相关帧信息后添加一些纠错比特， 纠错比特是经过特殊设计的， 在当前 传输的帧中出现错误时会在纠错比特中反映出来。 通过检查纠错比特可 以判定当前 MLT帧是否为错帧。 The error frame detection is to detect whether there is an error in the frame. This function can be implemented by FEC (Forward Error Correction) technology. FEC can be applied in the transport protocol or integrated in the encoding algorithm. After the MLT frame and its associated frame information transmitted by the encoding end to the decoding end, some error correction bits are added. The error correction bits are specially designed and reflected in the error correction bits when an error occurs in the currently transmitted frame. By checking the error correction bits, it can be determined whether the current MLT frame is a wrong frame.

步骤 605, 解码端根据错 /丢帧信息， 结合相关帧信息， 选择隐藏策 略。Step 605: The decoding end selects a hidden policy according to the error/lost frame information and the related frame information. slightly.

本步骤中， 错 /丢帧检测结果可以采用错 /丢帧错误标识表示。 该错 / 丢帧错误标识包括当前帧错误标识和前置帧错误标识。 当前帧错误标识 和前置帧错误标识都可以用 1或 0表示，因此错 /丢帧错误标识有几种组 合： 01 , 即前置 MLT帧为正常帧当前 MLT帧为错 /丢帧； 10, 即前置 MLT帧为错 /丢帧当前 MLT帧为正常帧； 11 ,即前置 MLT帧和当前 MLT 帧都为错 /丢帧。 In this step, the error/drop frame detection result may be represented by a wrong/lost frame error identifier. The error/drop frame error identifier includes a current frame error identifier and a pre-frame error identifier. Both the current frame error identifier and the pre-frame error identifier can be represented by 1 or 0. Therefore, there are several combinations of error/lost frame error identifiers: 01, that is, the pre-MLT frame is the normal frame, and the current MLT frame is the wrong/drop frame; That is, the pre-MLT frame is the wrong/lost frame, and the current MLT frame is the normal frame; 11 , that is, the pre-MLT frame and the current MLT frame are both wrong/lost frames.

解码端每进行一次隐藏策略选择， 便生成和更新一次前帧可靠性标 识。 以便在连续丢帧的情况下， 确定隐藏策略。 该前帧可靠性标识表示 根据本次选择的隐藏策略所恢复的错 /丢帧是否可靠。也可以在每次恢复 错 /丢帧后确定前帧可靠性标识。 Each time the decoder selects a hidden strategy, it generates and updates a previous frame reliability identifier. In order to continuously drop frames, determine the hidden strategy. The previous frame reliability indicator indicates whether the error/drop frame recovered according to the hidden policy selected this time is reliable. The previous frame reliability indicator can also be determined each time the error/drop frame is recovered.

在错 /丢帧错误标识为 11时， 根据前帧可靠性标识， 结合当前 MLT 帧相关帧信息， 确定隐藏策略。 确定隐藏策略采用逻辑判断的方法。 When the error/lost frame error flag is 11, the hidden policy is determined according to the previous frame reliability identifier and the current MLT frame related frame information. Determine the way in which hidden strategies use logical judgment.

在错 /丢帧错误标识为 01时， 根据当前 MLT帧相关帧信息， 查预置 在编码端的隐藏策略表， 确定隐藏策略。 该隐藏策略表存储了各种相关 帧信息对应的隐藏策略。 When the error/drop frame error flag is 01, according to the current MLT frame related frame information, the hidden strategy table preset on the encoding end is checked to determine the hidden policy. The hidden policy table stores hidden policies corresponding to various related frame information.

在错 /丢帧错误标识为 10时， 根据当前 MLT帧相关帧信息和前置 MLT帧相关帧信息， 查预置在编码端的隐藏策略表， 确定隐藏策略。 When the error/lost frame error flag is 10, according to the current MLT frame related frame information and the pre-MLT frame related frame information, the hidden strategy table preset at the encoding end is checked to determine the hidden policy.

步骤 606, 解码端根据选择的隐藏策略和解码波形历史緩沖中保存 的历史时域波形信息恢复出错 /丢帧。 Step 606: The decoding end recovers the error/drop frame according to the selected hidden policy and the historical time domain waveform information saved in the decoded waveform history buffer.

步骤 607, 解码端将恢复出的错 /丢帧存入解码波形历史緩沖。 Step 607: The decoding end stores the recovered error/lost frame into the decoding waveform history buffer.

以上就实现了丢帧隐藏。 The above achieves frame loss hiding.

其中， 步骤 605解码端根据错 /丢帧检测结果， 结合相关帧信息， 选 择隐藏策略是上述方法实现的重要步骤。 下面对该步骤进行详细说明。 Step 605: The decoding end selects the hidden strategy according to the error/drop frame detection result and the related frame information, which is an important step implemented by the foregoing method. This step is described in detail below.

图 7为本发明实施例解码端选择隐藏策略的方法流程图。 该方法包 括以下步骤：FIG. 7 is a flowchart of a method for selecting a hiding policy by a decoding end according to an embodiment of the present invention. Method package Including the following steps:

步骤 6051 ,判断前置 MLT帧和当前 MLT帧是否都为错 /丢帧；如果 是则执行步骤 6055 , 根据前帧可靠性标识， 结合当前 MLT帧的相关帧 信息， 为当前帧确定隐藏策略； 否则执行步骤 6052。 Step 6051, determining whether the pre-MLT frame and the current MLT frame are both wrong/lost frames; if yes, executing step 6055, determining a hidden policy for the current frame according to the previous frame reliability identifier and the associated frame information of the current MLT frame; Otherwise, step 6052 is performed.

本步骤中， 根据前置帧错误标识判断前置 MLT帧是否为错 /丢帧， 并根据当前帧错误标识判断当前 MLT帧是否为错 /丢帧。在前置 MLT帧 和当前 MLT帧都为错 /丢帧， 也就是出现连续错 /丢帧的情况时， 执行步 骤 6055, 根据前帧可靠性标识和当前帧相关帧信息， 采用逻辑推理的方 法， 为当前 MLT帧选择隐藏策略。 然后再执行图 6中的步骤 606。 In this step, it is determined whether the pre-MLT frame is an error/drop frame according to the pre-frame error identifier, and whether the current MLT frame is a wrong/lost frame according to the current frame error identifier. When both the pre-MLT frame and the current MLT frame are in the wrong/drop frame, that is, in the case of continuous error/drop frame, step 6055 is performed, and the logical reasoning method is adopted according to the previous frame reliability identifier and the current frame-related frame information. , Select a hidden policy for the current MLT frame. Step 606 in Figure 6 is then performed.

步骤 6052, 判断当前 MLT帧是否有错； 如果是则执行步骤 6056, 根据当前 MLT帧相关帧信息， 查预置在编码端的隐藏策略表， 确定隐 藏策略。 否则执行步骤 6053。 步骤 6056的隐藏策略确定方法， 属于前 置 MLT帧为正常帧但当前 MLT帧为错 /丢帧的情况。 Step 6052: Determine whether the current MLT frame is faulty. If yes, go to step 6056 to check the hidden strategy table preset on the encoding end according to the current MLT frame related frame information, and determine the hidden policy. Otherwise, go to step 6053. The hidden policy determining method of step 6056 belongs to the case where the pre-MLT frame is a normal frame but the current MLT frame is a wrong/lost frame.

以下属于前置 MLT帧为丢 /错帧但当前 MLT帧为正常帧的情况。 虽 然当前 MLT为正常帧， 但由于 IMLT变换用到了前置 MLT帧的重叠相 加时域帧信息， 因此也需要为当前 MLT帧选择隐藏策略。 The following is the case where the pre-MLT frame is a lost/error frame but the current MLT frame is a normal frame. Although the current MLT is a normal frame, since the IMLT transform uses the overlap-added time domain frame information of the pre-MLT frame, it is also necessary to select a hidden strategy for the current MLT frame.

步骤 6053 , 当前 MLT帧经 IDCT变换得到当前 MLT帧的重叠相加 时域帧。 Step 6053: The current MLT frame is subjected to IDCT transformation to obtain an overlap-added time domain frame of the current MLT frame.

本步骤中获得的当前 MLT 帧的重叠相加时域帧作为历史时域波形 的一部分， 用于根据隐藏策略和历史时域波形恢复错 /丢帧。 The overlap-added time domain frame of the current MLT frame obtained in this step is used as part of the historical time domain waveform to recover the error/drop frame based on the hidden strategy and the historical time domain waveform.

步骤 6054,根据当前 MLT帧相关帧信息和前置 MLT帧相关帧信息， 查预置在编码端的隐藏策略表， 确定隐藏策略。 Step 6054: According to the current MLT frame related frame information and the pre-MLT frame related frame information, check a hidden policy table preset at the encoding end to determine a hidden policy.

隐藏策略是预先设计并存放在解码端的。 对于隐藏策略选择过程中 的对各种情况判断的顺序， 不限于以上所述的顺序， 只要将各种情况区 分开即可， 并不影响判断结果。 为了便于说明隐藏策略的选取和根据不同的隐藏策略对丢帧 /错帧 进行恢复， 以下举例和附表进行说明。The hidden strategy is pre-designed and stored on the decoder side. The order of judging various situations in the hidden policy selection process is not limited to the order described above, and it is only necessary to distinguish the various cases without affecting the judgment result. In order to facilitate the selection of the hidden policy and the recovery of the lost frame/error frame according to different hidden policies, the following examples and attached tables are described.

图 8为本发明实施例隐藏策略原理示意图。 重叠相加时域波形为解 码端将 MLT帧经 IDCT变换得到的时域波形，最终时域波形为重叠相加 时域波形经逆向重叠相加后的时域波形。 FIG. 8 is a schematic diagram of a principle of a hidden policy according to an embodiment of the present invention The overlap-added time-domain waveform is a time-domain waveform obtained by IDCT transformation of the MLT frame by the decoding end, and the final time-domain waveform is the time-domain waveform after the overlap-added time-domain waveform is inversely overlapped.

由 MLT的重叠相加机制可知， 当错 /丢帧出现时， 最终时域波形的 当前帧和后续帧都受到了影响。 例如， 丢失图 8中所示的第 2帧， 即（、 d帧。 由于 3、 4的计算要依赖于 b, c; 4、 5的计算都要依赖于 d、 e, 因此最终时域波形中 3、 4、 5、 6都会因为（、 d的错 /丢而受到影响。 According to the overlapping and adding mechanism of MLT, when the error/drop frame occurs, the current frame and subsequent frames of the final time domain waveform are affected. For example, the second frame shown in Figure 8 is lost, that is, (, d frame. Since the calculation of 3, 4 depends on b, c; the calculation of 4, 5 depends on d, e, so the final time domain waveform In the 3, 4, 5, and 6 will be affected by (d, wrong / lost.

表 2为帧内、帧间相关性的所有可能组合所对应的丢帧隐藏策略表。 其中， "―"表示由前面紧邻的基音信息复制得到， "→ "表示由后面紧邻 的基音信息复制得到， "V"表示由重叠相加时域帧（即 a, b, c, d, e.. . ) 计算得到， "x"表示无法近似的恢复，只能依靠重复相邻信号的基音并将 其衰减来隐藏。 复制基音时， 可以采用一个基音复制得到整个半帧的波 形， 用于复制的基音根据 "―"或者"→ "从前或者从后选取。 Table 2 is a frame loss concealment strategy table corresponding to all possible combinations of intra-frame and inter-frame correlation. Where "―" means that the pitch information is immediately copied, "→" means that it is copied by the immediately adjacent pitch information, and "V" means that the time domain frame is overlapped (ie, a, b, c, d, e .. . ) Calculated, "x" means unrecoverable recovery, and can only be hidden by repeating the pitch of adjacent signals and attenuating them. When copying a pitch, you can use a pitch copy to get the waveform of the entire field. The pitch used for copying is selected from the front or back according to "-" or "→".

序 帧 3和 4相 帧 5和 6相 恢复的帧 恢复的帧 关帧信息 关帧信息 3、和 4、 5、和 6、 Sequence frame 3 and 4 phase frame 5 and 6 phase recovered frame recovered frame off frame information off frame information 3, and 4, 5, and 6,

帧 帧 帧 帧 4、 5、 内 间 内 间 Frame frame frame frame 4, 5, inner space

1 1¹ 1 11 1¹ 1 1

2 1¹ 1 02 1¹ 1 0

3 1¹ 0 13 1¹ 0 1

4 0¹ 1 1 ―4 0¹ 1 1 ―

5 1¹ 0 0 ― X5 1¹ 0 0 ― X

6 0¹ 0 1 7 0 1 1 0 →6 0¹ 0 1 7 0 1 1 0 →

8 0 1 0 0 X 8 0 1 0 0 X

9 1 0 1 1 X X 9 1 0 1 1 X X

10 0 0 1 1 X X → 10 0 0 1 1 X X →

11 0 0 1 0 X X → 11 0 0 1 0 X X →

12 1 0 0 1 X X X 12 1 0 0 1 X X X

13 1 0 1 0 X X → 13 1 0 1 0 X X →

14 1 0 0 0 X X X 14 1 0 0 0 X X X

15 1 0 0 1 X X X 15 1 0 0 1 X X X

16 0 0 0 0 X X X 16 0 0 0 0 X X X

表 2 Table 2

1) 前置 MLT帧为正常帧， 当前 MLT帧为错 /丢帧 1) The pre-MLT frame is a normal frame, and the current MLT frame is a wrong/lost frame.

在这种情况下， 帧 a和 b为正常帧， 帧 c和 d为错 /丢帧。 采用表 2 中隐藏策略， 根据帧 3和 4的相关帧信息恢复出帧 3、和 4、。 In this case, frames a and b are normal frames, and frames c and d are wrong/drop frames. Using the hidden strategy in Table 2, frames 3, and 4 are recovered based on the associated frame information of frames 3 and 4.

在当前 MLT帧的帧间相关性为强时，由于前置帧或前置帧的末尾具 有与当前帧相似的波形特征， 因此重复前置 MLT帧最终时域帧末尾处 的基音信息或者重复前置 MLT帧最终时域帧的后半帧可以大致获得一 个当前 MLT帧最终时域帧的前半帧。 When the inter-frame correlation of the current MLT frame is strong, since the end of the pre-frame or the pre-frame has a waveform feature similar to the current frame, repeating the pitch information at the end of the final time-domain frame of the pre-MLT frame or before repeating Setting the last half of the final time domain frame of the MLT frame can roughly obtain the first half of the final time domain frame of the current MLT frame.

