CN102237093A - Echo hiding method based on forward and backward echo kernels - Google Patents

Abstract

Translated fromChinese

本发明公开了一种基于前后向回声核的回声隐藏方法。该方法包括水印嵌入部分和水印提取部分，水印嵌入部分包括对音频载体信号进行分帧的步骤，以及在每一帧音频信号中引入延时为d的前后向回声的步骤；水印提取部分包括利用倒谱法对每一帧音频信号中的水印进行检测的步骤，以及根据回声的延时d提取水印的步骤；水印嵌入部分还包括对音频载体信号每一帧之间加入长度为d的时间间隔的步骤；水印提取部分还包括去除音频载体信号每一帧之间时间间隔的步骤。相比现有技术，本发明通过在音频载体信号每一帧之间加入长度与回声延时相等的时间间隔，消除了现有技术所存在的水印嵌入过程中的信号缺失，极大地提高了传信率和恢复率。

The invention discloses an echo concealment method based on forward and backward echo kernels. The method comprises a watermark embedding part and a watermark extracting part, the watermark embedding part includes the steps of dividing the audio carrier signal into frames, and introducing a forward and backward echo with a delay of d in each frame of the audio signal; the watermark extracting part includes using The cepstrum method detects the watermark in each frame of the audio signal, and extracts the watermark according to the delay d of the echo; the watermark embedding part also includes adding a time interval of length d between each frame of the audio carrier signal The step; the watermark extraction part also includes the step of removing the time interval between each frame of the audio carrier signal. Compared with the prior art, the present invention eliminates the signal loss in the watermark embedding process existing in the prior art by adding a time interval equal to the echo delay between each frame of the audio carrier signal, and greatly improves the transmission efficiency. letter rate and recovery rate.

Description

Translated fromChinese

一种基于前后向回声核的回声隐藏方法An Echo Concealment Method Based on Backward and Backward Echo Kernels

技术领域technical field

本发明涉及一种回声隐藏方法，尤其涉及一种基于前后向回声核的回声隐藏方法。属于信息安全技术领域。The invention relates to an echo concealment method, in particular to an echo concealment method based on forward and backward echo kernels. It belongs to the field of information security technology.

背景技术Background technique

随着科学技术的发展，人们对信息安全的需求越来越急迫。在信息安全中，最重要的一点是防止攻击者的攻击。一些主动攻击者发现保密信息后会立即发动攻击。一般情况下，为了防止攻击者的攻击，通常采用信息隐藏的办法。所谓信息隐藏就是把保密的信息隐藏在非保密的媒体数据中（如声音，图像等），从而使得主动攻击者发现不了信息的隐藏，与此同时，要保证信息隐藏技术与加密技术并不相互排斥，这些被隐藏的信息称之为数字水印。从另一个方面来说，攻击者即使发现了隐蔽信息也需要经过检测和解密两个步骤才能得到隐藏信息。要提高一个语音系统的安全性，应该注意以下几个方面。第一.语音载体应该尽量保持嵌入保密信息前后听不出差别。这样的话，攻击者就无法确定是否有隐藏信息。第二.即使被窃听者发现有隐藏信息，也不能轻易得让他提取出隐藏信息。第三.对保密信息的加密方法不能被轻易破解。With the development of science and technology, people's demand for information security is becoming more and more urgent. In information security, the most important point is to prevent attacks from attackers. Some active attackers launch attacks as soon as they discover classified information. In general, in order to prevent attackers from attacking, information hiding is usually used. The so-called information hiding is to hide confidential information in non-confidential media data (such as sound, image, etc.), so that active attackers cannot discover the hiding of information. At the same time, it is necessary to ensure that information hiding technology and encryption technology are not mutually exclusive. Exclusion, these hidden information are called digital watermarks. On the other hand, even if the attacker discovers the hidden information, he needs to go through two steps of detection and decryption to get the hidden information. To improve the security of a voice system, you should pay attention to the following aspects. First, the voice carrier should try to keep the difference before and after embedding confidential information inaudible. In this way, the attacker cannot determine whether there is hidden information. Second. Even if the eavesdropper finds that there is hidden information, he cannot easily extract the hidden information. Third. The encryption method for confidential information cannot be easily cracked.

