技术领域technical field
本发明涉及一种面向宽带通信抗干扰的分布式压缩感知系统及方法，属于无线通信技术领域。The invention relates to a distributed compressed sensing system and method for broadband communication anti-interference, belonging to the technical field of wireless communication.
背景技术Background technique
随着信息技术的快速发展，各种电磁干扰技术被广泛应用于军用无线对抗，如何有效对抗电磁干扰是军用无线通信需要解决的关键问题。在通信对抗中，对抗双方都需要进行快速且科学的决策以最大化对抗收益。以跳频通信为代表的传统抗干扰通信技术主要通过提高跳频速率和增加跳频带宽来提高系统抗干扰能力，其抗干扰能力强且信号难以截获，因此建立的跳频通信网被广泛应用于通信领域。目前大部分军用无线电台都采用纯跳频技术来提高网络在通信受限环境下的抗人为干扰能力，然而传统的跳频通信抗干扰技术的抗跟踪干扰能力弱，自适应能力差，尤其在通信资源受限的动态环境以及动作空间较大的宽带跳频场景下难以应用。此外，随着无线干扰技术的逐渐智能化，固定策略的跳频抗干扰技术已无法取得动态抗干扰优化性能。With the rapid development of information technology, various electromagnetic interference technologies are widely used in military wireless countermeasures. How to effectively counter electromagnetic interference is a key problem that military wireless communication needs to solve. In communication confrontation, both sides of the confrontation need to make quick and scientific decisions to maximize the benefit of the confrontation. The traditional anti-interference communication technology represented by frequency hopping communication mainly improves the anti-interference ability of the system by increasing the frequency hopping rate and increasing the frequency hopping bandwidth. Its anti-interference ability is strong and the signal is difficult to intercept, so the established frequency hopping communication network is widely used in the field of communications. At present, most military radio stations use pure frequency hopping technology to improve the anti-jamming ability of the network in a communication-limited environment. It is difficult to apply in the dynamic environment with limited communication resources and broadband frequency hopping scenarios with large action space. In addition, with the gradual intelligence of wireless jamming technology, the frequency hopping anti-jamming technology with fixed strategy can no longer achieve dynamic anti-jamming optimization performance.
目前，针对上述智能化的干扰信号主要采用基于深度强化学习的方式进行抗干扰智能决策，具体可以参考论文：Xin Liu,etc.,“Anti-jamming Communications UsingSpectrum Waterfall:A Deep Reinforcement Learning Approach”,IEEE CommunicationLetters,vol.22,no.5,May.2018。该方法首先对用户和干扰的混合信号离散化处理，然后进行整体的感知得到频谱，以此作为环境状态，神经网络根据环境状态经过多次学习和训练后，可以输出最优的抗干扰策略。然而，该方法仅适用于理论中的窄带通信场景，其特征是在频谱感知中对高速A/D的要求较低，且动作空间小，有利于决策算法的学习。而对于实际中的宽带通信场景，其动作空间大，在频谱感知中对高速A/D的要求高，传统抗干扰决策算法难以满足这些要求，这使得其抗干扰性能明显下降，无法快速做出决策，难以应用于实际抗干扰场景。At present, for the above-mentioned intelligent jamming signals, the anti-jamming intelligent decision-making based on deep reinforcement learning is mainly used. For details, please refer to the paper: Xin Liu, etc., "Anti-jamming Communications Using Spectrum Waterfall: A Deep Reinforcement Learning Approach", IEEE Communication Letters, vol.22, no.5, May.2018. This method first discretizes the mixed signal of the user and the interference, and then conducts the overall perception to obtain the spectrum, which is used as the environment state. The neural network can output the optimal anti-interference strategy after multiple learning and training according to the environment state. However, this method is only applicable to narrow-band communication scenarios in theory, which are characterized by low requirements for high-speed A/D in spectrum sensing and small action space, which is conducive to the learning of decision-making algorithms. For the actual broadband communication scene, the action space is large, and the requirements for high-speed A/D in spectrum sensing are high, and the traditional anti-jamming decision-making algorithm is difficult to meet these requirements, which makes its anti-jamming performance decline significantly, and it is impossible to make a quick decision. Decision-making is difficult to apply to actual anti-jamming scenarios.
公开于该背景技术部分的信息仅仅旨在增加对本发明的总体背景的理解，而不应当被视为承认或以任何形式暗示该信息构成已为本领域普通技术人员所公知的现有技术。The information disclosed in this Background section is only for enhancement of understanding of the general background of the present invention and should not be taken as an acknowledgment or any form of suggestion that the information constitutes the prior art that is already known to those of ordinary skill in the art.
发明内容Contents of the invention
本发明的目的在于克服现有技术中的不足，提供一种面向宽带通信抗干扰的分布式压缩感知系统及方法，能有效降低宽带感知的高速A/D要求，用于智能抗干扰决策时没有明显性能损耗，使得传统抗干扰决策算法适用于宽带通信抗干扰场景。The purpose of the present invention is to overcome the deficiencies in the prior art, to provide a distributed compressed sensing system and method for broadband communication anti-jamming, which can effectively reduce the high-speed A/D requirements of broadband sensing, and when used for intelligent anti-jamming decision-making, there is no The obvious performance loss makes the traditional anti-jamming decision-making algorithm suitable for broadband communication anti-jamming scenarios.
为达到上述目的，本发明是采用下述技术方案实现的：In order to achieve the above object, the present invention is achieved by adopting the following technical solutions:
一方面，本发明公开了一种面向宽带通信抗干扰的分布式压缩感知系统，包括感知器、编码器、二值化模块、译码器和融合中心，所述感知器、编码器、二值化模块和译码器的数量相同，且所述感知器、编码器、二值化模块、译码器的数量为多个；一个感知器连接一个对应的编码器，一个编码器连接一个对应的二值化模块，一个二值化模块连接一个对应的译码器；多个译码器连接一个融合中心；On the one hand, the present invention discloses a distributed compressed sensing system oriented to broadband communication anti-jamming, including a perceptron, an encoder, a binarization module, a decoder, and a fusion center. The perceptron, encoder, binary The number of modules and decoders is the same, and the number of perceptrons, encoders, binarization modules, and decoders is multiple; one perceptron is connected to a corresponding encoder, and one encoder is connected to a corresponding Binarization module, a binarization module is connected to a corresponding decoder; multiple decoders are connected to a fusion center;
其中，所述感知器，被配置为，感知待感知的宽带信号的子带信号，得到子带的短时频谱；Wherein, the sensor is configured to sense the sub-band signal of the wideband signal to be sensed, and obtain the short-term frequency spectrum of the sub-band;
所述编码器，被配置为，根据所述子带的短时频谱，进行压缩编码操作，得到编码向量；The encoder is configured to perform a compression encoding operation according to the short-term frequency spectrum of the subband to obtain an encoding vector;
所述二值化模块，被配置为，根据所述编码向量，进行二值化操作，得到二元频谱信息压缩帧；The binarization module is configured to perform a binarization operation according to the encoding vector to obtain a binary spectrum information compressed frame;
所述译码器，被配置为，根据所述二元频谱信息压缩帧，进行频谱解码操作，得到子带的重建短时频谱；The decoder is configured to perform a spectrum decoding operation according to the binary spectrum information compressed frame to obtain a reconstructed short-term spectrum of the subband;
所述融合中心，被配置为，根据所述子带的重建短时频谱，得到宽带频谱，以实现对待感知的宽带信号的分布式感知和压缩重建。The fusion center is configured to obtain a broadband spectrum according to the reconstructed short-term spectrum of the sub-band, so as to realize distributed sensing and compressed reconstruction of the broadband signal to be sensed.