对于当前 MLT帧最终时域帧的后半帧，该隐藏策略利用音频帧的时 域信号与其经 MLT重叠相加后的信号的代数关系， 通过估计出的当前 MLT帧最终时域帧的前半帧来求得剩下部分的时域音频信号。其具体实 现过程为： For the second half of the final time domain frame of the current MLT frame, the hidden strategy utilizes the algebraic relationship of the time domain signal of the audio frame with the signal added by the MLT overlap, and estimates the first half of the final time domain frame of the current MLT frame. To find the rest of the time domain audio signal. The specific implementation process is:

根据正向 MLT变换有 According to the forward MLT transform

v(n + N/2) = w[(N - 1) - n]x(N + n) _ w(n)x[(2N - l) - n] 0<n<N/2 v(n + N/2) = w[(N - 1) - n]x(N + n) _ w(n)x[(2N - l) - n] 0<n<N/2

将上式左右移项可得_r,_{M n} w[(N - l) - n] - x(N + n) - v(n + N/2)Move the above formula to the left and right_r ,_{M n} w[(N - l) - n] - x(N + n) - v(n + N/2)

x[(2N - 1) - n] = x[(2N - 1) - n] =

w(ⁿ) 0<n<N/2w(ⁿ ) 0<n<N/2

因此在本例中， x[(2N-l)-n]即为 4, x(N+n)即为 3, v(n+N/2)即为 b。 可见 3、 4和 b有唯一的确定关系。 Therefore, in this example, x[(2N-l)-n] is 4, x(N+n) is 3, and v(n+N/2) is b. It can be seen that 3, 4 and b have a unique relationship.

当 3的帧间相关性为强时， 重复 2的末尾处的基音获得 3、； 由于 3、 4和 b的唯一确定性关系， 采用 3、代替 3, 由 3、和 b可计算得到一个 4 的估计 4、， 4即可被大致恢复出来。为了确保 2和 3、之间的相位连续性， 在重复 2的基音时， 应从 2中选取相位与 2的末尾处相位相同的一点， 作为重复基音的起点。 这样使得 3、开始的相位与 2末尾的相位相同， 实 现了 2与 3、的相位连续。 一种可能的基音选取方法为： 由 2的末尾开始 向前选取一段基音的长度， 将此段作为重复的基音。 表 1中"—，，所示的 向前复制基音， 均可以采用这种选择基音的原理， 但不限于此。 When the inter-frame correlation of 3 is strong, the pitch at the end of repetition 2 is obtained by 3; due to the unique deterministic relationship of 3, 4, and b, 3 is used instead of 3, and 3 is calculated from 3, and b. The estimate of 4, 4 can be roughly restored. In order to ensure the phase continuity between 2 and 3, when repeating the pitch of 2, the same phase as the phase at the end of 2 should be selected from 2 as the starting point of the repeated pitch. This makes 3, the starting phase is the same as the phase at the end of 2, and the phase of 2 and 3 is continuous. A possible pitch selection method is: Select the length of a pitch from the end of 2 and use this segment as a repeated pitch. The principle of selecting a pitch can be adopted for the forward copying pitch shown in "-," in Table 1, but is not limited thereto.

此时，无论 3的帧间相关性是强还是弱都可以利用 3、和 b计算出 4、。 但由于利用 3、， 4和 b的关系计算得出 4、需要额外的增加计算量， 从筒 化算法降低成本的角度考虑， 增加额外的计算量是没有必要的。 因此在 帧内帧间都为强相关性时 3、、 4、用 2重复就可以了， 如表 2中序号为 1、 2、 3和 5的隐藏策略； 而在帧间相关性强而帧内相关性弱时才使用计算 的方法得到 4、， 3、仍采用重复 2的方法， 如表 2中序号为 4、 6、 7和 8 的隐藏策略。 At this time, regardless of whether the inter-frame correlation of 3 is strong or weak, 3, and b can be used to calculate 4 . However, due to the relationship between 3, 4 and b, 4 is needed to increase the amount of calculation. From the perspective of reducing the cost of the binning algorithm, it is not necessary to increase the additional calculation. Therefore, when there is strong correlation between intra frames, 3, 4, and 2 repetitions are possible, as in Table 2, the hidden strategies with numbers 1, 2, 3, and 5; and the correlation between frames is strong. When the internal correlation is weak, the calculation method is used to obtain 4, 3. The method of repeating 2 is still adopted, such as the hidden strategies of the numbers 4, 6, 7, and 8 in Table 2.

在当前 MLT帧的帧间相关性为弱时， 当前 MLT帧最终时域帧的前 半帧和后半帧都通过重复前置 MLT帧最终时域帧的后半帧的基音并进 行全帧衰减得到。 衰减强度从前半帧到后半帧逐渐增大。 衰减方法为将 预衰减的帧乘以衰减因子。该衰减因子表示衰减的强度，衰减因子越大， 表示衰减强度越大。 When the inter-frame correlation of the current MLT frame is weak, the first half frame and the second half frame of the final time domain frame of the current MLT frame are repeated by repeating the pitch of the second half of the final time domain frame of the pre-MLT frame and performing full frame attenuation. . The attenuation intensity gradually increases from the first half frame to the second half frame. The attenuation method is to multiply the pre-attenuated frame by the attenuation factor. The attenuation factor represents the intensity of the attenuation, and the larger the attenuation factor, the greater the attenuation intensity.

因此在本例中， 当 3的帧间相关性为弱时， 3、和 4、通过重复 2的基 音得到， 并将得到的 3、和 4、进行全帧衰减， 衰减强度从 3、的前端到 4、 的后端逐渐增强。如表 2中序号为 9-16的隐藏策略。 由于采用了衰减的 处理， 因此此时获得的恢复帧被认为不可靠。So in this example, when the inter-frame correlation of 3 is weak, 3, and 4, by repeating the basis of 2 The sound is obtained, and the obtained 3, and 4 are attenuated in full frame, and the attenuation intensity is gradually increased from the front end of 3 to the rear end of 4. As shown in Table 2, the hidden strategy with the number 9-16. The recovery frame obtained at this time is considered to be unreliable due to the processing using attenuation.

2) 前置 MLT帧为错 /丢帧， 当前 MLT帧为正常帧的情况 2) The pre-MLT frame is a wrong/lost frame, and the current MLT frame is a normal frame.

在这种情况下， 帧 c和 d为错 /丢帧， 帧 e和 f为正常帧。 采用表 2 中的隐藏策略， 根据帧 3和 4的相关帧信息、 帧 5和 6的相关帧信息， 恢复出帧 5、和 6、。 In this case, frames c and d are wrong/drop frames, and frames e and f are normal frames. Using the hidden strategy in Table 2, frames 5, and 6, are recovered based on the associated frame information of frames 3 and 4, and the associated frame information of frames 5 and 6.

2a) 当 5的帧间相关性为弱时， 5与前置帧的后半帧不相似。 5由 6 决定。 6的确定原理为： 由于窗口函数的加权作用， d和 e对 6的贡献是 不同的， e的作用要远大于 d。 因此， 即使将 d以其它的半帧代替， 由 e 的代替帧计算求出的 6、都与原波形是较为相似的， 尤其是 6、的后半帧波 形。 代替帧可以选用 f或者 b。 例如， 筒单的用 f代替 d, 根据 f和 e经 重叠相加后得到 6、。 2a) When the inter-frame correlation of 5 is weak, 5 is not similar to the second half of the pre-frame. 5 is determined by 6. The principle of determination of 6 is: Due to the weighting effect of the window function, the contribution of d and e to 6 is different, and the effect of e is much larger than d. Therefore, even if d is replaced by another field, the 6 calculated by the substitute frame of e is similar to the original waveform, especially the second half of the waveform. Instead of frame, you can use f or b. For example, instead of d, the cartridge is replaced by f, and the overlap is obtained according to f and e.

5、可以采用复制 6、基音的方法获得， 当 5的帧内相关性为强时， 不 对 5、进行衰减， 直接复制 6、的基音作为 5、 如表 2中序号为 2、 7、 11 和 13的恢复帧 5、和 6、的隐藏策略。 一种可能的基音选取方法为： 由 6、 的前端开始向后取一段基音的长度，将此段作为重复的基音。表 1中"→" 所示的向后复制基音， 均可以采用这种选择基音的原理， 但不限于此。 5, can be obtained by copying 6, pitch method, when the intra-frame correlation of 5 is strong, not attenuating 5, directly copying 6, the pitch as 5, as in Table 2, the serial numbers 2, 7, 11 and 13 hidden frames 5, and 6, hidden strategy. A possible pitch selection method is: taking the length of a pitch starting from the front end of 6, and using this segment as a repeated pitch. The principle of selecting a pitch can be adopted for the backward reproduction of the pitch shown by "→" in Table 1, but is not limited thereto.

当 5的帧间相关性为弱， 帧内相关性为弱时， 也采用复制 6、的基音 作为 5、但此时要对 5、进行衰减，衰减强度由 5、的末尾向前端逐渐增大， 如表 2中序号为 5、 8、 14和 16的恢复帧 5、和 6、的隐藏策略。 When the inter-frame correlation of 5 is weak and the intra-frame correlation is weak, the pitch of copy 6 is also used as 5, but at this time, 5 is attenuated, and the attenuation intensity is gradually increased from the end of 5 to the front end. , as shown in Table 2, the recovery strategies of the recovery frames 5, and 6, with serial numbers 5, 8, 14, and 16.

2b) 当 5的帧间相关性为强时， 如果 3的帧间相关性为强且 5的帧 内相关性为强， 则 5、与 4、相似， 且 5、和 6、相似。 因此 5、和 6、均由基音 4、的基音复制得到。如表 2中序号为 1和 4的恢复帧 5、和 6、的隐藏策略。 2b) When the inter-frame correlation of 5 is strong, if the inter-frame correlation of 3 is strong and the intra-frame correlation of 5 is strong, then 5, and 4 are similar, and 5, and 6, are similar. Therefore, 5, and 6, are all reproduced from the pitch of the pitch 4. The hidden strategies of recovery frames 5, and 6, as numbered 1 and 4 in Table 2.

当 5的帧间相关性为强， 如果 3的帧间相关性为强， 但 5的帧内相 关性为弱， 则 5、与 4、相似， 且 6、与 5、不相似。 因此 5、由 4、的基音复制 得到， 6、根据 f和 e经重叠相加后得到。 如表 2中序号为 3和 6的恢复 帧 5、和 6、的隐藏策略。 当然， 前述当 5的帧间相关性为强， 3的帧间相 关性为强且 5的帧内相关性为强时， 6、也可以采用计算得到， 但是为了 减少计算， 利用复制 4、即可得到 ^艮好的隐藏效果。When the inter-frame correlation of 5 is strong, if the inter-frame correlation of 3 is strong, but the intra-frame phase of 5 The relationship is weak, then 5, and 4 are similar, and 6, and 5 are not similar. Therefore, 5 is obtained by copying the pitch of 4, and 6 is obtained by overlapping and adding f and e. The hidden strategies of recovery frames 5, and 6, as numbered 3 and 6, in Table 2. Of course, when the inter-frame correlation of 5 is strong, the inter-frame correlation of 3 is strong, and the intra-frame correlation of 5 is strong, 6, it can also be calculated, but in order to reduce the calculation, copy 4 is used. You can get a good hidden effect.

当 3的帧间相关性为弱时， 无法利用 4、。 6、采用 f和 e经重叠相加 后得到。 When the inter-frame correlation of 3 is weak, 4 is not available. 6. Use f and e to add after overlapping.

因此， 当 5的帧间相关性为强、 3的帧间相关性为弱且 5的帧内相 关性为强， 则 5、由 6、的基音复制得到。 如表 2中序号为 9和 10的恢复 帧 5、和 6、的隐藏策略。 Therefore, when the inter-frame correlation of 5 is strong, the inter-frame correlation of 3 is weak, and the intra-frame correlation of 5 is strong, then 5. The pitch is reproduced by 6. The hidden strategies of recovery frames 5 and 6, as numbered 9 and 10 in Table 2.

当 5的帧间相关性为强、 3的帧间相关性为弱且 5的帧内相关性为 弱， 则 6、仍采用 f和 e经重叠相加后得到， 5、由 6、的基音复制并经衰减 得到， 衰减强度由 5、的末尾向前端逐渐增大。 如表 2中序号为 12和 15 的恢复帧 5、和 6、的隐藏策略。 When the inter-frame correlation of 5 is strong, the inter-frame correlation of 3 is weak, and the intra-frame correlation of 5 is weak, then 6, the f and e are still added by overlapping, and 5, the pitch of 6, It is reproduced and attenuated, and the attenuation intensity gradually increases from the end of the 5th to the front end. Hidden strategies for recovery frames 5 and 6, as numbered 12 and 15 in Table 2.

以上， 在计算 6、时， 还可以根据 b和 e计算 6、。 计算出的 6、同样与 原波形较为相似。 但是利用 b , 要求保存前置帧之前的一帧时域波形。 因此本实施例利用 f和 e计算 6、。 Above, when calculating 6, it can also be calculated according to b and e. The calculated 6. is also similar to the original waveform. But with b, it is required to save a frame time domain waveform before the pre-frame. Therefore, this embodiment uses f and e to calculate 6.

3) 前置 MLT帧有为错 /丢帧， 当前 MLT帧为错 /丢帧的情况。 3) The pre-MLT frame has error/lost frame, and the current MLT frame is wrong/drop frame.

在这种情况下， 即为出现连续丢帧 /错帧的情况。 丢帧序列的开始端 可根据第 1)种隐藏方法恢复。 对于中间的帧采用下述隐藏方法恢复。 In this case, it is the case that consecutive frame/error frames occur. The beginning of the frame loss sequence can be recovered according to the 1) hidden method. For the intermediate frame, use the following hidden method to recover.