要研究避免攻击者进行攻击，首先要了解人耳听觉掩蔽特性。人耳的听觉掩蔽特性分为频率掩蔽特性和时间掩蔽特性。时间掩蔽特性是指一个信号可以被之前和之后发出的信号掩蔽。当时间差和频率差都很小时，暂时掩蔽和同时掩蔽分别增加。基于回声的信息隐藏系统就是基于人耳掩蔽特性来隐藏信息的。回声隐藏利用了人耳的听觉掩蔽效应，是一种有效的音频信息隐藏方法。其目的是以添加回声的方式在原有音频信息中嵌入新信息，实现信息隐藏。回声核的数学模型表示如公式(1)：To study how to prevent attackers from attacking, we must first understand the auditory masking characteristics of the human ear. The auditory masking characteristics of the human ear are divided into frequency masking characteristics and time masking characteristics. The temporal masking property means that a signal can be masked by signals sent before and after it. When both the time difference and the frequency difference are small, temporal masking and simultaneous masking increase, respectively. Echo-based information hiding systems hide information based on the masking characteristics of the human ear. Echo hiding takes advantage of the auditory masking effect of the human ear and is an effective audio information hiding method. Its purpose is to embed new information in the original audio information by adding echo to realize information hiding. The mathematical model of the echo nucleus is expressed as formula (1):

                                                 (1) (1)

嵌入回声的声音

可表示为

和

的卷积，

和

分别为原始声音信号和回声核的单位脉冲响应。回声信号由

引入到原始声音当中，其中，

为延迟时间，为衰减系数。嵌入回声后的声音信号表示如下：Embedded echo sound

can be expressed as

and

the convolution,

and

are the unit impulse responses of the original sound signal and the echo kernel, respectively. The echo signal is generated by

introduced into the original sound, where,

is the delay time, is the attenuation coefficient. The sound signal after embedding the echo is represented as follows:

                                                                                                                       (2) (2)

回声隐藏的具体方法是：对一段声音信号数据，先将其分成若干包含相同样点数的片段，每个片段时间约为几个到几十个ms，样点数记为N。每段用来嵌入1比特隐藏信息。在信息嵌入过程中，对每段信号使用式(3.2)，选择

，则在信号中嵌 入隐藏信息比特“0”；选择

,则在信号中嵌入隐藏信息比特“1”。延时

和是根据人耳听觉掩蔽效应为准则进行选取的。最后，将所有含有隐藏信息的声音信号段串联成连续信号。The specific method of echo hiding is: for a piece of sound signal data, first divide it into several segments containing the same number of samples, the time of each segment is about several to tens of ms, and the number of samples is recorded as N. Each segment is used to embed 1 bit of hidden information. In the process of information embedding, use formula (3.2) for each segment of signal, select

, then the hidden information bit "0" is embedded in the signal; choose

, the hidden information bit "1" is embedded in the signal. time delay

and It is selected according to the criterion of human auditory masking effect. Finally, all sound signal segments containing hidden information are concatenated into a continuous signal.

嵌入信息的提取实际上就是确定回声延时。由于每段隐写声音信号都是一个卷积性组合信号，直接从时域或频域确定回声延时存在一定困难，可采用卷积同态滤波系统来处理，将这个卷积性组合信号变为加性组合信号。Bender等人用倒谱分析的方法来确定回声延时。The extraction of embedded information is actually to determine the echo delay. Since each segment of the steganographic sound signal is a convolutional combination signal, it is difficult to determine the echo delay directly from the time domain or frequency domain. The convolutional homomorphic filtering system can be used to process this convolutional combination signal. is an additive combination signal. Bender et al. used cepstrum analysis to determine the echo delay.

对于声音信号

,其倒谱描述如下for sound signal

, whose cepstrum is described as follows

                                                         

                                                     (3)

(3)

其中，

和

分别表示傅立叶变换和傅立叶反变换。于是，式(3)可以表示为如下形式：in,

and

represent the Fourier transform and the inverse Fourier transform, respectively. Therefore, formula (3) can be expressed as the following form:

                                                                                    (4) (4)

公式(4)为分别计算

和

的倒谱，然后求和，即Formula (4) is calculated separately

and

The cepstrum of , and then summed, that is

                                                           

                                                        (5)

(5)

对

求倒谱：right

Find the cepstrum:

                                                         

                                                     (6)

(6)

其中，

。in,

.