进一步的，所述待感知的宽带信号的频带范围为[FL，FU]，待感知的宽带信号的带宽为B＝FU-FL；Further, the frequency range of the broadband signal to be sensed is [FL ,FU ], and the bandwidth of the broadband signal to be sensed is B=FU -FL ;
感知器、编码器、二值化模块和译码器的数量相同，数量设定为M，M＞1；The number of perceptrons, encoders, binarization modules and decoders is the same, and the number is set to M, where M>1;
M个感知器独立分布，M个独立感知器的第m个感知器对应感知待感知的宽带信号的一个子带信号，即子频带(FL+(m-1)(FU-FL)/M，FL+m(FU-FL)/M)内的电磁频谱信号，得到一个对应的子带的短时频谱，m＝1，2，…，M。M perceptrons are independently distributed, and the mth perceptron of the M independent perceptrons corresponds to a subband signal of the wideband signal to be perceived, that is, the subband (FL + (m-1)(FU-FL ) /M, FL +m(FU -FL )/M) of the electromagnetic spectrum signal to obtain a short-time spectrum of a corresponding sub-band, m=1, 2, . . . , M.
进一步的，所述子带的短时频谱的表达式如下：Further, the expression of the short-time spectrum of the subband is as follows:
其中，表示第m个感知器在时间间隔[t，t+Δt]内得到的子带的短时频谱，1≤m≤M；/>表示第m个感知器得到的第l个离散频谱样本值，l＝1，2，…，L，l表示离散频谱样本值的索引，L表示短时频谱的长度。in, Indicates the short-term spectrum of the subband obtained by the mth perceptron in the time interval [t, t+Δt], 1≤m≤M; /> Indicates the lth discrete spectrum sample value obtained by the mth perceptron, l=1, 2, ..., L, l represents the index of the discrete spectrum sample value, and L represents the length of the short-time spectrum.
进一步的，所述离散频谱样本值的表达式如下：Further, the expression of the discrete spectrum sample value is as follows:
其中，log(·)表示对数函数；l表示离散频谱样本值的索引，l＝1，2，…，L，L表示短时频谱的长度；Δf表示频谱分辨率，Δf＝b/L，b表示子带带宽；fm表示子频带中心频率；是感知器接收端的功率谱密度，由自变量/>的功率谱密度函数计算得到；df表示对频率f的微分。Wherein, log ( ) represents a logarithmic function; l represents the index of the discrete spectrum sample value, l=1, 2,..., L, L represents the length of the short-time spectrum; Δf represents the spectral resolution, Δf=b/L, b represents the sub-band bandwidth; fm represents the center frequency of the sub-band; is the power spectral density at the receiving end of the perceptron, determined by the argument /> The power spectral density of The function is calculated; df represents the differential of the frequency f.
进一步的，所述编码器由深度神经网络搭建，包括依次连接的第一卷积块、第二卷积块、第三卷积块、第一全连接层、第二全连接层和第三全连接层；Further, the encoder is constructed by a deep neural network, including sequentially connected first convolutional block, second convolutional block, third convolutional block, first fully connected layer, second fully connected layer and third fully connected layer connection layer;
所述编码向量的表达式为：The expression of the encoding vector is:
其中，FE表示压缩编码操作；表示第m个感知器在时间间隔[t，t+Δt]内得到的子带的短时频谱；L′表示设定的编码向量压缩后的长度；/>表示第m个编码器得到的编码向量；Wherein, FE represents compression coding operation; Indicates the short-term spectrum of the subband obtained by the mth perceptron within the time interval [t, t+Δt]; L' indicates the length of the set coding vector after compression; /> Indicates the encoded vector obtained by the mth encoder;
其中，所述设定的编码向量压缩后的长度L′为48。Wherein, the compressed length L′ of the set encoding vector is 48.
进一步的，所述二元频谱信息压缩帧的表达式为：Further, the expression of the binary spectrum information compressed frame is:
其中，FB表示二值化操作；表示第m个二值化模块得到的二元频谱信息压缩帧。Among them, FB represents the binarization operation; Indicates the binary spectrum information compressed frame obtained by the mth binarization module.
进一步的，所述译码器由深度神经网络搭建，包括依次连接的第四全连接层、第五全连接层、第六全连接层、第四卷积块、第五卷积块、第六卷积块和一个第十卷积层；Further, the decoder is built by a deep neural network, including a fourth fully connected layer, a fifth fully connected layer, a sixth fully connected layer, a fourth convolutional block, a fifth convolutional block, and a sixth fully connected layer connected in sequence. convolutional block and a tenth convolutional layer;
所述子带的重建短时频谱的表达式为The expression of the reconstructed short-time spectrum of the subband is
其中，表示第m个译码器得到的子带的重建短时频谱；FD表示频谱解码操作。in, Represents the reconstructed short-term spectrum of the subband obtained by the m-th decoder; FD represents the spectral decoding operation.
进一步的，所述融合中心得到的宽带频谱st的表达式为：Further, the expression of the broadband spectrumst obtained by the fusion center is:
其中，表示第m个译码器得到的子带的重建短时频谱，m＝1,2,…,M。in, Represents the reconstructed short-time spectrum of the subband obtained by the mth decoder, m=1,2,...,M.
第二方面，本发明公开了一种面向宽带通信抗干扰的分布式压缩感知方法，适用于第一方面所述的面向宽带通信抗干扰的分布式压缩感知系统，所述方法包括：In the second aspect, the present invention discloses a distributed compressed sensing method for broadband communication anti-jamming, which is applicable to the distributed compressed sensing system for broadband communication anti-jamming described in the first aspect. The method includes:
基于所述分布式压缩感知系统的感知器，感知待感知的宽带信号的子带信号，得到子带的短时频谱；Based on the perceptron of the distributed compressed sensing system, the subband signal of the broadband signal to be sensed is sensed to obtain the short-term frequency spectrum of the subband;
基于所述分布式压缩感知系统的编码器，根据所述子带的短时频谱，进行压缩编码操作，得到编码向量；An encoder based on the distributed compressed sensing system performs a compression encoding operation according to the short-term frequency spectrum of the subband to obtain an encoding vector;
基于所述分布式压缩感知系统的二值化模块，根据所述编码向量，进行二值化操作，得到二元频谱信息压缩帧；Based on the binarization module of the distributed compressed sensing system, a binarization operation is performed according to the encoding vector to obtain a binary spectrum information compressed frame;
基于所述分布式压缩感知系统的译码器，根据所述二元频谱信息压缩帧，进行频谱解码操作，得到子带的重建短时频谱；Based on the decoder of the distributed compressed sensing system, compress the frame according to the binary spectrum information, perform a spectrum decoding operation, and obtain the reconstructed short-term spectrum of the sub-band;
基于所述分布式压缩感知系统的融合中心，根据所述子带的重建短时频谱，得到宽带频谱，以实现对待感知的宽带信号的分布式感知和压缩重建。Based on the fusion center of the distributed compressed sensing system, the wideband spectrum is obtained according to the reconstructed short-term spectrum of the subbands, so as to realize distributed sensing and compressed reconstruction of the wideband signal to be sensed.