由于每次进行隐藏策略的选择后， 都要更新前帧可靠性标识。 总的 来说，采用重复某一帧的波形，并进行衰减的恢复帧被认为是不可靠的， 前帧可靠性标识设置为不可靠； 采用公式计算得到， 或者采用重复某一 帧的波形， 并且不进行衰减的恢复帧认为是可靠的， 此时前帧可靠性标 识设置为可靠。 恢复连续丢帧 /错帧需要利用前帧可靠性标识、 前置 MLT帧最终时 域帧和当前 MLT帧的相关帧信息。 图 9为本发明实施例连续丢帧 /错帧 的隐藏策略选择方法流程图。 该方法的具体步骤如下：The front frame reliability indicator is updated each time the hidden policy is selected. In general, it is considered unreliable to use a waveform that repeats a certain frame and attenuated, and the front frame reliability flag is set to be unreliable; it is calculated by a formula, or a waveform of a certain frame is repeated. And the recovery frame that is not attenuated is considered to be reliable, and the front frame reliability flag is set to be reliable at this time. Restoring consecutive frame loss/error frames requires the use of the previous frame reliability indicator, the pre-MLT frame final time domain frame, and the associated frame information of the current MLT frame. FIG. 9 is a flowchart of a method for selecting a hidden frame of a continuous frame loss/error frame according to an embodiment of the present invention. The specific steps of the method are as follows:

步骤 801 , 判断前置帧是否可靠， 如果是则执行步骤 802; 否则执行 步骤 811。 Step 801: Determine whether the pre-frame is reliable. If yes, go to step 802; otherwise, go to step 811.

本步骤中，根据前帧可靠性标识判断前置 MLT帧的最终时域帧是否 可靠， 如果不可靠则执行步骤 811 , 当前 MLT帧最终时域帧重复前置帧 的最终时域帧， 并对当前 MLT帧最终时域帧进行全帧衰减。 以帧 5和 6 为例， 帧 5、和 6、由重复帧 3、和 4、得来， 全帧衰减时， 衰减强度由 5的前 端向 6的末尾逐渐增大。且获得的帧 5、和 6、的可靠性信息表示为不可靠。 由于前置帧为不可靠帧， 因此可以将 3、和 4、波形原样复制过来作为 5、 和 6、。 In this step, it is determined whether the final time domain frame of the pre- MLT frame is reliable according to the previous frame reliability identifier, and if not, step 811 is performed, and the final time domain frame of the current MLT frame repeats the final time domain frame of the pre-frame, and The final time domain frame of the current MLT frame is subjected to full frame attenuation. Taking frames 5 and 6 as an example, frames 5, and 6, from repeated frames 3, and 4, when the full frame is attenuated, the attenuation strength gradually increases from the front end of 5 to the end of 6. And the reliability information of the obtained frames 5, and 6, is expressed as unreliable. Since the pre-frame is an unreliable frame, the 3, and 4 waveforms can be copied as they are 5, and 6.

步骤 802, 判断帧间相关性是否为强； 是则执行步骤 803, 否则执行 步骤 812。 Step 802: Determine whether the inter-frame correlation is strong; if yes, go to step 803, otherwise go to step 812.

本步骤中， 如果判断帧间相关性为弱， 则不管帧内相关性是什么， 都执行步骤 812, 重复前置 MLT帧最终时域帧作为当前 MLT帧最终时 域帧， 并对得到的当前帧 MLT最终时域帧进行全帧衰减， 衰减强度由 前到后逐渐增大。 且获得的当前 MLT帧最终时域帧的可靠性信息标识 为不可靠。 由于帧间相关性为弱， 当前帧和帧 3、和 4、不相像， 因此可以 将 3、和 4、波形原样复制过来作为 5、和 6、。 In this step, if it is determined that the inter-frame correlation is weak, step 812 is performed regardless of the intra-frame correlation, and the final time-domain frame of the pre-existing MLT frame is repeated as the final time-domain frame of the current MLT frame, and the current current obtained is obtained. The frame MLT final time domain frame performs full frame attenuation, and the attenuation intensity gradually increases from front to back. And the reliability information of the final time domain frame of the current MLT frame obtained is not reliable. Since the inter-frame correlation is weak, the current frame and frame 3, and 4 are not similar, so 3, and 4, the waveform can be copied as the 5, and 6, respectively.

步骤 803, 判断帧内相关性是否为强； 是则执行步骤 804; 否则执行 步骤 813。 Step 803: Determine whether the intra-frame correlation is strong; if yes, go to step 804; otherwise, go to step 813.

本步骤中， 如果判断帧内相关性为弱， 则属于帧间相关性强且帧内 相关性弱的情况。 执行步骤 813 , 重复前置帧的最终时域帧的后半帧作 为当前 MLT帧的前半帧和后半帧， 得到的当前 MLT帧最终时域帧的前 半帧不衰减， 对后半帧进行衰减， 衰减强度由后半帧的前端向后半帧的 末尾逐渐增大。 且获得的当前 MLT帧最终时域帧的可靠性信息标识为 不可靠。In this step, if it is determined that the intra-frame correlation is weak, it is a case where the inter-frame correlation is strong and the intra-frame correlation is weak. Step 813 is performed to repeat the second half of the final time domain frame of the preamble frame as the first half frame and the second half frame of the current MLT frame, and obtain the current MLT frame before the final time domain frame. The half frame is not attenuated, and the latter half frame is attenuated, and the attenuation intensity is gradually increased from the front end of the second half frame to the end of the second half frame. And the reliability information of the final time domain frame of the current MLT frame obtained is identified as unreliable.

步骤 804, 重复前置 MLT帧最终时域帧的后半帧作为当前 MLT帧 的前半帧和后半帧， 不衰减。 且获得的当前 MLT帧最终时域帧的可靠 性信息标识为可靠。 Step 804, repeating the last half frame of the final time domain frame of the pre-MLT frame as the first half frame and the second half frame of the current MLT frame, without attenuation. And the reliability information of the final time domain frame of the current MLT frame obtained is identified as reliable.

在实际的音频信号中，相关帧信息的不同组合出现的概率是不同的， 实际情况下大多数的丢帧都可以被较好的隐藏。 In the actual audio signal, the probability of occurrence of different combinations of related frame information is different, and in reality, most of the dropped frames can be better hidden.

为实现本发明实施例的丢帧隐藏方法， 本发明实施例提供了丢帧隐 藏系统。 To implement the frame dropping concealment method in the embodiment of the present invention, the embodiment of the present invention provides a frame dropping hidden system.

图 10为本发明实施例丢帧隐藏系统组成框图，该系统包括编码装置 和解码装置。 编码装置将获得的相关帧信息以及当前 MLT帧发送给解 码装置。 解码装置对当前 MLT帧进行错 /丢帧检测， 生成错 /丢帧信息； 在进行丢帧隐藏处理时， 解码装置根据错 /丢帧信息， 结合相关帧信息和 历史时域波形恢复出错 /丢帧。 FIG. 10 is a block diagram of a frame dropping system according to an embodiment of the present invention, the system comprising an encoding device and a decoding device. The encoding device transmits the obtained related frame information and the current MLT frame to the decoding device. The decoding device performs error/drop frame detection on the current MLT frame to generate error/drop frame information. When performing frame loss concealment processing, the decoding device recovers the error/loss according to the error/lost frame information, the associated frame information and the historical time domain waveform. frame.

本实施例中的编码装置，采用图 11所示的本发明实施例提供的丢帧 隐藏编码装置的第一较佳实施例。 如图 11所示， 该编码装置包括 MLT 编码单元 110、 历史緩沖单元 120、 相关帧信息生成单元 130。 The coding apparatus in this embodiment adopts the first preferred embodiment of the frame loss concealment coding apparatus provided in the embodiment of the present invention shown in FIG. As shown in Fig. 11, the encoding apparatus includes an MLT encoding unit 110, a history buffer unit 120, and a correlation frame information generating unit 130.

其中， 历史緩沖单元 120, 用于緩沖当前帧的时域波形和前置帧的 时域波形。 历史緩沖单元 120包括当前帧历史緩沖 121和前置帧历史緩 沖 122。 The history buffer unit 120 is configured to buffer the time domain waveform of the current frame and the time domain waveform of the preamble frame. The history buffer unit 120 includes a current frame history buffer 121 and a pre-frame history buffer 122.

当前帧历史緩沖 121 , 用于緩沖当前帧的时域波形。 编码装置完成 MLT编码并输出后，该当前帧历史緩沖 121利用其保存的当前帧更新前 置帧历史緩沖 122, 并将后续帧保存在当前帧历史緩沖 121中。 The current frame history buffer 121 is used to buffer the time domain waveform of the current frame. After the encoding apparatus completes the MLT encoding and outputs, the current frame history buffer 121 updates the pre-frame history buffer 122 with its saved current frame, and saves the subsequent frame in the current frame history buffer 121.

前置帧历史緩沖 122, 用于緩沖前置帧的时域波形。 MLT编码单元 110, 用于根据编码装置获取的后续帧和从历史緩沖 单元 120中获取的当前帧进行 MLT变换， 生成 MLT帧， 并发送出去。The pre-frame history buffer 122 is used to buffer the time domain waveform of the pre-frame. The MLT encoding unit 110 is configured to perform an MLT transformation according to a subsequent frame acquired by the encoding device and a current frame acquired from the history buffer unit 120, generate an MLT frame, and send the MLT frame.

相关帧信息生成单元 130, 用于根据该编码装置获取的后续帧， 以 及从历史单元緩沖获取的当前帧和前置帧， 生成相关帧信息， 并发送出 去。 相关帧信息生成单元 130包括帧内相关性检测器 131和帧间相关性 检测器 132。 帧内相关性检测器 131 , 用于根据从当前帧历史緩沖 121 中获取的当前帧生成帧内相关性信息， 发送出去。 The related frame information generating unit 130 is configured to generate related frame information according to the subsequent frame acquired by the encoding device and the current frame and the pre-frame obtained from the history unit buffer, and send the related frame information. The related frame information generating unit 130 includes an intra correlation detector 131 and an inter correlation detector 132. The intra-frame correlation detector 131 is configured to generate intra-frame correlation information according to the current frame acquired from the current frame history buffer 121, and send the information.

帧间相关性检测器 132, 用于根据从当前帧历史緩沖 121获取的当 前帧和从前置帧历史緩沖 122获取的前置帧， 生成帧间相关性信息， 发 送出去。 The inter-frame correlation detector 132 is configured to generate inter-frame correlation information according to the current frame acquired from the current frame history buffer 121 and the pre-frame obtained from the pre-frame history buffer 122, and send the inter-frame correlation information.

以上所说的发送不是指直接的打包发送出去。 而是将各个相关帧信 息都发送到一个多路复用器， 再发送出去。 The above mentioned transmission does not mean direct packaging and sending out. Instead, each relevant frame information is sent to a multiplexer and sent out.

在发送 MLT帧、帧内相关性信息和帧间相关性信息时，需要将 MLT 帧及其相关帧信息分包发送， 或者将相关帧信息嵌入当前帧的前置帧或 者后续帧发送， 也可以在一个可分层的编码体系中在增强层中传输相关 帧信息， 在核心层传输 MLT帧。 When transmitting the MLT frame, the intra-frame correlation information, and the inter-frame correlation information, the MLT frame and its associated frame information need to be sent in packets, or the related frame information may be embedded in the pre-frame or the subsequent frame of the current frame, or The associated frame information is transmitted in the enhancement layer in a layered coding scheme, and the MLT frame is transmitted at the core layer.

当嵌入前置帧发送时， 需要增加一个 MLT 帧緩沖单元， 用于緩存 MLT编码单元生成的 MLT帧， 延时发送出去。 因此当前 MLT帧发送是 携带的是其后续帧的相关帧信息。 When the embedded preamble frame is sent, an MLT frame buffer unit needs to be added to buffer the MLT frame generated by the MLT coding unit, and the delay is sent out. Therefore, the current MLT frame transmission carries the relevant frame information of its subsequent frames.

当嵌入后续帧发送时， 需要增加一个相关帧信息緩沖单元， 用于緩 存相关帧信息生成单元生成的相关帧信息， 延时发送出去。 因此当前 MLT帧发送是携带的是其前置帧的相关帧信息。 When embedding the subsequent frame transmission, an associated frame information buffer unit is needed to buffer the relevant frame information generated by the related frame information generating unit, and the delay is sent out. Therefore, the current MLT frame transmission carries the relevant frame information of its preamble frame.

本实施例中的解码装置，采用图 12所示的本发明实施例提供的丢帧 隐藏解码装置的第一较佳实施例。 如图 12所示， 该解码装置包括帧错 检测器 201、 IMLT解码模块 202、 丢帧隐藏单元 220、 解码波形历史緩 沖 210。The decoding apparatus in this embodiment uses the first preferred embodiment of the frame loss concealment decoding apparatus provided in the embodiment of the present invention shown in FIG. As shown in FIG. 12, the decoding apparatus includes a frame error detector 201, an IMLT decoding module 202, a frame loss concealing unit 220, and a history of decoding waveforms. Rush 210.

其中， 帧错检测器 201 , 用于对从编码装置获取的 MLT帧进行错 / 丢帧检测， 生成错 /丢帧错误标识， 并发送给丢帧隐藏单元 220。 由于在 当前 MLT帧正常但前置 MLT帧为错 /丢帧时， 需要 IMLT解码模块 202 协助对当前 MLT帧进行 IDCT变换。 因此， 帧错检测器 201还要根据错 /丢帧错误标识进行判断：如果判断出当前 MLT帧正常且前置 MLT帧正 常， 则帧错检测器 201将当前 MLT帧发送到 IMLT解码模块 202; 如果 判断出当前 MLT帧没错， 但是前置 MLT帧为错 /丢帧， 则将错 /丢帧错 误标识和当前 MLT帧一起发送给 IMLT解码模块 202, 并向丢帧隐藏单 元 220发送错 /丢帧标识； 否则只需要将错 /丢帧错误标识发送给丢帧隐 藏单元 220。 The frame error detector 201 is configured to perform error/drop frame detection on the MLT frame acquired from the encoding device, generate an error/drop frame error identifier, and send the error/drop frame error identifier to the frame loss concealing unit 220. Since the current MLT frame is normal but the pre-MLT frame is a wrong/drop frame, the IMLT decoding module 202 is required to assist in IDCT transformation of the current MLT frame. Therefore, the frame error detector 201 further determines according to the error/drop frame error flag: if it is determined that the current MLT frame is normal and the pre-MLT frame is normal, the frame error detector 201 sends the current MLT frame to the IMLT decoding module 202; If it is determined that the current MLT frame is correct, but the pre-MLT frame is an error/drop frame, the error/drop frame error identifier is sent to the IMLT decoding module 202 together with the current MLT frame, and the frame loss concealing unit 220 is sent an error/ The frame loss identifier is sent; otherwise, the error/drop frame error identifier needs to be sent to the frame loss concealment unit 220.