由于

，又因衰减系数

，则：because

, and because of the attenuation coefficient

,but:

                                        

                                    (7)

(7)

所以，so,

                                   

                               (8)

(8)

因此，当回声核如式(8)时，嵌入回声后声音信号的倒谱表示如下：Therefore, when the echo kernel is as in formula (8), the cepstrum of the sound signal after embedding the echo is expressed as follows:

                                                      (9) (9)

式(9)中，

仅在

的整数倍处出现非零值，那么在信号的倒谱域

中，回声延迟处也会出现峰值，据此可确定嵌入回声的延迟大小，从而进一步确定隐藏信息是“0”或是“1”。In formula (9),

only at

Non-zero values appear at integer multiples of , then in the cepstrum domain of the signal

In, echo delay There will also be a peak at , from which the delay of the embedded echo can be determined, thereby further determining whether the hidden information is "0" or "1".

与其他音频信息隐藏方法(例如：LSB法，相位编码法，扩频调制法)相比，回声隐藏具有许多优点：①隐藏算法简单；②算法不产生噪声，隐藏效果好；③有时由于回声的引入，形成立体声效果，使声音听起来更加浑厚；④提取隐藏信息时不需要原始音频数据，能够实现隐藏信息的盲检测。但该方法的弱点也很明显，嵌入容量较小（一般秘密信息嵌入量为2b/s～64b/s,其大小与传输环境和参数设计有关）,提取率不是很让人满意，而且信道噪声、人为篡改都会降低提取准确率。Compared with other audio information hiding methods (for example: LSB method, phase coding method, spread spectrum modulation method), echo hiding has many advantages: ①The hiding algorithm is simple; ②The algorithm does not generate noise, and the hiding effect is good; ③Sometimes due to echo Introduce it to form a stereo effect and make the sound sound more vigorous; ④ No original audio data is needed when extracting hidden information, and blind detection of hidden information can be realized. But the weakness of this method is also obvious, the embedding capacity is small (generally, the embedding amount of secret information is 2b/s~64b/s, and its size is related to the transmission environment and parameter design), the extraction rate is not very satisfactory, and the channel noise , Human tampering will reduce the extraction accuracy.

不可感知性是包括回声隐藏在内的所有基于音频的信息隐藏理应具备的一个重要特性。在回声隐藏实现过程中，所引入回声的幅度直接决定了隐藏信息的不可感知性。Insensibility is an important property that all audio-based information hiding, including echo hiding, should possess. In the implementation process of echo hiding, the magnitude of the echo introduced directly determines the imperceptibility of the hidden information.

Xu等人（Xu. Proakis, Digital Communications. New York: McGraw-Hill,2001.）在Bender原始算法的基础上提出一种基于多回声核的回声隐藏方法。即用多回声核来引入多个小幅度回声来提高隐藏信息的不可感知性。但是，面对第三方的恶意攻击，该方法隐藏信息的恢复率并不理想，算法并不具备较强的鲁棒性。Xu et al.( Xu. Proakis, Digital Communications. New York: McGraw-Hill, 2001.) proposed an echo hiding method based on multiple echo kernels based on Bender's original algorithm. That is, the multi-echo kernel is used to introduce multiple small-amplitude echoes to improve the imperceptibility of hidden information. However, in the face of malicious third-party attacks, the recovery rate of hidden information in this method is not ideal, and the algorithm does not have strong robustness.

Oh在文献（Oh H2O, HyunWook Kim, JongWon Seok. Transparent and Robust Audio Watermarking with a New Echo Embedding Technique [ C ].ICME, 2001. 3172320.）中提出了一种基于双极性回声核的回声隐藏方法。该方法把人耳可感知的回声分为两部分echo和coloration。前者由于引入的回声延迟时间过大，影响听觉质量；后者听觉上可理解为对原始声音的赋色。Oh研究了在赋色域嵌入不同极性和个数的回声信号对原始声音所产生的影响。不同极性是针对式(10)中