与现有技术相比，本发明所达到的有益效果：Compared with the prior art, the beneficial effects achieved by the present invention are as follows:
本发明提出的面向宽带通信抗干扰的分布式压缩感知系统，通过感知器、编码器、二值化模块、译码器和融合中心的配合，能够有效降低宽带感知的高速采样要求，用于智能抗干扰决策时没有明显性能损耗，使传统抗干扰决策算法适用于宽带通信抗干扰场景。The anti-interference-oriented distributed compressed sensing system for broadband communication proposed by the present invention can effectively reduce the high-speed sampling requirements of broadband sensing through the cooperation of perceptrons, encoders, binarization modules, decoders and fusion centers, and is used for intelligent There is no obvious performance loss in anti-jamming decision-making, which makes the traditional anti-jamming decision-making algorithm suitable for broadband communication anti-jamming scenarios.
附图说明Description of drawings
图1为实施例提供的面向宽带通信抗干扰的分布式压缩感知系统的总体结构图；Fig. 1 is the overall structural diagram of the distributed compressive sensing system oriented to broadband communication anti-jamming provided by the embodiment;
图2为实施例提供的编码器、二值化模块和译码器的结构图；Figure 2 is a structural diagram of an encoder, a binarization module and a decoder provided by an embodiment;
图3为实施例提供的不同输出比特数下的频谱压缩失真的示意图；FIG. 3 is a schematic diagram of spectrum compression distortion under different output bit numbers provided by the embodiment;
图4为实施例提供的频谱压缩对抗干扰性能的归一化吞吐量示意图；Fig. 4 is a schematic diagram of the normalized throughput of spectrum compression anti-interference performance provided by the embodiment;
图5为实施例提供的宽带频谱在整个通信频带上的5种干扰模式的示意图。FIG. 5 is a schematic diagram of five interference modes of the wideband spectrum in the entire communication frequency band provided by the embodiment.
具体实施方式Detailed ways
下面结合附图对本发明作进一步描述。以下实施例仅用于更加清楚地说明本发明的技术方案，而不能以此来限制本发明的保护范围。The present invention will be further described below in conjunction with the accompanying drawings. The following examples are only used to illustrate the technical solution of the present invention more clearly, but not to limit the protection scope of the present invention.
实施例1Example 1
本实施例1提供了一种面向宽带通信抗干扰的分布式压缩感知系统，包括感知器、编码器、二值化模块、译码器和融合中心，感知器、编码器、二值化模块和译码器的数量相同，且感知器、编码器、二值化模块、译码器的数量为多个；一个感知器连接一个对应的编码器，一个编码器连接一个对应的二值化模块，一个二值化模块连接一个对应的译码器；多个译码器连接一个融合中心；Embodiment 1 provides a distributed compressed sensing system for broadband communication anti-interference, including a perceptron, an encoder, a binarization module, a decoder and a fusion center, a perceptron, an encoder, a binarization module and The number of decoders is the same, and there are multiple perceptrons, encoders, binarization modules, and decoders; one perceptron is connected to a corresponding encoder, and one encoder is connected to a corresponding binarization module. A binarization module is connected to a corresponding decoder; multiple decoders are connected to a fusion center;
其中，感知器，被配置为，感知待感知的宽带信号的子带信号，得到子带的短时频谱；Wherein, the sensor is configured to sense the subband signal of the wideband signal to be sensed, and obtain the short-term frequency spectrum of the subband;
编码器，被配置为，根据子带的短时频谱，进行压缩编码操作，得到编码向量；The encoder is configured to perform a compression encoding operation according to the short-term frequency spectrum of the subband to obtain an encoding vector;
二值化模块，被配置为，根据编码向量，进行二值化操作，得到二元频谱信息压缩帧；The binarization module is configured to perform a binarization operation according to the coding vector to obtain a binary spectrum information compressed frame;
译码器，被配置为，根据二元频谱信息压缩帧，进行频谱解码操作，得到子带的重建短时频谱；The decoder is configured to compress the frame according to the binary spectrum information, perform a spectrum decoding operation, and obtain the reconstructed short-term spectrum of the subband;
融合中心，被配置为，根据子带的重建短时频谱，得到宽带频谱，以实现对待感知的宽带信号的分布式感知和压缩重建。The fusion center is configured to obtain a wideband spectrum according to the reconstructed short-term spectrum of the subbands, so as to realize distributed sensing and compressed reconstruction of the wideband signal to be sensed.
本发明的技术构思为：针对传统抗干扰决策算法无法适用于宽带通信抗干扰场景，通过本发明的面向宽带通信抗干扰的分布式压缩感知系统，对宽带信号进行分布式感知，得到最终的宽带频谱，有效降低宽带感知的高速A/D要求，能够用于宽带通信抗干扰的决策，使得传统抗干扰决策算法适用于宽带通信抗干扰场景。The technical idea of the present invention is: for the traditional anti-jamming decision-making algorithm cannot be applied to the broadband communication anti-jamming scene, through the distributed compressed sensing system for broadband communication anti-jamming of the present invention, the broadband signal is distributed and sensed to obtain the final broadband communication Spectrum can effectively reduce the high-speed A/D requirements for broadband sensing, and can be used for broadband communication anti-jamming decisions, making traditional anti-jamming decision-making algorithms suitable for broadband communication anti-jamming scenarios.
如图1所示，本实施例首先根据干扰机模型和信道模型构建宽带通信抗干扰环境，根据分布式架构，采用多个感知器对宽带信号进行分布式感知，每个感知器只需感知一个子频带，得到子带频谱。然后采用基于编码器的频谱压缩技术对子带频谱进行压缩编码，得到由二进制比特流组成的二元频谱信息压缩帧。最后通过信息融合中心的译码器译码并融合后得到最终宽带频谱，可以用于宽带通信抗干扰的决策。As shown in Figure 1, this embodiment first builds a broadband communication anti-jamming environment based on the jammer model and the channel model. According to the distributed architecture, multiple sensors are used to perform distributed sensing on broadband signals, and each sensor only needs to sense one sub-band to get the sub-band spectrum. Then, the sub-band spectrum is compressed and coded by the coder-based spectrum compression technique, and a binary spectrum information compressed frame composed of a binary bit stream is obtained. Finally, it is decoded and fused by the decoder in the information fusion center to obtain the final broadband spectrum, which can be used for broadband communication anti-interference decision-making.
具体步骤如下。Specific steps are as follows.
S1：感知器感知待感知的宽带信号的子带信号，得到子带的短时频谱。S1: The sensor senses the sub-band signal of the broadband signal to be sensed, and obtains the short-term frequency spectrum of the sub-band.
S1.1：在本实施例中，待感知的宽带信号的频带范围为[FL，FU]，待感知的宽带信号的带宽为B＝FU-FL，FL和FU分别表示起始频率和截止频率。S1.1: In this embodiment, the frequency range of the broadband signal to be sensed is [FL ,FU ], and the bandwidth of the broadband signal to be sensed is B=FU -FL , whereFL andFU represent start frequency and cutoff frequency.