IMLT解码模块 202, 接收到当前 MLT帧时， 对其进行 IMLT变换。 将变换得到的当前 MLT帧重叠相加时域帧和最终时域帧存入解码波形 历史緩沖 210中。 当接收到当前 MLT帧及其错 /丢帧标识时， 只对当前 MLT帧进行 IDCT变换， 将得到的当前 MLT帧重叠相加时域帧存入解 码波形历史緩沖 210中， 最终时域帧则保存丢帧隐藏单元 220确定的恢 复帧。 The IMLT decoding module 202 performs IMLT transformation on the current MLT frame when it receives the current MLT frame. The transformed MLT frame overlap-added time domain frame and final time domain frame are stored in the decoded waveform history buffer 210. When the current MLT frame and its error/lost frame identifier are received, only the current MLT frame is subjected to IDCT transformation, and the obtained current MLT frame overlap-added time domain frame is stored in the decoded waveform history buffer 210, and the final time domain frame is The recovery frame determined by the lost frame hiding unit 220 is saved.

在实际应用时，上述帧错检测器 201也可以在当前 MLT帧为正常帧 时， 将错 /丢帧错误标识和当前 MLT帧一起发送给 IMLT解码模块 202 , 由 IMLT解码模块 202根据错 /丢帧错误标识确定后续的操作。 In actual application, the frame error detector 201 may also send the error/drop frame error identifier together with the current MLT frame to the IMLT decoding module 202 when the current MLT frame is a normal frame, and the IMLT decoding module 202 may be based on the error/loss. The frame error flag identifies subsequent operations.

IMLT解码模块 202根据帧错检测器 201发来的错 /丢帧错误标识， 对当前 MLT帧进行解码。 当错 /丢帧错误标识指示当前 MLT帧和前置 MLT帧均为正常帧时， 将解码得到的当前 MLT帧重叠相加时域帧和最 终时域帧存入解码波形历史緩沖 210 中。 当错 /丢帧错误标识指示当前 MLT帧正常但前置 MLT帧为错 /丢帧时， 只对当前 MLT帧进行 IDCT 变换，将得到的当前 MLT帧重叠相加时域帧存入解码波形历史緩沖 210 中， 最终时域帧则保存丢帧隐藏单元 220确定的恢复帧。The IMLT decoding module 202 decodes the current MLT frame based on the error/drop frame error flag sent by the frame error detector 201. When the error/drop frame error flag indicates that the current MLT frame and the pre-MLT frame are both normal frames, the decoded current MLT frame overlap-added time domain frame and the final time domain frame are stored in the decoded waveform history buffer 210. When the error/drop frame error flag indicates that the current MLT frame is normal but the pre-MLT frame is wrong/drop frame, only the current MLT frame is IDCT. The obtained MLT frame overlap-added time domain frame is stored in the decoded waveform history buffer 210, and the final time domain frame stores the restored frame determined by the lost frame hiding unit 220.

丢帧隐藏单元 220, 用于根据帧错检测器 201发来的错 /丢帧错误标 识， 结合接收的当前 MLT帧相关帧信息和緩存的前置 MLT帧相关帧信 息，确定丢帧隐藏策略；再根据该丢帧隐藏策略和解码波形历史緩沖 210 中保存的重叠相加时域帧和最终时域帧恢复出错 /丢帧。 The frame loss concealing unit 220 is configured to determine a frame loss concealment strategy according to the received error/drop frame error identifier sent by the frame error detector 201, combined with the received current MLT frame related frame information and the buffered pre-MLT frame related frame information; The error/drop frame is recovered according to the overlap-and-add time domain frame and the final time domain frame saved in the frame loss concealment strategy and the decoded waveform history buffer 210.

丢帧隐藏单元 220包括相关帧信息历史緩沖 221、 丢帧隐藏策略选 择模块 222和丢帧恢复模块 223。 The frame loss concealing unit 220 includes an associated frame information history buffer 221, a frame loss concealment strategy selection module 222, and a frame loss recovery module 223.

相关帧信息历史緩沖 221 , 用于緩沖当前 MLT帧相关帧信息和前置 MLT帧相关帧信息。 The related frame information history buffer 221 is configured to buffer the current MLT frame related frame information and the pre-MLT frame related frame information.

丢帧隐藏策略选择模块 222, 用于存储丢帧隐藏策略。 根据帧错检 测器 201发来的错 /丢帧错误标识，以及相关帧信息历史緩沖 221中的相 关帧信息， 确定丢帧隐藏策略。 丢帧隐藏策略事先设计并存储在该丢帧 隐藏策略选择模块 222中， 作为丢帧隐藏策略的结合， 可以采用表的形 式存在， 便于查找和使用。 如表 2所示的丢帧隐藏策略表。 该丢帧隐藏 策略选择模块 222还需要生成前帧可靠性标识，用于在出现连续丢帧时， 选择丢帧隐藏策略。 The frame loss hiding policy selection module 222 is configured to store a frame loss hiding policy. The frame loss concealment strategy is determined based on the error/drop frame error flag sent by the frame error detector 201 and the associated frame information in the associated frame information history buffer 221. The frame loss concealment strategy is pre-designed and stored in the frame loss concealment strategy selection module 222. As a combination of the frame loss concealment strategy, the frame loss can be used in the form of a table for easy searching and use. The frame loss hidden strategy table shown in Table 2. The frame loss concealment policy selection module 222 also needs to generate a pre-frame reliability identifier, which is used to select a frame loss concealment strategy when consecutive frame drops occur.

丢帧恢复模块 223 , 用于根据丢帧隐藏策略选择模块 222确定的丢 帧隐藏策略和解码波形历史緩沖 210中保存的重叠相加时域帧和最终时 域帧恢复出错 /丢帧。该恢复的错 /丢帧还要作为当前 MLT的最终时域帧， 发送给解码波形历史緩沖 The frame loss recovery module 223 is configured to recover the error/drop frame according to the frame loss concealment strategy determined by the frame loss concealment policy selection module 222 and the overlap addition time domain frame and the final time domain frame saved in the decoded waveform history buffer 210. The recovered error/drop frame is also sent to the decoded waveform history buffer as the final time domain frame of the current MLT.

解码波形历史緩沖 210, 用于存储重叠相加时域帧和最终时域帧。 该模块又细分为重叠相加时域帧历史緩沖 211 和最终时域帧历史緩沖 212。 The decoded waveform history buffer 210 is configured to store the overlap-added time domain frame and the final time domain frame. The module is further subdivided into an overlap-added time domain frame history buffer 211 and a final time domain frame history buffer 212.

重叠相加时域帧历史緩沖 211 ,用于緩沖两帧经 IMLT解码模块 202 解码的重叠相加时域帧。An overlap-added time domain frame history buffer 211 for buffering two frames via the IMLT decoding module 202 The decoded overlap adds the time domain frame.

最终时域帧历史緩沖 212, 用于緩沖经 IMLT解码模块 202解码的 当前 MLT帧最终时域帧， 或緩沖丢帧恢复模块 223恢复的错 /丢帧。 The final time domain frame history buffer 212 is configured to buffer the final time domain frame of the current MLT frame decoded by the IMLT decoding module 202, or the error/drop frame recovered by the buffered frame loss recovery module 223.

图 13 为本发明实施例丢帧隐藏方法的第二较佳实施例的方法流程 图。 该实施例采用编码装置的第一较佳实施例和解码装置的第一较佳实 施例组成丢帧隐藏系统， 并实现丢帧隐藏方法。 该方法具体步骤如下： 步骤 301 , 编码装置的 MLT编码单元 110根据接收的后续帧和当前 帧历史緩沖 121存储的当前帧进行 MLT变换， 生成 MLT帧。 FIG. 13 is a flow chart of a method of a second preferred embodiment of a method for dropping frames according to an embodiment of the present invention. This embodiment uses a first preferred embodiment of the encoding device and a first preferred embodiment of the decoding device to form a frame loss concealment system, and implements a frame dropping method. The specific steps of the method are as follows: Step 301: The MLT encoding unit 110 of the encoding apparatus performs MLT transformation according to the received subsequent frame and the current frame stored in the current frame history buffer 121 to generate an MLT frame.

步骤 302, 帧内相关性检测器 131对当前帧进行检测， 生成帧内相 关性信息； 帧间相关性检测器 132对当前帧和前置帧进行检测， 生成帧 间相关性信息。 Step 302: The intra-frame correlation detector 131 detects the current frame to generate intra-frame correlation information. The inter-frame correlation detector 132 detects the current frame and the pre-frame to generate inter-frame correlation information.

步骤 303, 编码装置将 MLT帧及帧内和帧间相关性信息发送给解码 装置。 Step 303: The encoding device sends the MLT frame and the intra-frame and inter-frame correlation information to the decoding device.

步骤 304, 解码装置的帧错检测器 201接收 MLT帧， 相关帧信息历 史緩沖 221緩沖帧内和帧间相关性信息。 Step 304: The frame error detector 201 of the decoding device receives the MLT frame, and the related frame information history buffer 221 buffers the intra-frame and inter-frame correlation information.

步骤 305, 帧错检测器 201对当前 MLT帧进行错 /丢帧检测，判断当 前 MLT帧或前置 MLT帧是否为错 /丢帧； 如果是则执行步骤 306; 否则 执行步骤 320和 321 , 帧错检测器 201将当前 MLT帧发送给 IMLT解码 模块 202 , 该模块对当前 MLT帧进行正常 MLT解码。 之后将当前 MLT 帧经 IDCT变换后得到重叠相加时域帧， 保存在重叠相加时域帧历史緩 沖 211 , 并将解码得到的最终时域波形保存在最终时域帧历史緩沖 212, 此时结束本流程。 Step 305, the frame error detector 201 performs error/drop frame detection on the current MLT frame, and determines whether the current MLT frame or the pre-MLT frame is an error/drop frame; if yes, step 306 is performed; otherwise, steps 320 and 321 are performed. The error detector 201 sends the current MLT frame to the IMLT decoding module 202, which performs normal MLT decoding on the current MLT frame. Then, the current MLT frame is subjected to IDCT transformation to obtain an overlap-added time domain frame, and is stored in the overlap-added time domain frame history buffer 211, and the decoded final time domain waveform is saved in the final time domain frame history buffer 212. End this process.

步骤 306, 帧错检测器 201进一步判断当前 MLT是否为正常帧， 如 果是则执行步骤 307和 308; 否则直接执行 308。 Step 306, the frame error detector 201 further determines whether the current MLT is a normal frame, and if yes, performs steps 307 and 308; otherwise, directly executes 308.

步骤 307,帧错检测器 201将错 /丢帧错误标识和当前 MLT帧发送给 IMLT解码模块 202, 该模块对当前 MLT帧进行 IDCT变换； 之后将当 前 MLT帧经 IDCT变换后得到重叠相加时域帧，保存在重叠相加时域帧 历史緩沖 211。Step 307, the frame error detector 201 sends the error/drop frame error identifier and the current MLT frame to The IMLT decoding module 202 performs IDCT transformation on the current MLT frame; then, the current MLT frame is subjected to IDCT transformation to obtain an overlap-added time domain frame, and is stored in the overlap-added time domain frame history buffer 211.

步骤 308, 帧错检测器 201将错 /丢帧错误标识发送给丢帧隐藏策略 选择模块 222。 Step 308: The frame error detector 201 sends the error/drop frame error identifier to the frame loss concealment strategy selection module 222.

本实施例中，错 /丢帧错误标识包括 1比特的前置帧错误标识和 1比 特的当前帧错误标识。 In this embodiment, the error/lost frame error identifier includes a 1-bit pre-frame error identifier and a 1-bit current frame error identifier.

如果当错 /丢帧错误标识为 01 , 则执行步骤 309、 312 - 314; 如果当 错 /丢帧错误标识为 10, 则执行步骤 310、 312 - 314; 如果当错 /丢帧错 误标识为 11 , 则执行步骤 311、 312 - 314。 If the error/drop frame error flag is 01, then steps 309, 312-314 are performed; if the error/drop frame error flag is 10, steps 310, 312-314 are performed; if the error/drop frame error flag is 11 Then, steps 311, 312-314 are performed.

步骤 309, 丢帧隐藏策略选择模块 222根据从相关帧信息历史緩沖 221获取的当前 MLT帧的帧内和帧间相关性信息， 选择丢帧隐藏策略。 Step 309, the frame loss concealment policy selection module 222 selects a frame loss concealment strategy according to intra-frame and inter-frame correlation information of the current MLT frame acquired from the related frame information history buffer 221 .

步骤 310, 丢帧隐藏策略选择模块 222根据从相关帧信息历史緩沖 221获取的前置 MLT帧和当前 MLT帧的帧内和帧间相关性信息， 选择 丢帧隐藏策略。 Step 310: The frame loss concealment policy selection module 222 selects a frame loss concealment strategy according to the intra-frame and inter-frame correlation information of the pre-MLT frame and the current MLT frame acquired from the related frame information history buffer 221.

步骤 311 , 丢帧隐藏策略选择模块 222根据前帧可靠性标识和从相 关帧信息历史緩沖 221获取的当前 MLT帧的帧内和帧间相关性信息， 选择丢帧隐藏策略。 Step 311: The frame loss concealment policy selection module 222 selects a frame loss concealment strategy according to the pre-frame reliability identifier and the intra-frame and inter-frame correlation information of the current MLT frame acquired from the related frame information history buffer 221 .

以上步骤 309、 310选择的隐藏策略具体参见表 2的内容。 步骤 311 选择的隐藏策略具体参见图 9。 The hidden policies selected in the above steps 309 and 310 are specifically shown in Table 2. Step 311 The hidden policy selected is shown in Figure 9.