的符号而言的。若

为正数则称回声为正极性；若

为负数则称回声为负极性。不同极性和个数的回声信号的频率响应是不同的。在音频信号中嵌入两个极性相反、不同延时的回声，能够增强回声的不可感知性。Huang发展了Oh的方法，提出了一种基于心理声学模型(Psychoacoustic model MPEG-1)分析的回声隐藏方法，进一步提高了算法的鲁棒性和回声的不可感知度。但是这类方法仍然存在缺陷：隐藏信息的恢复率较低，特别是引入回声幅度较小的时候。Oh proposed an echo hiding method based on a bipolar echo kernel in the literature( Oh H2O, HyunWook Kim, JongWon Seok. Transparent and Robust Audio Watermarking with a New Echo Embedding Technique [ C ]. ICME, 2001. 3172320.) . This method divides the echo perceivable by the human ear into two parts, echo and coloration. The former affects the auditory quality due to the excessive echo delay time introduced; the latter can be understood as the coloring of the original sound. Oh studied the effect of embedding different polarities and numbers of echo signals on the original sound in the color field. The different polarities are for the formula (10)

in terms of symbols. like

If it is a positive number, the echo is called positive polarity; if

If it is a negative number, the echo is called negative polarity. The frequency responses of echo signals with different polarities and numbers are different. Embedding two echoes with opposite polarities and different delays in the audio signal can enhance the imperceptibility of the echo. Huang developed Oh's method and proposed an echo hiding method based on the analysis of the psychoacoustic model (MPEG-1), which further improved the robustness of the algorithm and the imperceptibility of the echo. But this kind of method still has defects: the recovery rate of hidden information is low, especially when the amplitude of the echo introduced is small.

Kim提出了一种基于前后向回声核的回声隐藏方法（Hyoung Joong Kim, Yong Hee Choi. A Novel Echo2Hiding Scheme with Backward and Forward Kernels [ J ]. Circuits and Systems for Video Technology, 2003, 13 (8) : 8852889.）。在该方法中提出了一种新的回声“核”，该回声核由两个延时时间相同，但是方向相反的回声引入因子构成，称为前后向回声核。可表示为 Kim proposed an echo hiding method based on forward and backward echo kernels( Hyoung Joong Kim, Yong Hee Choi. A Novel Echo2Hiding Scheme with Backward and Forward Kernels [J]. Circuits and Systems for Video Technology, 2003, 13 (8): 8852889.). In this method, a new echo "kernel" is proposed. The echo kernel is composed of two echo introduction factors with the same delay time but opposite directions, which is called the front and rear echo kernels. can be expressed as

                                               

                                         (10)

(10)

其中

称后向回声引入因子，

称前向回声引入因子。Kim的研究结果表明，利用倒谱分析检测隐藏信息时，对应回声位置的峰值大小受三个因素的影响：①起最大决定作用的是回声核的倒谱响应

；②原始声音信号的倒谱响应

同样影响峰值大小；③延迟时间

的方向对峰值大小的影响也不容忽视。针对相同的原始声音信号，采用前后向回声核的隐藏算法在保证良好不可感知性的前提下能够获得较低的隐藏信息检测的误码率。但该方法在水印嵌入过程中，原始语音与回声叠加后，每一帧得到的信号是in

is called the backward echo introduction factor,

It is called the forward echo introduction factor. Kim's research results show that when cepstrum analysis is used to detect hidden information, the peak size corresponding to the echo position is affected by three factors: ① The cepstrum response of the echo kernel plays the most decisive role

; ② The cepstrum response of the original sound signal

Also affect the peak size; ③ delay time

The influence of the orientation of the peak on the peak size cannot be ignored. For the same original sound signal, the concealment algorithm using forward and backward echo kernels can obtain a lower bit error rate for hidden information detection under the premise of ensuring good imperceptibility. However, in the watermark embedding process of this method, after the original voice and the echo are superimposed, the signal obtained in each frame is

 

 

的语音信号，而理论上每帧信号应该为speech signal, and theoretically each frame signal should be

 

                  

 

                  

实际应用中比理论上少了

的部分。换而言之，这使得嵌入过程失去了理论上的精确性。Less in practice than in theory

part. In other words, this makes the embedding process lose its theoretical precision.

发明内容Contents of the invention

本发明所要解决的技术问题在于克服现有技术存在的水印嵌入时，每帧信号会部分缺失的不足，提供一种基于前后向回声核的回声隐藏方法，该方法能够消除水印嵌入过程中的信号缺失。The technical problem to be solved by the present invention is to overcome the deficiency that each frame signal will be partially missing when the watermark is embedded in the prior art, and provide an echo hiding method based on the forward and backward echo kernels, which can eliminate the signal in the process of watermark embedding missing.