感知器、编码器、二值化模块和译码器的数量相同，数量设定为M，M＞1；The number of perceptrons, encoders, binarization modules and decoders is the same, and the number is set to M, where M>1;
M个感知器独立分布，M个独立感知器的第m个感知器对应感知待感知的宽带信号的一个带宽b＝B/M的子带信号，即子频带(FL+(m-1)(FU-FL)/M，FL+m(FU-FL)/M)内的电磁频谱信号，得到一个对应的子带的短时频谱，m＝1，2，…，M。M perceptrons are independently distributed, and the mth perceptron of the M independent perceptrons corresponds to a sub-band signal of a bandwidth b=B/M of the broadband signal to be perceived, that is, the sub-band (FL +(m-1) (FU -FL )/M, FL +m(FU -FL )/M) within the electromagnetic spectrum signal, get a short-time spectrum of a corresponding sub-band, m=1,2,...,M .
S1.2：在时域中，根据S1.1所述，感知器m(m＝1，2，…，M)在时间t接收到的第m个子带信号可以表示为：S1.2: In the time domain, according to S1.1, the mth sub-band signal received by the perceptron m (m=1, 2, ..., M) at time t can be expressed as:
其中x(t)和xj(t)分别表示用户和干扰机的基带信号，ωt和分别表示用户和干扰机的基带信号在t时刻的瞬时角频率，/>和/>分别表示用户和干扰机的基带信号的随机初始相位，n(t)表示加性随机噪声，gu，m是用户信号从发射机到第m个感知机的信道功率增益，gj，m是干扰信号从干扰机到第m个感知机的信道功率增益。where x(t) and xj (t) represent the baseband signals of the user and the jammer respectively, ωt and represent the instantaneous angular frequencies of the baseband signals of the user and the jammer at time t, respectively, /> and /> denote the random initial phases of the baseband signals of the user and the jammer respectively, n(t) denotes the additive random noise, gu,m is the channel power gain of the user signal from the transmitter to the mth sensor, gj,m is The channel power gain of the jamming signal from the jammer to the mth sensor.
S1.3、感知器m能感知的第m个子带信号的带宽为[fm-b/2，fm+b/2]，本地振荡器产生的子带中心频率为fm＝FL+(m-1/2)×b，低通滤波器的截止频率为b/2，ADC采样率为fs＝b。则根据步骤S1.2中ym(t)可以计算感知器m接收端的功率谱密度(PSD)函数S1.3. The bandwidth of the mth sub-band signal that sensor m can perceive is [fm -b/2, fm +b/2], and the center frequency of the sub-band generated by the local oscillator is fm =FL + (m-1/2)×b, the cutoff frequency of the low-pass filter is b/2, and the sampling rate of the ADC is fs =b. Then according to ym (t) in step S1.2, the power spectral density (PSD) function of the receiving end of sensor m can be calculated
其中，Ut(x-ft)表示基带信号的功率谱密度；表示干扰信号的功率谱密度；Nt(x)表示噪声信号的功率谱密度，功率谱密度可以通过P-Welch算法进行估计；ft表示t时刻基带信号的中心频率；/>表示t时刻干扰信号的中心频率；gu，m表示用户信号从发射机到第m个感知机的信道功率增益；gj，m表示干扰信号从干扰机到第m个感知机的信道功率增益；J表示干扰频点总数。Wherein, Ut (xft ) represents the power spectral density of the baseband signal; Represents the power spectral density of the interference signal; Nt (x) represents the power spectral density of the noise signal, which can be estimated by the P-Welch algorithm; ft represents the center frequency of the baseband signal at time t;/> Indicates the center frequency of the interference signal at time t; gu, m represents the channel power gain of the user signal from the transmitter to the mth sensor; gj, m represents the channel power gain of the interference signal from the jammer to the mth sensor ; J represents the total number of interference frequency points.
S1.4、根据步骤S1.3的功率谱密度函数使其自变量/>从而计算离散频谱样本值/>离散频谱样本值/>的表达式如下：S1.4, according to the power spectral density function of step S1.3 make its argument /> Thus computing the discrete spectrum sample values /> discrete spectrum sample values /> The expression of is as follows:
其中，log(·)表示对数函数；/表示离散频谱样本值的索引，l＝1，2，…，L，L表示短时频谱的长度；Δf表示频谱分辨率，Δf＝b/L，b表示子带带宽；fm表示子频带中心频率；是感知器接收端的功率谱密度，由步骤S1.3中的自变量/>的功率谱密度/>函数计算得到；df表示对频率f的微分。Wherein, log ( ) represents a logarithmic function; / represents the index of the discrete spectrum sample value, l=1, 2, ..., L, L represents the length of the short-time spectrum; Δf represents the spectral resolution, Δf=b/L, b represents the sub-band bandwidth; fm represents the center frequency of the sub-band; is the power spectral density at the receiving end of the perceptron, determined by the argument in step S1.3 /> Power Spectral Density /> The function is calculated; df represents the differential of the frequency f.
则在时间间隔[t，t+Δt]内感知到的第m个子带的短时频谱可以表示为：Then the short-term spectrum of the mth subband perceived in the time interval [t, t+Δt] can be expressed as:
其中，表示第m个感知器在时间间隔[t，t+Δt]内得到的子带的短时频谱，1≤m≤M；/>表示第m个感知器得到的第l个离散频谱样本值，l＝1，2，…，L，l表示离散频谱样本值的索引，L表示短时频谱的长度。in, Indicates the short-term spectrum of the subband obtained by the mth perceptron in the time interval [t, t+Δt], 1≤m≤M; /> Indicates the lth discrete spectrum sample value obtained by the mth perceptron, l=1, 2, ..., L, l represents the index of the discrete spectrum sample value, and L represents the length of the short-time spectrum.
至此完成对宽带信号的分布式感知。So far, the distributed sensing of broadband signals has been completed.
S2：编码器，根据子带的短时频谱，进行压缩编码操作，得到编码向量。S2: The encoder performs a compression encoding operation according to the short-term frequency spectrum of the sub-band to obtain an encoding vector.
如图2所示，编码器由深度神经网络搭建，包括依次连接的第一卷积块、第二卷积块、第三卷积块、第一全连接层、第二全连接层和第三全连接层；As shown in Figure 2, the encoder is built by a deep neural network, including the first convolutional block, the second convolutional block, the third convolutional block, the first fully connected layer, the second fully connected layer and the third convolutional block connected in sequence. fully connected layer;
其中，第一卷积块包括一个第一卷积层和一个第一最大池化层组成。第一卷积层的通道数为32，核为1×3，步长为1，填充为1；第一最大池化层的核为1×2，步长为1。Wherein, the first convolutional block includes a first convolutional layer and a first maximum pooling layer. The number of channels of the first convolutional layer is 32, the kernel is 1×3, the stride is 1, and the padding is 1; the kernel of the first maximum pooling layer is 1×2, and the stride is 1.
第二卷积块包括一个第二卷积层和一个第二最大池化层组成。第二卷积层的通道数为64，核为1×3，步长为1，填充为1；第二最大池化层的核为1×2，步长为1。The second convolutional block consists of a second convolutional layer and a second maximum pooling layer. The number of channels of the second convolutional layer is 64, the kernel is 1×3, the stride is 1, and the padding is 1; the kernel of the second maximum pooling layer is 1×2, and the stride is 1.