步骤 312, 丢帧隐藏策略选择模块 222根据选择的隐藏策略更新前 帧可靠性标识。 该步骤业务可以在步骤 313之后执行。 Step 312: The frame loss concealment policy selection module 222 updates the pre-frame reliability identifier according to the selected concealment policy. This step of the service can be performed after step 313.

步骤 313 , 丢帧恢复模块 223根据选择的隐藏策略、 重叠相加时域 帧历史緩沖 211中的重叠相加时域帧和最终时域帧历史緩沖 212中的最 终时域帧恢复错 /丢帧。 步骤 314, 丢帧恢复模块 223将恢复出的错 /丢帧输出， 并存入最终 时域帧历史緩沖 212。Step 313, the frame loss recovery module 223 recovers the error/drop frame according to the selected hidden policy, the overlap-added time domain frame in the overlap-added time domain frame history buffer 211, and the final time domain frame in the final time domain frame history buffer 212. . Step 314, the frame loss recovery module 223 outputs the recovered error/drop frame and stores it in the final time domain frame history buffer 212.

以上便完成了丢帧隐藏。 The above completes the frame loss hiding.

为了较好地识别短音符， 本发明在编码端利用前置帧和后续帧对当 前帧进行短音符检测， 输出短音符信息。 该短音符信息作为相关帧信息 的一部分， 包括短音符标识或者包括短音符标识和短音符特征参数。 解 码端根据该短音符信息恢复错 /丢帧。 In order to better recognize short notes, the present invention performs short note detection on the current frame using the pre-frame and subsequent frames at the encoding end, and outputs short note information. The short note information is included as part of the associated frame information, including short note identification or including short note identification and short note feature parameters. The decoding end recovers the error/drop frame based on the short note information.

在实际应用中， 短音符信息中的短音符标识可以用 1比特的二进制 数表示， 1为短音符， 0为非短音符。 短音符特征参数可以有多种形式， 如可以是一组代表短音符基音， 幅度包络等特征的参数， 也可以是一个 由另一个精度低的编码器编码所得的粗糙版本的帧等。 In practical applications, short note identifiers in short note information can be represented by a 1-bit binary number, 1 for short notes and 0 for non-short notes. The short note characteristic parameters can be in various forms, such as a set of parameters representing a short note pitch, an amplitude envelope, or the like, or a coarse version of the frame encoded by another low precision encoder.

在确定当前帧为短音符帧时， 根据短音符帧的定义， 一个短音符帧 有较强的帧内相关性， 但与前后相邻帧的相关性却较弱。 因此在短音符 检测中首先对当前帧判断其帧内相关性。 如果帧内相关性为强， 则分别 进一步检查其与前置帧及后续帧的相关性。 如果与它们的相关性均为 弱， 则判定当前帧为短音符帧。 相关性的计算可用自相关， 互相关函数 的方法， 如前所述。 When it is determined that the current frame is a short note frame, according to the definition of the short note frame, a short note frame has a strong intra-frame correlation, but the correlation with the adjacent frames is weak. Therefore, in the short note detection, the intra frame correlation is first judged for the current frame. If the intra-frame correlation is strong, the correlation with the pre-frame and subsequent frames is further checked separately. If the correlation with them is weak, it is determined that the current frame is a short note frame. Correlation calculations can be performed using autocorrelation, cross-correlation functions, as previously described.

确定了当前帧为短音符帧，就要确定该短音符帧的短音符特征参数。 在短音符特征参数为短音符基音和短音符幅度包络的情况下， 短音 符基音表示的是一段周期性强的音频信号中最小的一段周期。 短音符幅 度包络表示的是这一帧波形的形状。 当短音符特征参数为短音符基音， 幅度包络等特征参数时， 恢复的方法为： 在整个帧内重复短音符基音， 然后根据幅度包络信息调整帧包络的形状。 It is determined that the current frame is a short note frame, and the short note characteristic parameter of the short note frame is determined. In the case where the short note characteristic parameter is a short note pitch and a short note amplitude envelope, the short note pitch represents the smallest period of a periodic strong audio signal. The short note amplitude envelope represents the shape of this frame waveform. When the short note characteristic parameter is a short note pitch, amplitude envelope and other characteristic parameters, the recovery method is: repeating the short note pitch in the entire frame, and then adjusting the shape of the frame envelope according to the amplitude envelope information.

当短音符特征参数为粗糙编码的帧时， 此时的特征参数实际上是由 一个精度较低的编码器在编码端编码形成。 实际上这种方式下编码端有 两个编码器， 高精度编码器对每一帧都编码， 低精度编码器只用来对短 音符帧进行编码。 当高精度编码的短音符帧丢失时， 就对低精度编码的 该帧进行解码。 解码端也是有两个解码器的。When the short note feature parameter is a coarsely encoded frame, the feature parameter at this time is actually formed by the encoder with a lower precision at the encoding end. In fact, in this way, the encoding end has Two encoders, high-precision encoders encode each frame, and low-precision encoders are only used to encode short note frames. When a high-precision encoded short note frame is lost, the frame of low-precision encoding is decoded. The decoder also has two decoders.

图 14 为本发明实施例丢帧隐藏编码装置第二较佳实施例的组成框 图。 如图 14所示， 该编码装置包括 MLT编码单元 110、 历史緩沖单元 120、 相关帧信息生成单元 730。 与图 11所示的编码装置的第一较佳实 施例的不同之处在于， 本实施例中， 相关帧信息生成单元 730不仅包括 帧内相关性检测器 731和帧间相关性检测器 732, 还包括短音符检测器 733, 用于根据编码装置接收的后续帧、 从当前帧历史緩沖 121 获取的 当前帧以及从前置帧历史緩沖 122获取的前置帧， 生成短音符信息。 具 体是根据前置帧、 当前帧和后续帧这 3帧生成短音符标识， 根据当前帧 生成短音符特征参数。 FIG. 14 is a block diagram showing a second preferred embodiment of a frame dropping concealment coding apparatus according to an embodiment of the present invention. As shown in FIG. 14, the encoding apparatus includes an MLT encoding unit 110, a history buffer unit 120, and a correlation frame information generating unit 730. The difference from the first preferred embodiment of the encoding apparatus shown in FIG. 11 is that, in this embodiment, the related frame information generating unit 730 includes not only the intra-correlation detector 731 and the inter-frame correlation detector 732. Also included is a short note detector 733 for generating short note information based on a subsequent frame received by the encoding device, a current frame acquired from the current frame history buffer 121, and a preamble frame acquired from the preamble history buffer 122. Specifically, a short note identifier is generated according to the three frames of the pre-frame, the current frame, and the subsequent frame, and the short note feature parameter is generated according to the current frame.

因此， 编码装置通过分组通信网络发给解码装置的信息包括当前 MLT帧、 帧内相关性信息、 帧间相关性信息和短音符信息。 Therefore, the information transmitted by the encoding device to the decoding device through the packet communication network includes the current MLT frame, intra-frame correlation information, inter-frame correlation information, and short note information.

解码装置可以采用图 12 示出的解码装置的组成结构作为解码装置 的第二较佳实施例。 与解码装置的第一较佳实施例不同之处在于， 本实 施例中， 相关帧信息历史緩沖 221还保存短音符信息。 当丢帧隐藏策略 选择模块 222收到帧错检测器 201发来的错 /丢帧错误标识，且根据相关 帧信息历史緩沖 211所緩存的当前帧短音符信息发现， 当前错 /丢帧为短 音符帧， 则只需根据短音符信息恢复丢帧即可。 The decoding apparatus can employ the constituent structure of the decoding apparatus shown in Fig. 12 as the second preferred embodiment of the decoding apparatus. The difference from the first preferred embodiment of the decoding apparatus is that, in the embodiment, the related frame information history buffer 221 also holds short note information. When the frame loss concealment policy selection module 222 receives the error/drop frame error identifier sent by the frame error detector 201, and finds the current frame short note information buffered by the associated frame information history buffer 211, the current error/lost frame is short. For note frames, you only need to recover lost frames based on short note information.

图 15为本发明实施例丢帧隐藏方法第三较佳实施例的方法流程图。 该方法采用编码装置的第二较佳实施例和解码装置的第二较佳实施例 组成丢帧隐藏系统， 并实现丢帧隐藏方法。 该方法包括以下步骤： FIG. 15 is a flowchart of a method for deleting a frame hiding method according to a third preferred embodiment of the present invention. The method uses a second preferred embodiment of the encoding device and a second preferred embodiment of the decoding device to form a frame loss hiding system, and implements a frame dropping method. The method includes the following steps:

步骤 401 , 编码装置的 MLT编码单元 110根据接收的后续帧和历史 緩沖 121存储的当前帧进行 MLT变换， 生成 MLT帧。 步骤 402, 帧内相关性检测器 731对当前帧进行检测， 生成帧内相 关性信息； 帧间相关性检测器 732对当前帧和前置帧进行检测， 生成帧 间相关性信息。 短音符相关性检测器 733根据前置帧和后续帧对当前帧 进行检测， 生成短音符信息。Step 401: The MLT encoding unit 110 of the encoding apparatus performs MLT transformation according to the received subsequent frame and the current frame stored in the history buffer 121 to generate an MLT frame. Step 402: The intra-frame correlation detector 731 detects the current frame to generate intra-frame correlation information. The inter-frame correlation detector 732 detects the current frame and the pre-frame to generate inter-frame correlation information. The short note correlation detector 733 detects the current frame based on the pre-frame and the subsequent frame to generate short note information.

本实施例中， 短音符信息包括短音符标识和短音符特征参数。 如果 短音符相关性检测器检测出当前帧为短音符帧， 则将短音符标识设定为 In this embodiment, the short note information includes a short note identifier and a short note feature parameter. If the short note correlation detector detects that the current frame is a short note frame, set the short note identification to

1 , 并产生短音符特征参数。 如果短音符相关性检测器检测出当前帧不 为短音符帧， 则将短音符标识设定为 0, 也不产生短音符特征参数。1 , and produce short note feature parameters. If the short note correlation detector detects that the current frame is not a short note frame, the short note identification is set to 0 and no short note feature parameters are generated.

步骤 403, 编码装置将 MLT帧及其帧内和帧间相关性信息、 短音符 信息发送给解码装置。 Step 403: The encoding device sends the MLT frame and its intra and inter-frame correlation information and short note information to the decoding device.

步骤 404, 解码装置的帧错检测器 201接收 MLT帧， 相关帧信息历 史緩沖 221接收帧内和帧间相关性信息、 以及短音符信息。 Step 404, the frame error detector 201 of the decoding device receives the MLT frame, and the related frame information history buffer 221 receives the intra and interframe correlation information, and the short note information.

步骤 405, 帧错检测器 201对当前 MLT帧进行错 /丢帧检测，判断当 前 MLT帧或前置 MLT帧是否有错 /丢帧， 生成错 /丢帧信息； 如果有则 执行步骤 406; 否则执行步骤 420和 421。 步骤 420和 421与前述步骤 320和 321相同。 Step 405: The frame error detector 201 performs error/drop frame detection on the current MLT frame, determines whether the current MLT frame or the pre-MLT frame has an error/drop frame, and generates error/lost frame information; if yes, step 406; otherwise Steps 420 and 421 are performed. Steps 420 and 421 are the same as steps 320 and 321 described above.

步骤 406, 帧错检测器 201进一步判断当前 MLT是否为正常帧， 如 果是则执行步骤 407和 408； 否则直接执行 408。 Step 406, the frame error detector 201 further determines whether the current MLT is a normal frame, and if yes, performs steps 407 and 408; otherwise, directly executes 408.

步骤 407与前述步骤 307相同。 Step 407 is the same as step 307 described above.

步骤 408, 帧错检测器 201将错 /丢帧错误标识发送给丢帧隐藏策略 选择模块 222。 Step 408, the framing error detector 201 sends the error/lost frame error identifier to the frame loss concealment strategy selection module 222.

步骤 409, 丢帧隐藏策略选择模块 222根据从相关帧信息历史緩沖 221 获得的短音符信息判断当前错 /丢帧是否为短音符。 如果当前 MLT 帧为错 /丢帧且是短音符， 则执行步骤 430, 丢帧隐藏策略选择模块 222 将短音符信息发送给丢帧恢复模块 223, 由丢帧恢复模块 223根据短音 符信息， 恢复丢失帧。 且该恢复帧被认为不可靠， 然后执行步骤 413。 否则， 执行步骤 410。Step 409, the frame loss concealment policy selection module 222 determines whether the current error/lost frame is a short note based on the short note information obtained from the related frame information history buffer 221. If the current MLT frame is a wrong/drop frame and is a short note, step 430 is executed, the frame loss concealment policy selection module 222 sends the short note information to the frame loss recovery module 223, and the frame loss recovery module 223 is based on the short tone. Symbol information, recover lost frames. And the recovery frame is considered unreliable, and then step 413 is performed. Otherwise, step 410 is performed.

以上步骤 409中， 判断当前错帧是否为短音符时， 可以采用短音符 信息中短音符标识来判断。 In the above step 409, when it is judged whether the current frame is a short note, the short note identifier in the short note information can be used for judging.

步骤 410, 丢帧隐藏策略选择模块 222根据错 /丢帧错误检测标识， 结合相关帧信息和前帧可靠性标识， 选择隐藏策略。 具体选择方法参见 图 13的步骤 309、 310和 311。 Step 410: The frame loss concealment policy selection module 222 selects a concealment strategy according to the error/lost frame error detection identifier, combined with the related frame information and the previous frame reliability identifier. See steps 309, 310, and 311 of Figure 13 for specific selection methods.

步骤 411 , 丢帧隐藏策略选择模块 222根据选择的隐藏策略更新前 帧可靠性标识。 Step 411: The frame loss hiding policy selection module 222 updates the previous frame reliability identifier according to the selected hiding policy.

步骤 412, 丢帧恢复模块 223根据选择的隐藏策略、 重叠相加时域 帧历史緩沖 211中的重叠相加时域帧和最终时域帧历史緩沖 212中的最 终时域帧恢复出错 /丢帧。 Step 412, the frame loss recovery module 223 recovers the error/drop frame according to the selected hidden policy, the overlap-added time domain frame in the overlap-added time domain frame history buffer 211, and the final time domain frame in the final time domain frame history buffer 212. .

