Blind Identification Of Channel Codes Based On Deep Learning And Multi-Dimensional Feature Extraction

Blind identification of channel codes is a critical technology in non-cooperative signal processing and modern intelligent communication,and also plays an vital role in wireless physical layer security.The identification and analysis of channel codes refers to the estimation of coding types and parameters by using the intercepted codeword sequence in the non-cooperative communication environment with little or no prior information,so as to decode the codeword sequence and recover the data transmitted by the codeword sequence.At present,blind identification of channel coding technology is extensively applied in information countermeasure and information interception research.In scenarios such as military information confrontation and electronic communication surveillance,the third communication party does not have prior information such as encoding type and encoding parameter used by the communication parties in the communication process,and the non-cooperative receiver needs to intercept the signal transmitted by the transmitter and extract relevant information from it.This paper investigates the use of deep learning for blind channel coding recognition in non-cooperative communication scenarios.First of all,the existing blind channel coding recognition algorithms have high computational complexity,poor performance in a low signal-to-noise ratio environment,and there will be signal distortion caused by channel fading in the actual communication environment,which will affect the recognition performance.Secondly,current researches on blind channel coding recognition based on deep learning methods focus on parameter recognition of channel codes,and regard coding types as prior information without considering the problem of coding type recognition.However,in a completely blind environment,it is necessary to consider the coding types used in codeword sequences.Therefore,the main work of this paper is as follows:(1)For solving the problem of distinguishing channel coding parameters in low signal-to-noise ratio and fading environment,this paper applies deep learning technology to blind channel coding recognition,proposes a channel coding recognizer based on deep residual shrinkage network,and realizes the channel coding parameter estimation.It is verified in additive white Gaussian noise,single-path Rayleigh fading,and multi-path Rayleigh fading channels.The experimental results show that the recognizer performs well in low signal-to-noise ratio and channel fading environments.The accuracy can exceed90% when the signal-to-noise ratio is not less than 4d B in the additive white Gaussian noise channel,and the precision can be improved by 5% to 15% compared with the existing research results.(2)To solve the problem of channel coding type recognition,this paper presents a data randomness-based multi-dimensional feature extraction algorithm,and realizes the identification of coding types and coding parameters by facilitating the efficient channel attention mechanism convolutional neural network.The problem of blind channel coding recognition is divided into three sub-problems: coding type recognition,coding length recognition,and coding rate recognition,then the three sub-problems are simulated.The experimental results illustrate that the system has a decent property in the three sub-problems: coding type recognition,coding length recognition,and coding rate recognition.If the signal-to-noise ratio is not lower than 11 d B,the precision can reach more than 80% in coding type recognition.Compared to the prior research,the accuracy can be improved by 10% to 30% in coding length and coding rate recognition.