第三卷积块包括一个第三卷积层和一个第三最大池化层组成。第三卷积层的通道数为128，核为1×3，步长为1，填充为1；第三最大池化层的核为1×2，步长为1。The third convolutional block consists of a third convolutional layer and a third maximum pooling layer. The number of channels of the third convolutional layer is 128, the kernel is 1×3, the stride is 1, and the padding is 1; the kernel of the third maximum pooling layer is 1×2, and the stride is 1.
再经过一系列的卷积块，使用三个全连接层(Fc)来获得信息嵌入向量。在编码器中，第三全连接层的激活函数为tanh，维度为L′以生成L′个[-1，1]的输出，第一全连接层和第二全连接层的激活函数为ReLU，维度分别为1024和512。After a series of convolutional blocks, three fully connected layers (Fc) are used to obtain the information embedding vector. In the encoder, the activation function of the third fully connected layer is tanh, and the dimension is L' to generate L' outputs of [-1, 1]. The activation functions of the first fully connected layer and the second fully connected layer are ReLU , with dimensions 1024 and 512, respectively.
将感知器得到的子带的短时频谱作为编码器的输入，并输出长度为L′的编码向量，编码向量的表达式为：The short-term spectrum of the subband obtained by the perceptron As the input of the encoder, and output the encoding vector of length L′, the expression of the encoding vector is:
其中，FE表示压缩编码操作；表示第m个感知器在时间间隔[t，t+Δt]内得到的子带的短时频谱；L′表示设定的编码向量压缩后的长度；/>表示第m个编码器得到的编码向量；Wherein, FE represents compression coding operation; Indicates the short-term spectrum of the subband obtained by the mth perceptron within the time interval [t, t+Δt]; L' indicates the length of the set coding vector after compression; /> Indicates the encoded vector obtained by the mth encoder;
其中，设定的编码向量压缩后的长度L′为48。Wherein, the length L′ of the coded vector after compression is set to be 48.
S3：二值化模块，根据编码向量，进行二值化操作，得到二元频谱信息压缩帧。S3: The binarization module performs a binarization operation according to the encoding vector to obtain a binary spectrum information compressed frame.
将编码器输出的编码向量作为二值化模块的输入，得到二元频谱信息压缩帧，表达式如下：The encoding vector output by the encoder is used as the input of the binarization module to obtain a binary spectrum information compressed frame, the expression is as follows:
其中，FB表示二值化操作；表示第m个二值化模块得到的二元频谱信息压缩帧。Among them, FB represents the binarization operation; Indicates the binary spectrum information compressed frame obtained by the mth binarization module.
类似于AlexNet神经网络，本发明使用随机二值化技术将编码器的输出映射到L′个比特。二值化可以通过简单地对输出的比特数施加约束来控制压缩率，并且在无线信道上的比特向量的传输是可串行化且可解串行化的。随机二值化函数FB(x)，x∈[-1，1]定义为：Similar to the AlexNet neural network, the present invention uses a stochastic binarization technique to map the output of the encoder to L' bits. Binarization can control the compression rate by simply imposing a constraint on the number of output bits, and the transmission of the bit vector over the wireless channel is serializable and deserializable. The stochastic binarization function FB (x), x ∈ [-1, 1] is defined as:
FB(x)＝x+σ(x)∈{-1，1}FB (x) = x + σ (x) ∈ {-1, 1}
其中σ(x)是量化噪声，取值为1-x或1+x。Where σ(x) is the quantization noise, and the value is 1-x or 1+x.
至此完成对分布式感知后子带频谱的压缩。So far, the compression of the sub-band spectrum after distributed sensing is completed.
S4：译码器，根据二元频谱信息压缩帧，进行频谱解码操作，得到子带的重建短时频谱。S4: Decoder, which compresses the frame according to the binary spectrum information, performs spectrum decoding operation, and obtains the reconstructed short-time spectrum of the sub-band.
如图2所示，译码器由深度神经网络搭建，包括依次连接的第四全连接层、第五全连接层、第六全连接层、第四卷积块、第五卷积块、第六卷积块和一个第十卷积层。第四卷积块、第五卷积块、第六卷积块能够将三个全连接层生成的高阶表示转换成1×L的向量。As shown in Figure 2, the decoder is built by a deep neural network, including the fourth fully connected layer, the fifth fully connected layer, the sixth fully connected layer, the fourth convolutional block, the fifth convolutional block, and the fourth fully connected layer. Six convolutional blocks and a tenth convolutional layer. The fourth convolutional block, the fifth convolutional block, and the sixth convolutional block can convert the high-order representation generated by the three fully connected layers into a 1×L vector.
其中，第四全连接层、第五全连接层和第六全连接层的维度分别为512、1024和3200。Among them, the dimensions of the fourth fully connected layer, the fifth fully connected layer and the sixth fully connected layer are 512, 1024 and 3200 respectively.
第四卷积块包括一个第四卷积层和一个第五卷积层，第四卷积层的通道数为64，核为1×3，步长为1，填充为1；第五卷积层的通道数为64，核为1×3，步长为2，填充为1。The fourth convolutional block includes a fourth convolutional layer and a fifth convolutional layer. The number of channels of the fourth convolutional layer is 64, the kernel is 1×3, the step size is 1, and the padding is 1; the fifth convolutional layer The number of channels of the layer is 64, the kernel is 1×3, the stride is 2, and the padding is 1.
第五卷积块包括一个第六卷积层和一个第七卷积层，第六卷积层的通道数为32，核为1×3，步长为1，填充为1；第七卷积层的通道数为32，核为1×3，步长为2，填充为1。The fifth convolutional block includes a sixth convolutional layer and a seventh convolutional layer. The number of channels of the sixth convolutional layer is 32, the kernel is 1×3, the step size is 1, and the padding is 1; the seventh convolutional layer The number of channels of the layer is 32, the kernel is 1×3, the stride is 2, and the padding is 1.
第六卷积块包括一个第八卷积层和一个第九卷积层，第八卷积层的通道数为16，核为1×3，步长为1，填充为1；第九卷积层的通道数为16，核为1×3，步长为2，填充为1。The sixth convolutional block includes an eighth convolutional layer and a ninth convolutional layer. The number of channels of the eighth convolutional layer is 16, the kernel is 1×3, the step size is 1, and the padding is 1; the ninth convolutional layer The number of channels of the layer is 16, the kernel is 1×3, the stride is 2, and the padding is 1.
第十卷积层的通道数为1，核为1×3，步长为1，填充为1。The channel number of the tenth convolutional layer is 1, the kernel is 1×3, the stride is 1, and the padding is 1.
第四卷积层、第五卷积层、第六卷积层、第七卷积层、第八卷积层和第九卷积层的激活函数都采用ReLU，第十卷积层不使用激活函数。The activation functions of the fourth convolutional layer, the fifth convolutional layer, the sixth convolutional layer, the seventh convolutional layer, the eighth convolutional layer, and the ninth convolutional layer all use ReLU, and the tenth convolutional layer does not use activation function.