步骤 413, 丢帧恢复模块 223将恢复出的错 /丢帧输出， 并存入最终 时域帧历史緩沖 212。 Step 413, the frame loss recovery module 223 outputs the recovered error/drop frame and stores it in the final time domain frame history buffer 212.

本流程结束。 This process ends.

以上的各实施例中， 相关帧信息包括帧间和帧内相关性信息和短音 符信息。 当然还可以产生更多的相关帧信息种类与 MLT帧一同传送。 这些新的相关帧信息可以更详细的描述帧内和帧间相关性， 例如前置帧 与后续帧的相关性等； 新的相关帧信息还可以是一组音频帧特征参数。 当采用复制其它帧波形并衰减的方法恢复错 /丢帧时，为这个不可靠的恢 复帧提供更加详细的特征信息， 根据这组特征信息恢复错 /丢帧， 将会大 大提高错 /丢帧恢复的正确性。其丢帧隐藏方法与本发明实施例阐述的方 法原理相同。 In the above embodiments, the correlation frame information includes inter-frame and intra-frame correlation information and short note information. Of course, more related frame information types can be generated and transmitted together with the MLT frame. These new related frame information can describe intra- and inter-frame correlation in more detail, such as the correlation between the pre-frame and subsequent frames, etc. The new related frame information can also be a set of audio frame feature parameters. When the error/drop frame is recovered by copying other frame waveforms and attenuating, the more reliable feature information is provided for the unreliable recovery frame, and the error/drop frame is recovered according to the set of feature information, which will greatly improve the error/drop frame. The correctness of the restoration. The frame dropping method is the same as the method described in the embodiment of the present invention.

由以上所述可以看出， 本发明实施例所提供的丢帧隐藏的方法、 系 统、 编码装置和解码装置， 能够选取不同的隐藏策略， 恢复错 /丢帧。 该 隐藏策略利用音频帧的时域信号与其经 MLT重叠相加后的信号的代数 关系， 通过估计出的部分时域音频信号来求得剩下部分的时域音频信 号； 还利用 MLT重叠相加过程中的窗口函数特性， 利用恢复帧中一部 分保存较完好的帧得到错 /丢帧的恢复帧。因此本发明实施例能够比较好 的恢复出错 /丢帧， 提高了丢帧隐藏能力， 避免了时域信号混淆的现象。It can be seen from the above that the method, system, encoding device and decoding device for frame loss concealment provided by the embodiments of the present invention can select different hidden policies and recover errors/drop frames. The The hidden strategy utilizes the algebraic relationship between the time domain signal of the audio frame and the signal added by the MLT overlap, and obtains the remaining part of the time domain audio signal by estimating the partial time domain audio signal; and also utilizes the MLT overlap addition process. In the window function feature, a recovery frame with a wrong/drop frame is obtained by using a part of the restored frame to hold a better frame. Therefore, the embodiment of the present invention can better recover errors/drop frames, improve the frame hiding capability, and avoid the confusion of time domain signals.

本发明实施例在生成相关帧信息时， 还生成短音符信息， 标识当前 In the embodiment of the present invention, when generating related frame information, short note information is also generated, and the current identifier is identified.

MLT 帧是否为短音符帧， 同时针对短音符帧还生成一组短音符特征参 数， 用以解码装置在该帧丢失时， 能够较好的恢复出短音符错 /丢帧。Whether the MLT frame is a short note frame, and a short note feature parameter is also generated for the short note frame, so that the decoding device can recover the short note error/drop frame when the frame is lost.

综上所述， 以上仅为本发明的较佳实施例而已， 并非用于限定本发 明的保护范围。 凡在本发明的精神和原则之内， 所作的任何修改、 等同 替换、 改进等， 均应包含在本发明的保护范围之内。 In conclusion, the above is only the preferred embodiment of the present invention and is not intended to limit the scope of the present invention. Any modifications, equivalents, improvements, etc. made within the spirit and scope of the present invention are intended to be included within the scope of the present invention.

Claims

权利要求书 Claim

1、 一种丢帧隐藏方法， 其特征在于， 在解码端存储历史时域波形， 该方法包括：A method for hiding a frame loss, characterized in that a historical time domain waveform is stored at a decoding end, and the method comprises:

编码端将获得的相关帧信息以及当前重叠调制变换 MLT 帧发送给 解码端； The encoding end sends the obtained related frame information and the current overlapping modulation transform MLT frame to the decoding end;

所述解码端对接收的所述当前 MLT帧进行错 /丢帧检测， 生成错 /丢 帧信息； 根据所述错 /丢帧信息， 所述相关帧信息和所述历史时域波形恢 复出错 /丢帧。 The decoding end performs error/drop frame detection on the received current MLT frame, and generates error/lost frame information. According to the error/drop frame information, the related frame information and the historical time domain waveform recovery error/ Drop frame.

2、如权利要求 1所述的方法， 其特征在于， 所述编码端将获得的相 关帧信息以及当前 MLT帧发送给解码端是分包发送的； 或将所述相关 帧信息嵌在当前 MLT帧的相邻帧发送； 或将所述相关帧信息作为增强 层数据， 将所述当前 MLT帧作为核心层数据发送。 The method according to claim 1, wherein the encoding end sends the obtained related frame information and the current MLT frame to the decoding end to be packetized, or embeds the related frame information in the current MLT. The adjacent frame of the frame is transmitted; or the related frame information is used as enhancement layer data, and the current MLT frame is transmitted as core layer data.

3、 如权利要求 1所述的方法， 其特征在于， 所述生成错 /丢帧信息 包括： 3. The method according to claim 1, wherein the generating the error/drop frame information comprises:

所述解码端对所述当前 MLT帧进行错 /丢帧检测， 将检测结果作为 当前 MLT帧的当前帧错 /丢帧信息； The decoding end performs error/drop frame detection on the current MLT frame, and uses the detection result as current frame error/lost frame information of the current MLT frame;

所述解码端将所述当前 MLT帧前一帧的错 /丢帧信息作为当前 MLT 帧的前置帧错 /丢帧信息。 The decoding end uses the error/lost frame information of the previous frame of the current MLT frame as the pre-frame error/drop frame information of the current MLT frame.

4、 如权利要求 1所述的方法， 其特征在于， 所述相关帧信息包括帧 内相关性信息和帧间相关性信息； 4. The method according to claim 1, wherein the related frame information includes intra-frame correlation information and inter-frame correlation information;

所述获得相关帧信息以及当前 MLT帧包括： The obtaining related frame information and the current MLT frame includes:

编码端根据当前帧及其后续帧对当前帧进行 MLT 变换， 生成当前 MLT帧； The encoding end performs MLT transformation on the current frame according to the current frame and subsequent frames to generate a current MLT frame.

根据当前帧进行帧内相关性检测， 生成帧内相关性信息； 根据当前帧及其前置帧进行帧间相关性检测 ,生成帧间相关性信息。Performing intra-frame correlation detection according to the current frame to generate intra-frame correlation information; Inter-frame correlation detection is performed according to the current frame and its preamble frame, and inter-frame correlation information is generated.

5、 如权利要求 4所述的方法， 其特征在于， 在前置 MLT帧为正常 帧且当前帧为错 /丢帧时， 所述恢复出错 /丢帧为： The method according to claim 4, wherein when the pre-MLT frame is a normal frame and the current frame is a wrong/drop frame, the recovery error/drop frame is:

解码端才 据当前 MLT帧的帧间相关性信息、帧内相关性信息和所述 历史时域波形恢复出错 /丢帧。 The decoding end recovers the error/drop frame according to the inter-frame correlation information, the intra-frame correlation information and the historical time domain waveform of the current MLT frame.

6、如权利要求 5所述的方法， 其特征在于， 所述历史时域波形包括 前置 MLT帧最终时域帧和前置 MLT帧重叠相加时域帧； The method according to claim 5, wherein the historical time domain waveform comprises a pre-MLT frame final time domain frame and a pre-MLT frame overlap-add time domain frame;

解码端才 据当前 MLT帧的帧间相关性信息、帧内相关性信息和所述 历史时域波形恢复出的错 /丢帧为当前 MLT帧最终时域帧的前半帧和后 半帧， 具体恢复过程为： The decoding end recovers the error/drop frame according to the inter-frame correlation information, the intra-frame correlation information and the historical time domain waveform of the current MLT frame as the first half frame and the second half frame of the final time domain frame of the current MLT frame, specifically The recovery process is:

在当前 MLT帧的帧间相关性为强， 当前 MLT帧的帧内相关性为强 时： When the inter-frame correlation of the current MLT frame is strong, and the intra-frame correlation of the current MLT frame is strong:

重复前置 MLT帧最终时域帧的后半帧， 作为当前 MLT帧最终时域 帧的前半帧和后半帧； Repeating the second half of the final time domain frame of the pre-MLT frame as the first half frame and the second half frame of the final time domain frame of the current MLT frame;

在当前 MLT帧的帧间相关性为强， 当前 MLT帧的帧内相关性为弱 时： When the inter-frame correlation of the current MLT frame is strong, and the intra-frame correlation of the current MLT frame is weak:

重复前置 MLT帧最终时域帧的后半帧， 作为当前 MLT帧最终时域 帧的前半帧； 根据前置 MLT帧重叠相加时域帧的后半帧、 当前 MLT帧 最终时域帧的前半帧与当前 MLT帧最终时域帧的后半帧的代数关系， 确定所述当前 MLT帧最终时域帧的后半帧； Repeating the second half of the final time domain frame of the pre-MLT frame as the first half of the final time domain frame of the current MLT frame; and adding the second half of the time domain frame according to the pre-MLT frame overlap, and the final time domain frame of the current MLT frame Determining the algebraic relationship between the first half frame and the second half of the final time domain frame of the current MLT frame, and determining the second half of the final time domain frame of the current MLT frame;

在当前 MLT帧的帧间相关性为弱时： When the inter-frame correlation of the current MLT frame is weak:

重复前置 MLT帧最终时域帧的后半帧， 作为当前 MLT帧最终时域 帧的前半帧和后半帧， 并对得到的当前 MLT帧最终时域帧进行全帧衰 减。 The second half of the final time domain frame of the current MLT frame is repeated, and the first and second half frames of the final time domain frame of the current MLT frame are repeated, and the final frame of the current MLT frame is subjected to full frame attenuation.

7、 如权利要求 4所述的方法， 其特征在于， 在前置 MLT帧为错 / 丢帧且当前帧为正常帧时， 所述恢复出错 /丢帧为：7. The method of claim 4, wherein the pre-MLT frame is wrong/ When the frame is dropped and the current frame is a normal frame, the recovery error/drop frame is:

解码端根据当前 MLT帧的帧间相关性信息、帧内相关性信息和前置 MLT帧的帧间相关性信息， 以及历史时域波形恢复出错 /丢帧。 The decoding end recovers the error/drop frame according to the inter-frame correlation information of the current MLT frame, the intra-frame correlation information, and the inter-frame correlation information of the pre-MLT frame, and the historical time domain waveform.

8、 如权利要求 7所述的方法， 其特征在于， 所述历史时域波形包括 前置 MLT帧最终时域帧和所述当前 MLT帧经逆向 DCT变换得到的当 前 MLT 帧重叠相加时域帧； 或者包括前置 MLT帧最终时域帧和前置 MLT帧前一帧的重叠相加时域帧； The method according to claim 7, wherein the historical time domain waveform comprises a final time domain frame of a pre-MLT frame and a current MLT frame overlap addition time domain obtained by inverse DCT transform of the current MLT frame. a frame; or an overlap-added time domain frame including a final time domain frame of the pre-MLT frame and a previous frame of the pre-MLT frame;

所述解码端根据当前 MLT帧的帧间相关性信息、帧内相关性信息和 前置 MLT帧的帧间相关性信息， 以及历史时域波形恢复出的错 /丢帧为 当前 MLT帧最终时域帧的前半帧和后半帧， 具体恢复过程为： The decoding end is based on the inter-frame correlation information of the current MLT frame, the intra-frame correlation information, and the inter-frame correlation information of the pre-MLT frame, and the error/drop frame recovered by the historical time domain waveform is the final time of the current MLT frame. The first half frame and the second half frame of the domain frame, the specific recovery process is:

在当前 MLT帧的帧间相关性为弱、帧内相关性为强，或者当前 MLT 帧的帧间相关性为强、 帧内相关性为强且前置 MLT帧的帧间相关性为 弱时： When the inter-frame correlation of the current MLT frame is weak, the intra-frame correlation is strong, or the inter-frame correlation of the current MLT frame is strong, the intra-frame correlation is strong, and the inter-frame correlation of the pre-MLT frame is weak. :

将所述当前 MLT帧重叠相加时域帧的后半帧代替前置 MLT帧重叠 相加时域帧的后半帧， 对所述当前 MLT帧重叠相加时域帧的前半帧和 后半帧进行逆向 MLT变换得到当前 MLT帧最终时域帧的后半帧；或者， 将所述前置 MLT帧前一帧的重叠相加时域帧的后半帧代替前置 MLT帧 重叠相加时域帧的后半帧， 对所述当前 MLT帧重叠相加时域帧的前半 帧和所述前置 MLT帧前一帧的重叠相加时域帧的后半帧进行逆向 MLT 变换， 得到当前 MLT帧最终时域帧的后半帧； Substituting the current MLT frame overlap and adding the second half of the time domain frame to replace the pre-MLT frame and adding the second half of the time domain frame, and adding and adding the first half and the second half of the time domain frame to the current MLT frame The frame is subjected to inverse MLT transformation to obtain the second half of the final time domain frame of the current MLT frame; or, the overlap of the previous frame of the pre-MLT frame is added to the second half of the time domain frame instead of the pre-MLT frame. The latter half of the domain frame, performing inverse MLT transformation on the first half of the current MLT frame and the first half of the pre-MLT frame and the second half of the pre-MLT frame The last half of the final time domain frame of the MLT frame;

重复得到的所述当前 MLT帧最终时域帧的后半帧， 作为当前 MLT 帧最终时域帧的前半帧； And repeating the obtained second frame of the final time domain frame of the current MLT frame as the first half frame of the final time domain frame of the current MLT frame;