具体而言，本发明采用以下技术方案：Specifically, the present invention adopts the following technical solutions:

一种基于前后向回声核的回声隐藏方法，包括水印嵌入部分和水印提取部分，水印嵌入部分包括对音频载体信号进行分帧的步骤，以及在每一帧音频信号中引入延时为d的前后向回声的步骤；水印提取部分包括利用倒谱法对每一帧音频信号中的水印进行检测的步骤，以及根据回声的延时d提取水印的步骤；An echo concealment method based on forward and backward echo kernels, including a watermark embedding part and a watermark extraction part, the watermark embedding part includes the step of dividing the audio carrier signal into frames, and introducing a delay of d into each frame of audio signal The step to the echo; the watermark extraction part includes the step of utilizing the cepstrum method to detect the watermark in each frame of the audio signal, and the step of extracting the watermark according to the delay d of the echo;

所述水印嵌入部分还包括对音频载体信号每一帧之间加入长度为d的时间间隔的步骤；The watermark embedding part also includes the step of adding a time interval of length d between each frame of the audio carrier signal;

所述水印提取部分还包括去除音频载体信号每一帧之间时间间隔的步骤。Said watermark extraction part also includes the step of removing the time interval between each frame of the audio carrier signal.

本发明通过在音频载体信号每一帧之间加入长度与回声延时相等的时间间隔，消除了现有技术所存在的水印嵌入过程中的信号缺失，极大地提高了传信率和恢复率。The invention eliminates the signal loss in the watermark embedding process in the prior art by adding a time interval equal to the echo delay between each frame of the audio carrier signal, and greatly improves the transmission rate and recovery rate.

附图说明Description of drawings

图1为本发明的水印嵌入原理示意图；Fig. 1 is a schematic diagram of the watermark embedding principle of the present invention;

图2为信元数变化时现有方法和本发明方法的恢复率比较，其中“——”为现有方法，“―·―”为本发明方法；Fig. 2 compares the recovery rate between the existing method and the method of the present invention when the number of cells changes, wherein "—" is the existing method, and "----" is the method of the present invention;

图3为衰减率变化时现有方法和本发明方法的恢复率比较，其中“——”为现有方法，“―·―”为本发明方法；Fig. 3 is the recovery rate comparison of the existing method and the method of the present invention when the attenuation rate changes, wherein "—" is the existing method, and "——" is the method of the present invention;

图4为延时点数变化时现有方法和本发明方法的恢复率比较，其中“——”为现有方法，“―·―”为本发明方法；Fig. 4 compares the recovery rate between the existing method and the method of the present invention when the number of delay points changes, wherein "—" is the existing method, and "----" is the method of the present invention;

图5为本发明方法的鲁棒性测试结果，其中A为原始水印，B1为原始水印通过现有方法传输并且不受攻击情况下在接收端得到的水印，B2为原始水印通过本发明方法传输并且在不受攻击情况下得到的水印，C1为原始水印通过现有方法传输并经高斯白噪声攻击在接收端所得到的水印，C2为原始水印通过本发明方法传输并经高斯白噪声攻击在接收端所得到的水印，D1为原始水印通过现有方法传输并经重采样攻击在接收端所得到的水印，D2为原始水印通过本发明方法传输并经重采样攻击在接收端所得到的水印，E1为原始水印通过现有方法传输并经重滤波攻击在接收端所得到的水印，E2为原始水印通过本发明方法传输并经重滤波攻击在接收端所得到的水印。Figure 5 is the robustness test result of the method of the present invention, where A is the original watermark, B1 is the watermark obtained at the receiving end when the original watermark is transmitted through the existing method and is not attacked, and B2 is the original watermark transmitted through the method of the present invention And the watermark obtained under the condition of not being attacked, C1 is the original watermark transmitted by the existing method and attacked by Gaussian white noise at the receiving end, and C2 is the original watermark transmitted by the method of the present invention and attacked by Gaussian white noise at the receiving end The watermark obtained at the receiving end, D1 is the original watermark transmitted by the existing method and the watermark obtained at the receiving end by resampling attack, D2 is the watermark obtained by the original watermark transmitted by the method of the present invention and obtained at the receiving end by resampling attack , E1 is the watermark obtained at the receiving end after the original watermark is transmitted through the existing method and subjected to heavy filtering attack, and E2 is the watermark obtained at the receiving end after the original watermark is transmitted through the method of the present invention and subjected to heavy filtering attack.