M个译码器接收到这些二元频谱信息压缩帧后，进行频谱译码，得到子带的重建短时频谱，表达式为After M decoders receive these binary spectrum information compressed frames, they decode the spectrum to obtain the reconstructed short-time spectrum of the subband, the expression is
其中，表示第m个译码器得到的子带的重建短时频谱；FD表示频谱解码操作；/>表示第m个译码器得到的第l个重建的离散频谱样本值，l＝1,2,…,L。in, Represents the reconstructed short-term spectrum of the subband obtained by the m-th decoder; FD represents the spectral decoding operation; /> Indicates the l-th reconstructed discrete spectrum sample value obtained by the m-th decoder, l=1,2,...,L.
S5：融合中心，根据子带的重建短时频谱，得到宽带频谱，以实现对待感知的宽带信号的分布式感知和压缩重建。S5: The fusion center obtains the broadband spectrum according to the reconstructed short-term spectrum of the sub-band, so as to realize the distributed sensing and compressed reconstruction of the broadband signal to be sensed.
融合中心得到的宽带频谱st的表达式为：The expression of the broadband spectrum st obtained by fusing the centers is:
其中，表示第m个译码器得到的子带的重建短时频谱，m＝1,2,…,M。in, Represents the reconstructed short-time spectrum of the subband obtained by the mth decoder, m=1,2,...,M.
如图1所示，融合中心得到的宽带频谱，可以作为环境状态输入到DRL网络并对其进行训练，DRL网络输出合适的频点给发射机，作为宽带通信抗干扰的决策。As shown in Figure 1, the broadband spectrum obtained by the fusion center can be used as an environmental state input to the DRL network for training, and the DRL network outputs appropriate frequency points to the transmitter as a decision for broadband communication anti-jamming.
需要说明的是，本实施例中的编码器和译码器都是预先构建并训练好的最优网络参数的编码器和译码器。It should be noted that the encoder and decoder in this embodiment are pre-built and trained with optimal network parameters.
具体的训练方法为，将均方误差损失(MSELoss)作为编码器和译码器的损失函数LMSE：The specific training method is to use the mean square error loss (MSELoss) as the loss function LMSE of the encoder and decoder:
其中N是训练样本的数量。然后可以使用梯度下降法根据损失函数更新编码器和译码器神经网络的参数，找到最优的网络参数。where N is the number of training samples. The parameters of the encoder and decoder neural networks can then be updated according to the loss function using the gradient descent method to find the optimal network parameters.
训练好的最优网络参数的编码器和译码器，能够最小化编码器引入的失真，以提高整个系统的精确性。The encoder and decoder with well-trained optimal network parameters can minimize the distortion introduced by the encoder to improve the accuracy of the whole system.
对于图1所示的分布式感知，每个感知器每隔Δt向信息融合中心发送一次压缩的窄带信号频谱。然后信息融合中心将重建窄带信号频谱并将这些频谱融合成宽带频谱。信息融合中心采用端到端训练程序，其输入是所有感知器收集的全频带短时频谱。通常压缩后的窄带频谱所需的通信开销远小于直接感知到的窄带频谱所需的开销，然而压缩可能导致重构的短时频谱出现严重的失真，并影响抗干扰性能。假设表示每个压缩的窄带短时频谱的比特长度为L′，则通信开销与L′成比例，频谱压缩的失真可以表示为：For the distributed sensing shown in Figure 1, each perceptron sends a compressed narrowband signal spectrum to the information fusion center every Δt. The information fusion center will then reconstruct the narrowband signal spectrum and fuse these spectrums into wideband spectrum. The information fusion center employs an end-to-end training procedure whose input is the full-band short-term spectrum collected by all perceptrons. Generally, the communication overhead required by the compressed narrowband spectrum is much smaller than that required by the directly perceived narrowband spectrum. However, the compression may cause severe distortion in the reconstructed short-term spectrum and affect the anti-jamming performance. Assuming that the bit length of each compressed narrow-band short-term spectrum is L', the communication overhead is proportional to L', and the distortion of spectrum compression can be expressed as:
其中是从L′个比特重建的/>中的第l个元素。in is reconstructed from L′ bits /> The lth element in .
显然，随着L′的减小，失真增加，这使得抗干扰性能下降。为了探索压缩率和失真之间的关系，本发明在同一数据集上训练了几个输出长度L′不同的编码器，，并计算每个频谱压缩模型的平均失真。图3显示了L′对频谱压缩失真的影响。可以看出，随着编码器输出比特数的增加，失真减小。当L′≥48时，失真不再显著降低。因此，所需输出比特数的最优选择是L′＝48，此时的失真约为0.06。Obviously, as L' decreases, the distortion increases, which degrades the anti-interference performance. In order to explore the relationship between compression rate and distortion, the present invention trains several encoders with different output length L' on the same data set, and calculates the average distortion of each spectrum compression model. Figure 3 shows the effect of L' on spectral compression distortion. It can be seen that as the number of encoder output bits increases, the distortion decreases. When L'≥48, the distortion is no longer significantly reduced. Therefore, the optimal choice of the required output bit number is L'=48, and the distortion at this time is about 0.06.
请参阅图4，为了研究使用重建的短时频谱对所提出方案的抗干扰性能的影响，使用L′＝48比特来重建短时频谱。定义归一化吞吐量为：Please refer to FIG. 4 , in order to study the impact of using the reconstructed short-time spectrum on the anti-jamming performance of the proposed scheme, L'=48 bits are used to reconstruct the short-time spectrum. Define the normalized throughput as:
其中S0是初始状态，π*是最优的策略。whereS0 is the initial state and π* is the optimal policy.
图4给出了使用无压缩感知的信号、用4bit量化的信号和用L′＝48的分布式压缩感知系统的信号训练的基于深度强化学习(Deep Reinforcement Learning，DRL)的抗干扰归一化吞吐量对比图。仿真结果表明，随着训练迭代次数的增加，三种场景下的归一化吞吐量逐渐收敛到一个相同的上限0.98。这意味着直接量化方法和我们提出的L′＝48的分布式压缩感知系统都不会导致任何明显的性能下降。同时，本发明所提出的分布式压缩感知系统的通信开销比直接量化方法的通信开销低了(4×L)/L′≈17倍。Figure 4 shows the anti-interference normalization based on Deep Reinforcement Learning (DRL) using the signal without compressed sensing, the signal quantized with 4bit and the signal trained with the distributed compressed sensing system with L'=48 Throughput comparison graph. The simulation results show that as the number of training iterations increases, the normalized throughput under the three scenarios gradually converges to a same upper limit of 0.98. This means that neither the direct quantization method nor our proposed distributed compressed sensing system with L′=48 will lead to any noticeable performance degradation. At the same time, the communication overhead of the distributed compressed sensing system proposed by the present invention is (4×L)/L′≈17 times lower than that of the direct quantization method.
对于图4中的结果，本实施案例中使用了一种基于深度Q学习(Deep Q-learningNetwork，DQN)的DRL抗干扰决策算法。DQN通常使用深度卷积神经网络来拟合动作值：For the results in Figure 4, a DRL anti-interference decision-making algorithm based on Deep Q-learning Network (DQN) is used in this implementation case. DQN typically uses deep convolutional neural networks to fit action values:
其中是即时奖励值，θ是网络参数。网络由两个具有16和32个滤波器的Conv层、两个具有2048和50个节点的Fc层实现，除输出层之外，每一层都采用ReLU激活函数。in is the immediate reward value, and θ is the network parameter. The network is implemented by two Conv layers with 16 and 32 filters, two Fc layers with 2048 and 50 nodes, and ReLU activation function is adopted in each layer except the output layer.