在当前 MLT 帧的帧间相关性为弱、 帧内相关性为弱时， 或者当前 MLT帧的帧间相关性为强、 帧内相关性为弱且前置 MLT帧的帧间相关 性为弱时： 将所述当前 MLT帧重叠相加时域帧的后半帧代替前置 MLT帧重叠 相加时域帧的后半帧， 对所述当前 MLT帧重叠相加时域帧的前半帧和 后半帧进行逆向 MLT变换得到当前 MLT帧最终时域帧的后半帧；或者， 将所述前置 MLT帧前一帧的重叠相加时域帧的后半帧代替前置 MLT帧 重叠相加时域帧的后半帧， 对所述当前 MLT帧重叠相加时域帧的前半 帧和所述前置 MLT帧前一帧的重叠相加时域帧的后半帧进行逆向 MLT 变换， 得到当前 MLT帧最终时域帧的后半帧；When the inter-frame correlation of the current MLT frame is weak and the intra-frame correlation is weak, or the inter-frame correlation of the current MLT frame is strong, the intra-frame correlation is weak, and the inter-frame correlation of the pre-MLT frame is weak. Time: Substituting the current MLT frame overlap and adding the second half of the time domain frame to replace the pre-MLT frame and adding the second half of the time domain frame, and adding and adding the first half and the second half of the time domain frame to the current MLT frame The frame is subjected to inverse MLT transformation to obtain the second half of the final time domain frame of the current MLT frame; or, the overlap of the previous frame of the pre-MLT frame is added to the second half of the time domain frame instead of the pre-MLT frame. The latter half of the domain frame, performing inverse MLT transformation on the first half of the current MLT frame and the first half of the pre-MLT frame and the second half of the pre-MLT frame The last half of the final time domain frame of the MLT frame;

重复得到的所述当前 MLT帧最终时域帧的后半帧，经衰减后作为当 前 MLT帧最终时域帧的前半帧； And repeating the obtained second frame of the final time domain frame of the current MLT frame, and attenuating the first half frame of the final time domain frame of the current MLT frame;

在当前 MLT帧的帧间相关性为强、 帧内相关性为强， 且前置 MLT 帧的帧间相关性为强时： When the inter-frame correlation of the current MLT frame is strong, the intra-frame correlation is strong, and the inter-frame correlation of the pre-MLT frame is strong:

重复所述前置 MLT帧最终时域帧的后半帧， 作为当前 MLT帧最终 时 i或帧的前半帧和后半帧； Repeating the second half of the final time domain frame of the pre-MLT frame as the first half frame and the second half frame of the current MLT frame;

在当前 MLT帧的帧间相关性为强、 帧内相关性为弱， 且前置 MLT 帧的帧间相关性为强时： When the inter-frame correlation of the current MLT frame is strong, the intra-frame correlation is weak, and the inter-frame correlation of the pre-MLT frame is strong:

重复所述前置 MLT帧最终时域帧的后半帧， 作为当前 MLT帧最终 时域帧的前半帧； Repeating the second half of the final time domain frame of the pre-MLT frame as the first half of the final time domain frame of the current MLT frame;

将所述当前 MLT帧重叠相加时域帧的后半帧代替前置 MLT帧重叠 相加时域帧的后半帧， 对所述当前 MLT帧重叠相加时域帧的前半帧和 后半帧进行逆向 MLT变换得到当前 MLT帧最终时域帧的后半帧；或者， 将所述前置 MLT帧前一帧的重叠相加时域帧的后半帧代替前置 MLT帧 重叠相加时域帧的后半帧， 对所述当前 MLT 帧重叠相加时域帧的前半 帧和所述前置 MLT帧前一帧的重叠相加时域帧的后半帧进行逆向 MLT 变换， 得到当前 MLT帧最终时域帧的后半帧。 Substituting the current MLT frame overlap and adding the second half of the time domain frame to replace the pre-MLT frame and adding the second half of the time domain frame, and adding and adding the first half and the second half of the time domain frame to the current MLT frame The frame is subjected to inverse MLT transformation to obtain the second half of the final time domain frame of the current MLT frame; or, the overlap of the previous frame of the pre-MLT frame is added to the second half of the time domain frame instead of the pre-MLT frame. The latter half of the domain frame, performing inverse MLT transformation on the first half of the current MLT frame overlap and the previous half of the pre-MLT frame and the second half of the pre-MLT frame The last half of the final time domain frame of the MLT frame.

9、 如权利要求 4所述的方法， 其特征在于， 所述恢复出错 /丢帧进 一步包括确定所述恢复出的错 /丢帧是否可靠， 生成前帧可靠性标识； 在前置 MLT帧和当前 MLT帧都为错 /丢帧时， 所述恢复出错 /丢帧 为：9. The method according to claim 4, wherein the recovery error/drop frame advances The step of determining whether the recovered error/drop frame is reliable, and generating a front frame reliability identifier; when both the pre-MLT frame and the current MLT frame are wrong/drop frames, the recovery error/drop frame is:

解码端 ^^据所述前帧可靠性标识、当前 MLT帧的帧间相关性信息和 当前 MLT帧的帧内相关性信息恢复出错 /丟帧。 The decoding end recovers the error/drop frame according to the previous frame reliability identifier, the inter-frame correlation information of the current MLT frame, and the intra-frame correlation information of the current MLT frame.

10、 如权利要求 9所述的方法， 其特征在于， 所述历史时域波形包 括前置 MLT帧最终时域帧； 10. The method of claim 9, wherein the historical time domain waveform comprises a final time domain frame of a pre-MLT frame;

解码端^^据所述前帧可靠性标识、 当前 MLT帧的帧间相关性信息、 当前 MLT帧的帧内相关性信息恢复出的错 /丢帧为当前 MLT帧最终时域 帧的前半帧和后半帧， 具体恢复过程为： The decoding end recovers the error/drop frame according to the previous frame reliability identifier, the inter-frame correlation information of the current MLT frame, and the intra-frame correlation information of the current MLT frame as the first field of the final time domain frame of the current MLT frame. And the second half of the frame, the specific recovery process is:

在前置 MLT帧最终时域帧为不可靠帧， 或者前置 MLT帧最终时域 帧为可靠帧且当前 MLT帧的帧间相关性为弱时： When the final time domain frame of the pre-MLT frame is an unreliable frame, or the final MLT frame final time domain frame is a reliable frame and the inter-frame correlation of the current MLT frame is weak:

重复所述前置 MLT帧最终时域帧， 经全帧衰减后作为当前 MLT帧 最终时域帧； 且得到的所述当前 MLT帧最终时域帧为不可靠帧； And repeating, the final time domain frame of the pre-MLT frame is used as the final time domain frame of the current MLT frame after being attenuated by the full frame; and the obtained final time domain frame of the current MLT frame is an unreliable frame;

前置 MLT帧最终时域帧为可靠帧、 当前 MLT帧的帧间相关性为强 且当前 MLT帧的帧内相关性为弱时： When the final time domain frame of the pre-MLT frame is a reliable frame, the inter-frame correlation of the current MLT frame is strong, and the intra-frame correlation of the current MLT frame is weak:

重复所述前置 MLT帧最终时域帧的后半帧， 作为当前 MLT帧最终 时域帧的前半帧； 重复前置 MLT帧最终时域帧的后半帧， 经衰减后作 为当前 MLT帧最终时域帧的后半帧； 且得到的所述当前 MLT帧最终时 域帧为不可靠帧； Repeating the second half of the final time domain frame of the pre-MLT frame as the first half of the final time domain frame of the current MLT frame; repeating the second half of the final time domain frame of the pre-MLT frame, and attenuating the final MLT frame as the final MLT frame The last half of the time domain frame; and the resulting final time domain frame of the current MLT frame is an unreliable frame;

在前置 MLT帧最终时域帧为可靠帧、 当前 MLT帧的帧间相关性为 强、 当前 MLT帧的帧内相关性为强时： When the final time domain frame of the pre-MLT frame is a reliable frame, the inter-frame correlation of the current MLT frame is strong, and the intra-frame correlation of the current MLT frame is strong:

重复所述前置 MLT帧最终时域帧的后半帧， 作为当前 MLT帧最终 时域帧的前半帧和后半帧； 且得到的所述当前 MLT帧最终时域帧为可 靠帧。And repeating the second half frame of the final time domain frame of the pre-MLT frame as the first half frame and the second half frame of the final time domain frame of the current MLT frame; and the obtained final time domain frame of the current MLT frame is a reliable frame.

11、 如权利要求 6、 8或 10所述的方法， 其特征在于， 所述重复所 述前置 MLT帧最终时域帧的后半帧为： 从所述前置 MLT帧最终时域帧 的后半帧的末尾向前选取一段基音， 并重复该段基音。The method according to claim 6, 8 or 10, wherein the repeating the last field of the final time domain frame of the pre-MLT frame is: from the final time domain frame of the pre-MLT frame At the end of the second half of the frame, a pitch is selected forward and the pitch is repeated.

12、 如权利要求 8所述的方法， 其特征在于， 所述重复得到的所述 当前 MLT帧最终时域帧的后半帧为： 从所述当前 MLT帧最终时域帧的 后半帧的前端向后选取一段基音， 并重复该段基音。 The method according to claim 8, wherein the second half of the final time domain frame of the current MLT frame obtained by the repetition is: from the last half of the final time domain frame of the current MLT frame The front end selects a pitch backwards and repeats the pitch.

13、 如权利要求 4所述的方法， 其特征在于， 13. The method of claim 4, wherein

所述根据当前帧进行帧内相关性检测， 生成帧内相关性信息包括： 当所述当前帧的帧内基音稳定、 基音周期性强， 则所述帧内相关性 强； 否则为弱； The performing the intra-correlation detection according to the current frame, and generating the intra-frame correlation information includes: when the intra-substation of the current frame is stable and the pitch is periodic, the intra-frame correlation is strong; otherwise, the intra-correlation is weak;

所述根据当前帧及其前置帧进行帧间相关性检测， 生成帧间相关性 信息包括： The performing inter-frame correlation detection according to the current frame and the pre-frame thereof, and generating inter-frame correlation information includes:

对所述当前帧的前半帧进行帧内相关性检测， 如果所述当前帧的前 半帧的帧内相关性为弱， 则所述帧间相关性为弱； 否则才 据所述当前帧 的前半帧和所述前置帧的后半帧进行帧间相关性检测， 检测结果为相关 性强， 则所述帧间相关性为强； 否则为弱。 Performing intra-correlation detection on the first half of the current frame, if the intra-frame correlation of the first frame of the current frame is weak, the inter-frame correlation is weak; otherwise, according to the first half of the current frame The frame and the second half of the pre-frame are inter-frame correlation detection, and if the detection result is strong correlation, the inter-frame correlation is strong; otherwise, it is weak.

14、 如权利要求 4所述的方法， 其特征在于， 所述相关帧信息进一 步包括短音符信息； The method according to claim 4, wherein the related frame information further includes short note information;

所述获得的相关帧信息以及当前 MLT帧进一步包括,根据所述后续 帧、 所述当前帧和所述前置帧， 生成所述当前 MLT帧的短音符信息。 The obtained related frame information and the current MLT frame further include: generating short note information of the current MLT frame according to the subsequent frame, the current frame, and the pre-frame.

15、 如权利要求 14所述的方法， 其特征在于， 所述恢复出错 /丢帧 为： 15. The method of claim 14, wherein the recovery error/drop frame is:

在当前 MLT帧为错 /丢帧，且根据所述短音符信息确定当前 MLT帧 为短音符帧时， 根据所述短音符信息恢复出错 /丢帧。 When the current MLT frame is a wrong/drop frame, and the current MLT frame is determined to be a short note frame according to the short note information, the error/drop frame is recovered according to the short note information.

16、如权利要求 15所述的方法， 其特征在于， 所述短音符信息包括 所述短音符标识用于确定当前 MLT帧是否为短音符；所述短音符特 征参数用于恢复错 /丢帧；The method according to claim 15, wherein said short note information comprises The short note identifier is used to determine whether the current MLT frame is a short note; the short note feature parameter is used to recover the wrong/lost frame;

所述^^据后续帧、当前帧和前置帧生成当前 MLT帧的短音符信息包 括： The short note information of the current MLT frame generated by the subsequent frame, the current frame, and the preamble frame includes:

根据所述后续帧、所述当前帧和所述前置帧生成所述当前 MLT帧的 短音符标识； 根据所述当前帧生成所述当前 MLT帧的短音符特征参数。 Generating a short note identifier of the current MLT frame according to the subsequent frame, the current frame, and the preamble frame; and generating a short note feature parameter of the current MLT frame according to the current frame.

17、 如权利要求 1所述的方法， 其特征在于， 所述恢复出错 /丢帧之 后进一步包括： 所述解码端将恢复出的错 /丢帧保存为历史时域波形。 The method according to claim 1, wherein the recovering the error/drop frame further comprises: the decoding end saving the recovered error/lost frame as a historical time domain waveform.

18、 如权利要求 1所述的方法， 其特征在于， 所述解码端对接收的 所述当前 MLT帧进行错 /丢帧检测后， 该方法包括： 判定前置 MLT帧和 当前 MLT帧都为正常帧， 则根据所述当前 MLT帧进行逆向 MLT解码， 获得当前 MLT帧重叠相加时域帧和当前 MLT帧最终时域帧， 并保存。 The method according to claim 1, wherein after the decoding end performs error/drop frame detection on the received current MLT frame, the method includes: determining that both the pre-MLT frame and the current MLT frame are The normal frame is then subjected to inverse MLT decoding according to the current MLT frame, and the current MLT frame overlap-added time domain frame and the current MLT frame final time domain frame are obtained and saved.

19、 一种丢帧隐藏编码装置， 其特征在于， 该编码装置包括 MLT 编码单元、 历史緩沖单元和相关帧信息生成单元； 19. A frame loss concealment coding apparatus, the coding apparatus comprising an MLT coding unit, a history buffer unit, and an associated frame information generating unit;

所述 MLT编码单元，用于根据该编码装置获取的后续帧和从所述历 史緩沖单元获取的当前帧进行 MLT变换， 生成 MLT帧， 并发送出去； 所述历史緩沖单元， 用于緩沖所述当前帧和所述前置帧； The MLT encoding unit is configured to perform MLT transformation according to a subsequent frame acquired by the encoding device and a current frame acquired from the history buffer unit, generate an MLT frame, and send the MLT frame, and send the data out to the history buffer unit. a current frame and the preamble frame;

所述相关帧信息生成单元， 用于^ ^据该编码装置获取的后续帧， 以 及从所述历史单元緩沖获取的所述当前帧和所述前置帧， 生成相关帧信 息， 并发送出去。 The related frame information generating unit is configured to generate related frame information according to a subsequent frame acquired by the encoding device, and the current frame and the pre-frame obtained from the history unit buffer, and send the related frame information.