具体实施方式Detailed ways

下面结合附图对本发明的技术方案进行详细说明：The technical scheme of the present invention is described in detail below in conjunction with accompanying drawing:

假设x1（n）为原始音信号中的一帧信号，且

。Supposex 1(n ) is a frame signal in the original tone signal, and

.

在现有水印嵌入方式中，原始语音与回声叠加后，每一帧得到的信号是In the existing watermark embedding method, after the original voice and the echo are superimposed, the signal obtained in each frame is

 

 

的语音信号，而理论上每帧信号应该为speech signal, and theoretically each frame signal should be

 

                  

 

                  

实际应用中比理论上少了的部分。换而言之，这使得嵌入过程失去了理论上的精确性。Less in practice than in theory part. In other words, this makes the embedding process lose its theoretical precision.

在本发明的嵌入方式中，如图1所示，通过对音频载体信号每一帧之间加入长度为d的时间间隔，其中d为回声的延时，每一帧中得到In the embedding method of the present invention, as shown in Figure 1, by adding a time interval of length d between each frame of the audio carrier signal, where d is the delay of the echo, each frame obtains

                                                      

的信号，与理论上精确计算的值相同，避免了与理论上回声隐藏信号在数值上的误差。通过改进以后的嵌入方式虽然会在一定程度上降低声音质量，但是可以极大地提高传信率和恢复率。而且这种方式降低的声音质量在硬件实现时可以通过硬件的方式来弥补。The signal is the same as the theoretically accurately calculated value, avoiding the numerical error with the theoretical echo concealment signal. Although the sound quality will be reduced to a certain extent by improving the future embedding method, the transmission rate and recovery rate can be greatly improved. Moreover, the sound quality reduced in this way can be compensated by means of hardware during hardware implementation.

为了验证本发明的有效性，对现有方法及本发明方法进行了对比：In order to verify the effectiveness of the present invention, the existing method and the method of the present invention are compared:

一、性能对比：1. Performance comparison:

采用15s的wave格式歌曲s作为载体，采样频率22.05KHZ，分别用传统型和本发明方法进行传输。附图3—5显示了对比实验的结果。The song s in the wave format of 15s is used as the carrier, and the sampling frequency is 22.05KHZ, and the traditional type and the method of the present invention are respectively used for transmission. Accompanying drawing 3-5 has shown the result of comparative experiment.

以s1作为载体音频，取衰减系数为0.05，采用延时45点表示“1”，延时55点表示“0”，不断改变所传输信元数，比较传统方法和本发明方法的恢复率，所得结果见附图2。可以看出，当所传信元数少于30时，现有方法和本发明方法的恢复率都比较低，不满足实际需要；当所传信元数多于30时，本发明方法的恢复率明显高于现有方法；因此，当信元数不断变化时，本发明方法的恢复率性能比现有方法的更好。Take s1 as the carrier audio frequency, get the attenuation coefficient as 0.05, adopt 45 points of delay to represent "1", and 55 points of delay to represent "0", constantly change the number of transmitted cells, compare the recovery rate of the traditional method and the method of the present invention, The results obtained are shown in Figure 2. It can be seen that when the number of transmitted cells is less than 30, the recovery rate of the existing method and the method of the present invention is relatively low, which does not meet actual needs; when the number of transmitted cells is more than 30, the recovery rate of the method of the present invention is obvious higher than the existing method; therefore, when the number of cells is constantly changing, the recovery rate performance of the method of the present invention is better than that of the existing method.

以s1作为载体音频，传输的信元数取为340，采用延时45点表示“1”，延时55点表示“0”。不断改变衰减系数，比较传统方法和本发明方法的恢复率比较，所得结果见附图3。由图可知，当衰减系数在小于0.1时，恢复率性能差不多，但都低于0.7不能满足实际需要；当衰减系数在大于0.1时，本发明方法的恢复率性能好于现有方法，并且恢复率满足实际应用；因此，本发明方法比现有方法更能满足实际应用。Taking s1 as the carrier audio, the number of transmitted cells is taken as 340, a delay of 45 points is used to represent "1", and a delay of 55 points is used to represent "0". Constantly changing the attenuation coefficient, comparing the recovery rate of the traditional method and the method of the present invention, the results obtained are shown in Figure 3. As can be seen from the figure, when the attenuation coefficient is less than 0.1, the recovery rate performance is almost the same, but all less than 0.7 can not meet the actual needs; when the attenuation coefficient is greater than 0.1, the recovery rate performance of the method of the present invention is better than the existing method, and the restoration The rate satisfies the practical application; therefore, the method of the present invention can satisfy the practical application better than the existing method.