为了更新DQN网络，智能体将收集转移并存储在经验回放池中，然后随机选择一个小批量的转移计算损失：To update the DQN network, the agent will collect the transfer And store in the experience replay pool, and then randomly select a small batch of transfer calculation loss:
在仿真中，用户和干扰机都工作在[100MHz,200MHz]的频带上，即带宽B＝100MHz。分布式感知系统中有M＝5个感知器，每个感知器以Δf＝100kHz(即L＝b/Δf＝200)进行部分频带(带宽b＝b/M＝20MHz)感知，并每Δt＝1ms向信息融合中心发送长度为L′的压缩子频带短时频谱。智能体在T＝200ms(即NT＝T/ΔT＝200)内保留重建的频谱数据。因此，频谱瀑布Sn的大小为1000×200。合法信号的带宽设置为2MHz，其中心频率以2MHz的步长每τ＝10ms(即Nτ＝τ/Δt＝10)跳变一次，动作的数量为|A|＝50。智能体可以在每个时隙从A＝{101Mhz,103Mhz,...,199Mhz}中选择动作。频率切换成本设置为λ＝0.05。合法信号和干扰信号均采用滚降系数为α＝0.4的根升余弦脉冲整形滤波器进行整形，合法信号的发射功率设置为0dBm。In the simulation, both the user and the jammer work in the [100MHz, 200MHz] frequency band, that is, the bandwidth B=100MHz. There are M=5 perceptrons in the distributed sensing system, and each perceptron performs partial frequency band (bandwidth b=b/M=20MHz) sensing with Δf=100kHz (that is, L=b/Δf=200), and every Δt= 1 ms to send the compressed sub-band short-time spectrum of length L' to the information fusion center. The agent retains the reconstructed spectrum data within T=200ms (ie NT=T/ΔT=200). Therefore, the size of the spectral waterfall Sn is 1000×200. The bandwidth of the legal signal is set to 2MHz, its center frequency jumps once every τ=10ms (ie Nτ=τ/Δt=10) with a step size of 2MHz, and the number of actions is |A|=50. The agent can choose an action from A={101Mhz, 103Mhz,...,199Mhz} at each time slot. The frequency switching cost is set to λ=0.05. Both the legal signal and the interference signal are shaped by a root-raised cosine pulse shaping filter with a roll-off coefficient of α=0.4, and the transmission power of the legal signal is set to 0dBm.
为了评估所提出的分布式感知架构中基于深度强化学习的频率选择方案的抗干扰性能，在训练过程中的超参数设置如下：折扣因子为γ＝0.95，学习率为β＝10-4，转移采样批次大小为mini-batch＝256，经验回放池的大小为1000，目标网络的更新频率为1000。In order to evaluate the anti-jamming performance of the frequency selection scheme based on deep reinforcement learning in the proposed distributed sensing architecture, the hyperparameters during the training process are set as follows: the discount factor is γ=0.95, the learning rate is β=10-4 , and the transfer The sampling batch size is mini-batch=256, the size of the experience playback pool is 1000, and the update frequency of the target network is 1000.
图4可以看出，随着训练迭代次数的增加，本发明所提出方案的归一化吞吐量也可以逐渐收敛到0.98的上限，本发明所提出的方法的抗干扰性能非常接近使用理想频谱和基于SINR的奖励的抗干扰方案。It can be seen from Fig. 4 that as the number of training iterations increases, the normalized throughput of the scheme proposed by the present invention can also gradually converge to the upper limit of 0.98, and the anti-jamming performance of the method proposed by the present invention is very close to using the ideal spectrum and An anti-jamming scheme for SINR-based rewards.
图5给出了本实施案例1中宽带频谱在整个通信频带上的5种干扰模式，具体如下：Figure 5 shows five interference modes of the wideband spectrum in the entire communication frequency band in this implementation case 1, as follows:
(1)全频段干扰：干扰信号的带宽为整个选定频带，干扰功率为40dBm；(1) Full frequency band interference: the bandwidth of the interference signal is the entire selected frequency band, and the interference power is 40dBm;
(2)扫频干扰：扫频速度为0.5GHz/s，干扰功率为50dBm；(2) Frequency sweep interference: frequency sweep speed is 0.5GHz/s, interference power is 50dBm;
(3)单音干扰：干扰带宽为5MHz，干扰功率为40dBm，干扰机每20ms改变一次中心频率；(3) Single-tone interference: the interference bandwidth is 5MHz, the interference power is 40dBm, and the jammer changes the center frequency every 20ms;
(4)开关梳状干扰：每隔2MHz发射一个干扰信号，干扰信号每100ms改变中心频率，干扰功率为40dBm；(4) Switch comb interference: transmit an interference signal every 2MHz, the interference signal changes the center frequency every 100ms, and the interference power is 40dBm;
(5)跟随干扰：干扰中心频率与用户最近的通信频率相同，如果用户信号未出现在跟随干扰的范围内，则干扰机选择随机频率进行干扰。干扰带宽为5MHz，干扰功率为50dBm。(5) Follow-up interference: The center frequency of interference is the same as the user's nearest communication frequency. If the user signal does not appear within the range of follow-up interference, the jammer selects a random frequency for interference. The interference bandwidth is 5MHz, and the interference power is 50dBm.
通信频带被划分为五个20Mhz的子频带(即100MHz～120MHz，120MHz～140MHz，140MHz～160MHz，160MHz～180MHz和180MHz～200MHz)，每个干扰模式影响一个子频带。所以本申请采用了面向宽带通信抗干扰的分布式压缩感知系统，针对不同的干扰模式，能有效降低宽带感知的高速采样要求，用于智能抗干扰决策时没有明显性能损耗。The communication frequency band is divided into five 20Mhz sub-bands (ie, 100MHz-120MHz, 120MHz-140MHz, 140MHz-160MHz, 160MHz-180MHz and 180MHz-200MHz), and each interference pattern affects one sub-band. Therefore, this application adopts a distributed compressed sensing system for broadband communication anti-jamming, which can effectively reduce the high-speed sampling requirements of broadband sensing for different interference modes, and has no obvious performance loss when used for intelligent anti-jamming decision-making.
实施例2Example 2
本实施例提供了一种面向宽带通信抗干扰的分布式压缩感知方法，适用于实施例1的面向宽带通信抗干扰的分布式压缩感知系统，方法包括：This embodiment provides a distributed compressed sensing method for broadband communication anti-interference, which is applicable to the distributed compressed sensing system for broadband communication anti-interference in Embodiment 1. The method includes:
基于分布式压缩感知系统的感知器，感知待感知的宽带信号的子带信号，得到子带的短时频谱；Based on the perceptron of the distributed compressed sensing system, the subband signal of the broadband signal to be sensed is sensed, and the short-term spectrum of the subband is obtained;
基于分布式压缩感知系统的编码器，根据子带的短时频谱，进行压缩编码操作，得到编码向量；The encoder based on the distributed compressed sensing system performs compression encoding operation according to the short-term frequency spectrum of the sub-band to obtain the encoding vector;
基于分布式压缩感知系统的二值化模块，根据编码向量，进行二值化操作，得到二元频谱信息压缩帧；Based on the binarization module of the distributed compressed sensing system, the binarization operation is performed according to the coding vector to obtain the binary spectrum information compressed frame;
基于分布式压缩感知系统的译码器，根据二元频谱信息压缩帧，进行频谱解码操作，得到子带的重建短时频谱；Based on the decoder of the distributed compressed sensing system, the frame is compressed according to the binary spectrum information, and the spectrum decoding operation is performed to obtain the reconstructed short-term spectrum of the sub-band;
基于分布式压缩感知系统的融合中心，根据子带的重建短时频谱，得到宽带频谱，以实现对待感知的宽带信号的分布式感知和压缩重建。Based on the fusion center of the distributed compressed sensing system, the broadband spectrum is obtained according to the reconstructed short-term spectrum of the sub-bands, so as to realize the distributed sensing and compressed reconstruction of the broadband signal to be sensed.