20、如权利要求 19所述的编码装置， 其特征在于， 所述历史緩沖单 元包括当前帧历史緩沖和前置帧历史緩沖； The encoding apparatus according to claim 19, wherein the history buffer unit includes a current frame history buffer and a pre-frame history buffer;

所述当前帧历史緩沖， 用于緩沖所述当前帧， 并采用接收的后续帧 更新当前帧历史緩沖内容； 所述前置帧历史緩沖； 用于緩沖所述前置帧， 并采用接收自所述当 前帧历史緩沖的当前帧更新前置帧历史緩沖内容。The current frame history buffer is configured to buffer the current frame, and update the current frame history buffer content by using the received subsequent frame; The pre-frame history buffer is configured to buffer the pre-frame, and update the pre-frame history buffer content by using the current frame received from the current frame history buffer.

21、如权利要求 19所述的编码装置， 其特征在于， 所述相关帧信息 生成单元包括帧内相关性检测器和帧间相关性检测器； The encoding apparatus according to claim 19, wherein said related frame information generating unit comprises an intra-frame correlation detector and an inter-frame correlation detector;

所述帧内相关性检测器， 用于根据从所述当前帧历史緩沖中获取的 当前帧生成帧内相关性信息， 发送出去； The intra-frame correlation detector is configured to generate intra-frame correlation information according to the current frame acquired from the current frame history buffer, and send the information;

所述帧间相关性检测器， 用于根据从所述当前帧历史緩沖中获取的 当前帧，和从所述前置帧历史緩沖获取的前置帧，生成帧间相关性信息， 发送出去。 The inter-frame correlation detector is configured to generate inter-frame correlation information according to a current frame acquired from the current frame history buffer and a pre-frame obtained from the pre-frame history buffer, and send the information.

22、如权利要求 21所述的编码装置， 其特征在于， 所述相关帧信息 生成单元进一步包括短音符检测器， 用于根据所述后续帧、 从所述当前 帧历史緩沖中获取的当前帧和从所述前置帧历史緩沖获取的前置帧， 生 成短音符信息， 发送出去。 The encoding apparatus according to claim 21, wherein the related frame information generating unit further comprises a short note detector, configured to acquire a current frame from the current frame history buffer according to the subsequent frame And the pre-frame obtained from the pre-frame history buffer generates short note information and sends it out.

23、如权利要求 19所述的编码装置， 其特征在于， 该编码装置进一 步包括 MLT帧緩沖单元， 用于緩沖所述 MLT编码单元生成的 MLT帧， 延时一帧发送出去。 The encoding apparatus according to claim 19, wherein the encoding apparatus further comprises an MLT frame buffer unit for buffering the MLT frame generated by the MLT encoding unit, and transmitting one frame delayed.

24、如权利要求 19所述的编码装置， 其特征在于， 该编码装置进一 步包括相关帧信息緩沖单元， 用于緩沖所述相关帧信息生成单元生成的 相关帧信息， 延时一帧发送出去。 The encoding apparatus according to claim 19, wherein the encoding apparatus further comprises a correlation frame information buffering unit for buffering the associated frame information generated by the related frame information generating unit, and transmitting the frame delayed by one frame.

25、 一种丢帧隐藏解码装置， 其特征在于， 该解码装置包括帧错检 测器、 丢帧隐藏单元和解码波形历史緩沖； 25. A frame loss concealment decoding apparatus, comprising: a frame error detector, a frame loss concealing unit, and a decoding waveform history buffer;

所述解码波形历史緩沖， 用于存储历史时域波形； The decoded waveform history buffer is used to store a historical time domain waveform;

所述帧错检测器， 用于对当前 MLT帧进行错 /丢帧检测， 生成错 /丢 帧信息； 并将所述错 /丢帧信息发送给所述丢帧隐藏单元； The frame error detector is configured to perform error/drop frame detection on the current MLT frame, generate error/lost frame information, and send the error/drop frame information to the frame loss concealment unit;

所述丢帧隐藏单元，用于根据接收自所述帧错检测器的所述错 /丢帧 信息、 接收自编码装置的相关帧信息和所述解码波形历史緩沖中的历史 时域波形恢复出错 /丢帧。The frame loss concealing unit is configured to use the error/drop frame received from the frame error detector The information, the associated frame information received from the encoding device, and the historical time domain waveform in the decoded waveform history buffer recover errors/drop frames.

26、如权利要求 25所述的解码装置， 其特征在于， 所述解码波形历 史緩沖包括重叠相加时域帧历史緩沖和最终时域帧历史緩沖； The decoding apparatus according to claim 25, wherein said decoding waveform history buffer comprises an overlap-added time domain frame history buffer and a final time domain frame history buffer;

所述重叠相加时域帧历史緩沖， 用于緩沖重叠相加时域帧； 所述最终时域帧历史緩沖， 用于緩沖最终时域帧， 或緩沖所述丢帧 隐藏单元恢复的错 /丢帧。 The overlap-added time domain frame history buffer is used to buffer the overlap-added time domain frame; the final time domain frame history buffer is used to buffer the final time domain frame, or buffer the error of the lost frame hidden unit recovery/ Drop frame.

27、如权利要求 26所述的解码装置，其特征在于， 当所述当前 MLT 帧为正常帧时， 该解码装置进一步包括逆向 MLT解码模块； The decoding apparatus according to claim 26, wherein when the current MLT frame is a normal frame, the decoding apparatus further includes a reverse MLT decoding module;

所述帧错检测器进一步用于，当根据所述错 /丢帧信息判定当前 MLT 帧为正常帧时， 向所述逆向 MLT解码模块发送所述错 /丢帧信息和所述 当前 MLT帧； The frame error detector is further configured to: when determining that the current MLT frame is a normal frame according to the error/drop frame information, sending the error/drop frame information and the current MLT frame to the reverse MLT decoding module;

当逆向 MLT解码模块根据所述错 /丢帧信息判定，前置 MLT帧和当 前 MLT帧都为正常帧时， 所述逆向 MLT解码模块对所述当前 MLT帧 进行逆向 MLT变换， 将生成的当前 MLT帧重叠相加时域帧存入所述重 叠相加时域帧历史緩沖， 将生成的当前 MLT 帧最终时域帧存入所述最 终时域帧历史緩沖； 并输出所述当前 MLT帧最终时域帧； When the reverse MLT decoding module determines that the pre-MLT frame and the current MLT frame are normal frames, the reverse MLT decoding module performs inverse MLT transformation on the current MLT frame, and the generated current The MLT frame overlap-added time domain frame is stored in the overlap-added time domain frame history buffer, and the generated current MLT frame final time domain frame is stored in the final time domain frame history buffer; and the current MLT frame is finally output. Time domain frame

当逆向 MLT解码模块根据所述错 /丢帧信息判定，前置 MLT帧为错 /丢帧且当前 MLT帧为正常帧时， 所述逆向 MLT解码模块对所述当前 MLT帧进行逆向 DCT变换，将生成的当前 MLT帧重叠相加时域帧存入 所述重叠相加时域帧历史緩沖。 When the reverse MLT decoding module determines that the pre-MLT frame is a wrong/drop frame and the current MLT frame is a normal frame, the reverse MLT decoding module performs inverse DCT transform on the current MLT frame, The generated current MLT frame overlap-added time domain frame is stored in the overlap-added time domain frame history buffer.

28、如权利要求 26所述的解码装置，其特征在于， 当所述当前 MLT 帧为正常帧时， 该解码装置还包括逆向 MLT解码模块； The decoding apparatus according to claim 26, wherein when the current MLT frame is a normal frame, the decoding apparatus further includes an inverse MLT decoding module;

所述帧错检测器进一步用于， 根据所述错 /丢帧信息判定前置 MLT 帧和当前 MLT帧的错 /丢帧情况； 当所述帧错检测器判定，前置 MLT帧和当前 MLT帧都为正常帧时， 向所述逆向 MLT解码模块发送当前 MLT帧；The frame error detector is further configured to determine, according to the error/drop frame information, a wrong/drop frame condition of the pre-MLT frame and the current MLT frame; When the frame error detector determines that both the pre-MLT frame and the current MLT frame are normal frames, send the current MLT frame to the reverse MLT decoding module;

所述逆向 MLT解码模块对所述当前 MLT帧进行逆向 MLT变换， 将生成的当前 MLT帧重叠相加时域帧存入所述重叠相加时域帧历史緩 沖， 将生成的当前 MLT帧最终时域帧存入所述最终时域帧历史緩沖； 并输出所述当前 MLT帧最终时域帧； The reverse MLT decoding module performs inverse MLT transformation on the current MLT frame, and stores the generated current MLT frame overlap-added time domain frame into the overlap-added time domain frame history buffer, and the generated current MLT frame final time The domain frame is stored in the final time domain frame history buffer; and outputs the final time domain frame of the current MLT frame;

当所述帧错检测器判定，前置 MLT帧为错 /丢帧且当前 MLT帧为正 常帧时， 向所述逆向 MLT解码模块发送所述错 /丢帧信息和所述当前 MLT帧； When the frame error detector determines that the pre-MLT frame is a wrong/lost frame and the current MLT frame is a normal frame, transmitting the error/lost frame information and the current MLT frame to the reverse MLT decoding module;

所述逆向 MLT解码模块根据所述错 /丢帧信息，对所述当前 MLT帧 进行逆向 DCT变换， 将生成的当前 MLT帧重叠相加时域帧存入所述重 叠相加时域帧历史緩沖。 The reverse MLT decoding module performs inverse DCT transform on the current MLT frame according to the error/lost frame information, and stores the generated current MLT frame overlap-added time domain frame into the overlap-added time domain frame history buffer. .

29、如权利要求 25所述的解码装置， 其特征在于， 所述丢帧隐藏单 元包括相关帧信息历史緩沖、 丢帧隐藏策略选择模块和丢帧恢复模块； 所述相关帧信息历史緩沖，用于緩沖当前 MLT帧相关帧信息和前置 MLT帧相关帧信息； The decoding apparatus according to claim 25, wherein the frame loss concealing unit comprises an associated frame information history buffer, a frame loss concealment policy selection module, and a frame loss recovery module; Buffering the current MLT frame related frame information and the pre-MLT frame related frame information;

所述丢帧隐藏策略选择模块，用于存储解码装置根据所述错 /丢帧信 息， 所述相关帧信息和历史时域波形恢复错 /丢帧的隐藏策略； 用于根据 从所述帧错检测器接收的所述错 /丢帧信息，和从所述相关帧信息历史緩 沖接收的相关帧信息， 选择隐藏策略； 之后根据选择的隐藏策略更新前 帧可靠性标识； The frame loss concealment strategy selection module is configured to store, by the decoding device, a hidden strategy for recovering an error/drop frame according to the error/drop frame information, the related frame information and a historical time domain waveform; And selecting, by the detector, the error/drop frame information, and the related frame information received from the related frame information history buffer, selecting a hidden policy; and then updating the previous frame reliability identifier according to the selected hidden policy;

所述丢帧恢复模块， 用于根据从所述丢帧隐藏策略选择模块获取的 所述隐藏策略和所述解码波形历史緩沖中的历史时域波形恢复出错 /丢 帧。 The frame loss recovery module is configured to recover an error/lost frame according to the hidden policy acquired from the frame loss hiding policy selection module and the historical time domain waveform in the decoding waveform history buffer.

30、 如权利要求 29所述的解码装置， 其特征在于， 所述相关帧信息 包括短音符信息；The decoding device according to claim 29, wherein the related frame information Includes short note information;

所述丢帧隐藏策略选择模块进一步用于， 根据从所述帧错检测器接 收的所述错 /丢帧信息， 和从所述相关帧信息历史緩沖接收的短音符信 息， 判断当前 MLT帧是否为错 /丢帧且为短音符帧， 如果不是则为所述 当前 MLT帧选择隐藏策略； 如果是则将所述短音符信息发送给所述丢 帧恢复模块； The frame loss concealment policy selection module is further configured to: determine, according to the error/drop frame information received from the frame error detector, and the short note information received from the related frame information history buffer, whether the current MLT frame is Error/drop frame and short note frame, if not, select a hiding strategy for the current MLT frame; if yes, send the short note information to the frame loss recovery module;

所述丢帧恢复模块进一步用于， 根据所述短音符信息恢复出错 /丢 帧。 The frame loss recovery module is further configured to recover an error/drop frame according to the short note information.

31、 一种丢帧隐藏系统， 其特征在于， 该系统包括编码装置和解码 装置： 31. A frame loss concealment system, characterized in that the system comprises an encoding device and a decoding device:

所述编码装置，用于将获得的相关帧信息以及当前 MLT帧发送给解 码装置； The encoding device is configured to send the obtained related frame information and the current MLT frame to the decoding device;

所述解码装置， 用于对接收的所述当前 MLT帧进行错 /丢帧检测， 生成错 /丢帧信息； 根据所述错 /丢帧信息， 所述相关帧信息和历史时域 波形恢复出错 /丢帧。 The decoding device is configured to perform error/drop frame detection on the received current MLT frame, and generate error/drop frame information; according to the error/drop frame information, the related frame information and the historical time domain waveform recovery error / Drop frame.

32、如权利要求 31所述的系统， 其特征在于， 所述编码装置将所述 当前 MLT帧和所述相关帧信息分包发送给所述解码装置； 或者编码装 置将所述相关帧信息嵌在所述当前 MLT 帧的相邻帧发送给所述解码装 置； 或编码装置将所述相关帧信息作为增强层数据， 将所述当前 MLT 帧作为核心层数据发送给所述解码装置。 The system according to claim 31, wherein the encoding device transmits the current MLT frame and the related frame information to the decoding device, or the encoding device embeds the related frame information An adjacent frame of the current MLT frame is sent to the decoding device; or the encoding device uses the related frame information as enhancement layer data, and transmits the current MLT frame as core layer data to the decoding device.