取载体音频为s1，所传的信元数为340。设采用延时n点表示“1”，延时n+10点表示“0”，不断改变n的大小，比较传统方法和本发明方法的恢复率，所得结果见图4。由图可以得到：本发明方法的恢复率性能好于现有方法，能够达到传统型不能达到的高恢复率；因此，本发明方法比现有方法更能满足实际中的应用需求。Take the carrier audio as s1, and the number of transmitted cells is 340. Assuming that "1" is represented by n points of delay, "0" is represented by n+10 points of delay, the size of n is constantly changed, and the recovery rate of the traditional method and the method of the present invention are compared, and the obtained results are shown in Figure 4. It can be obtained from the figure that the recovery rate performance of the method of the present invention is better than that of the existing method, and can reach a high recovery rate that cannot be achieved by the traditional method; therefore, the method of the present invention can better meet the actual application requirements than the existing method.

二．鲁棒性测试：two. Robustness test:

采用180s的wave格式歌曲作为载体，采样频率22.05KHZ，水印名取为“数字水印”，像素为99×95。对水印做各种攻击检测鲁棒性。其结果如图5所示，其中A为原始水印，B1为原始水印通过现有方法传输并且不受攻击情况下在接收端得到的水印，B2为原始水印通过本发明方法传输并且在不受攻击情况下得到的水印，C1为原始水印通过现有方法传输并经高斯白噪声攻击在接收端所得到的水印，C2为原始水印通过本发明方法传输并经高斯白噪声攻击在接收端所得到的水印，D1为原始水印通过现有方法传输并经重采样攻击在接收端所得到的水印，D2为原始水印通过本发明方法传输并经重采样攻击在接收端所得到的水印，E1为原始水印通过现有方法传输并经重滤波攻击在接收端所得到的水印，E2为原始水印通过本发明方法传输并经重滤波攻击在接收端所得到的水印。The 180s wave format song is used as the carrier, the sampling frequency is 22.05KHZ, the watermark name is "digital watermark", and the pixel size is 99×95. Robustness for detection of various attacks on watermarks. The result is shown in Figure 5, where A is the original watermark, B1 is the watermark obtained at the receiving end when the original watermark is transmitted by the existing method and is not attacked, and B2 is the original watermark transmitted by the method of the present invention and is not attacked The watermark obtained in the case, C1 is the original watermark transmitted by the existing method and the watermark obtained at the receiving end by Gaussian white noise attack, C2 is the original watermark transmitted by the method of the present invention and obtained by the Gaussian white noise attack at the receiving end Watermark, D1 is the original watermark transmitted by the existing method and obtained at the receiving end by resampling attack, D2 is the watermark obtained by the original watermark transmitted by the method of the present invention and obtained at the receiving end by resampling attack, E1 is the original watermark The watermark transmitted by the existing method and obtained at the receiving end through heavy filtering attack, E2 is the original watermark transmitted through the method of the present invention and obtained at the receiving end through heavy filtering attack.

选取100幅二值图像作为实验用图，取第70个图像作为水印嵌入到音频载体中，求取原水印和提取水印的归一化相关系数。NC值（归一化相关系数）是对提取水印和原始水印的相似度进行定量的评价。Select 100 binary images as the experimental images, take the 70th image as the watermark and embed it into the audio carrier, and calculate the normalized correlation coefficient between the original watermark and the extracted watermark. The NC value (normalized correlation coefficient) is a quantitative evaluation of the similarity between the extracted watermark and the original watermark.

表1  各种攻击下的NC值Table 1 NC values under various attacks

 theAAB1B1B2B2C1C1C2C2D1D1D2D2E1E1E2E2NC值NC value110.9570.957110.9570.9570.9990.9990.9320.9320.9940.9940.8170.8170.9900.990

从表1可以看出，在各种攻击下，本发明方法得到的NC值比传统型的高。因此本发明方法比现有方法在各种攻击下的恢复率高。It can be seen from Table 1 that under various attacks, the NC value obtained by the method of the present invention is higher than that of the traditional one. Therefore, the method of the present invention has a higher recovery rate under various attacks than the existing method.

表2为各种攻击下，所得到的SNR（信噪比）和PSNR（峰值信噪比），用来衡量算法的鲁棒性。Table 2 shows the obtained SNR (signal-to-noise ratio) and PSNR (peak signal-to-noise ratio) under various attacks, which are used to measure the robustness of the algorithm.

表2  各种攻击下的SNR和PSNR值Table 2 SNR and PSNR values under various attacks

 theAAB1B1B2B2C1C1C2C2D1D1D2D2E1E1E2E2SNRSNR∞∞11.7411.74∞∞11.7411.7438.8738.879.759.7519.6319.635.685.6817.7317.73PSNRPSNR87.8687.8611.6511.6587.8687.8611.6511.6539.7339.739.729.7220.3420.345.585.581818

从表2可以看出，在各种攻击下，本发明方法得到的NC值比传统型的高。因此本发明方法比现有方法从SNR和PSNR的角度来看具有更好的性能。It can be seen from Table 2 that under various attacks, the NC value obtained by the method of the present invention is higher than that of the traditional one. Therefore, the method of the present invention has better performance than the existing methods from the perspective of SNR and PSNR.

三、容纳能力测试3. Capacity test

在STEP 2001 音频信息隐藏评估测试中，将隐藏数据2 b /15 s的嵌入量视为满足声音作品拷贝控制的标准；72 b /30 s的隐藏信息嵌入量视为版权管理达标标准。实验中，对声音信号进行分段处理后每段包含52个采样点，而A、B、C、D、E和F这六段声音采样频率都为22.05 KHz，本发明方法的隐藏信息容纳能力为400 bps，远远超过STEP 2001标准。In the STEP 2001 audio information hiding evaluation test, the embedding amount of hidden data of 2 b /15 s is regarded as meeting the standard of copy control of sound works; the embedding amount of hidden information of 72 b /30 s is regarded as the standard of copyright management. In the experiment, after the sound signal is processed in sections, each section contains 52 sampling points, and the six sections of sound sampling frequency of A, B, C, D, E and F are all 22.05 KHz, the hidden information capacity of the inventive method It is 400 bps, far exceeding the STEP 2001 standard.

四、 主观评价4. Subjective evaluation

为了测试采用不同方法嵌入隐藏信息后的声音信号感知质量,进行了如下实验：取5段流行音乐A、B、C、D、E和F。对20名平均年龄为24岁无专业音乐背景的学生播放这五段原始声音信号及其对应的嵌入隐藏信息的声音信号，让他们区分两者之间的差别并按照主观区分度（Subjective Difference Grades, SDG）打分，取其平均值，结果见表3。In order to test the perceptual quality of the sound signal after embedding hidden information by different methods, the following experiments are carried out: 5 pieces of pop music A, B, C, D, E and F are taken. Play these five original sound signals and their corresponding sound signals embedded with hidden information to 20 students with an average age of 24 years and no professional music background, let them distinguish the difference between the two and according to the subjective difference (Subjective Difference Grades , SDG) and take the average value, the results are shown in Table 3.

表3  不可感知性的测试结果Table 3 Test results of imperceptibility

从实验结果可以看出，本发明方法相对现有方法对应的SDG差一些，但音频质量的降低还是在可以接受的范围之内。It can be seen from the experimental results that the SDG corresponding to the method of the present invention is worse than that of the existing method, but the degradation of the audio quality is still within an acceptable range.

Claims (1)

1. A echo hiding method based on a forward and backward echo core comprises a watermark embedding part and a watermark extracting part, wherein the watermark embedding part comprises a step of framing an audio carrier signal and a step of introducing forward and backward echoes with time delay d into each frame of audio signal; the watermark extraction part comprises a step of detecting the watermark in each frame of audio signal by using a cepstrum method and a step of extracting the watermark according to the delay d of the echo; it is characterized in that the preparation method is characterized in that,

the watermark embedding part further comprises a step of adding a time interval of length d between each frame of the audio carrier signal;

the watermark extraction section further comprises the step of removing the time interval between each frame of the audio carrier signal.

Images