本领域内的技术人员应明白，本申请的实施例可提供为方法、系统、或计算机程序产品。因此，本申请可采用完全硬件实施例、完全软件实施例、或结合软件和硬件方面的实施例的形式。而且，本申请可采用在一个或多个其中包含有计算机可用程序代码的计算机可用存储介质(包括但不限于磁盘存储器、CD-ROM、光学存储器等)上实施的计算机程序产品的形式。Those skilled in the art should understand that the embodiments of the present application may be provided as methods, systems, or computer program products. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) having computer-usable program code embodied therein.
本申请是参照根据本申请实施例的方法、设备(系统)、和计算机程序产品的流程图和/或方框图来描述的。应理解可由计算机程序指令实现流程图和/或方框图中的每一流程和/或方框、以及流程图和/或方框图中的流程和/或方框的结合。可提供这些计算机程序指令到通用计算机、专用计算机、嵌入式处理机或其他可编程数据处理设备的处理器以产生一个机器，使得通过计算机或其他可编程数据处理设备的处理器执行的指令产生用于实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能的装置。The present application is described with reference to flowcharts and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the present application. It should be understood that each procedure and/or block in the flowchart and/or block diagram, and a combination of procedures and/or blocks in the flowchart and/or block diagram can be realized by computer program instructions. These computer program instructions may be provided to a general purpose computer, special purpose computer, embedded processor, or processor of other programmable data processing equipment to produce a machine such that the instructions executed by the processor of the computer or other programmable data processing equipment produce a An apparatus for realizing the functions specified in one or more procedures of the flowchart and/or one or more blocks of the block diagram.
这些计算机程序指令也可存储在能引导计算机或其他可编程数据处理设备以特定方式工作的计算机可读存储器中，使得存储在该计算机可读存储器中的指令产生包括指令装置的制造品，该指令装置实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能。These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing apparatus to operate in a specific manner, such that the instructions stored in the computer-readable memory produce an article of manufacture comprising instruction means, the instructions The device realizes the function specified in one or more procedures of the flowchart and/or one or more blocks of the block diagram.
这些计算机程序指令也可装载到计算机或其他可编程数据处理设备上，使得在计算机或其他可编程设备上执行一系列操作步骤以产生计算机实现的处理，从而在计算机或其他可编程设备上执行的指令提供用于实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能的步骤。These computer program instructions can also be loaded onto a computer or other programmable data processing device, causing a series of operational steps to be performed on the computer or other programmable device to produce a computer-implemented process, thereby The instructions provide steps for implementing the functions specified in the flow chart or blocks of the flowchart and/or the block or blocks of the block diagrams.
以上所述仅是本发明的优选实施方式，应当指出：对于本技术领域的普通技术人员来说，在不脱离本发明原理的前提下，还可以做出若干改进和润饰，这些改进和润饰也应视为本发明的保护范围。The above is only a preferred embodiment of the present invention, it should be pointed out that for those of ordinary skill in the art, without departing from the principle of the present invention, some improvements and modifications can also be made, and these improvements and modifications are also possible. It should be regarded as the protection scope of the present invention.
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310501884.8ACN116545574A (en) | 2023-05-06 | 2023-05-06 | Broadband communication anti-interference distributed compressed sensing system and method |
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310501884.8ACN116545574A (en) | 2023-05-06 | 2023-05-06 | Broadband communication anti-interference distributed compressed sensing system and method |
| Publication Number | Publication Date |
|---|---|
| CN116545574Atrue CN116545574A (en) | 2023-08-04 |
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310501884.8APendingCN116545574A (en) | 2023-05-06 | 2023-05-06 | Broadband communication anti-interference distributed compressed sensing system and method |
| Country | Link |
|---|---|
| CN (1) | CN116545574A (en) |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN119853734A (en)* | 2024-12-30 | 2025-04-18 | 中国电子科技集团公司第二十研究所 | Binary array space-time-frequency code communication anti-interference method based on intelligent perception decision |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN119853734A (en)* | 2024-12-30 | 2025-04-18 | 中国电子科技集团公司第二十研究所 | Binary array space-time-frequency code communication anti-interference method based on intelligent perception decision |
| Publication | Publication Date | Title |
|---|---|---|
| Xu et al. | Wireless image transmission using deep source channel coding with attention modules | |
| Shao et al. | Communication-computation trade-off in resource-constrained edge inference | |
| Choi et al. | Universal deep neural network compression | |
| Zhang et al. | iflow: Numerically invertible flows for efficient lossless compression via a uniform coder | |
| Wiedemann et al. | Entropy-constrained training of deep neural networks | |
| CN102665221A (en) | Cognitive radio frequency spectrum perception method based on compressed sensing and BP (back-propagation) neural network | |
| US11881227B2 (en) | Audio signal compression method and apparatus using deep neural network-based multilayer structure and training method thereof | |
| CN113129911A (en) | Audio signal coding compression and transmission method and electronic equipment | |
| CN116545574A (en) | Broadband communication anti-interference distributed compressed sensing system and method | |
| Beitollahi et al. | FLAC: Federated learning with autoencoder compression and convergence guarantee | |
| EP3069449B1 (en) | Split gain shape vector coding | |
| Wu et al. | An efficient heterogeneous edge-cloud learning framework for spectrum data compression | |
| Li et al. | Content adaptive distributed joint source-channel coding for image transmission with hyperprior | |
| Ye et al. | Lightweight generative joint source-channel coding for semantic image transmission with compressed conditional GANs | |
| Jiang et al. | Lightweight vision model-based multi-user semantic communication systems | |
| Xuan et al. | Deep joint source-channel coding for transmission of correlated sources over AWGN channels | |
| CN112669857B (en) | Voice processing method, device and equipment | |
| Aghamalek et al. | An Improved soft decision method in Viterbi decoder using artificial neural networks | |
| Liao et al. | Structured neural network with low complexity for MIMO detection | |
| Flynn et al. | Reducing bandwidth for robust distributed speech recognition in conditions of packet loss | |
| Gayen et al. | Improving Network Resource Utilization for Distributed Wireless Sensing Applications | |
| US20250285627A1 (en) | Receiving device for up-scaling decoded frame, based on neural network corresponding to codec information, and method of operating receiving device | |
| Arvinte et al. | Wideband and Entropy-Aware Deep Soft Bit Quantization | |
| Amr et al. | A new deep spiking architecture for reconstruction of compressed data in cognitive radio networks | |
| Chen et al. | Latent Feature-Guided Conditional Diffusion for Generative Image Semantic Communication |
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |