Channel Coding Recognition And Scrambling Analysis Based On Deep Learning

In order to ensure the reliable and effective transmission of signals during transmission,it is necessary to use channel coding technology to process signals in the actual communication environment.Currently,most of the research on channel coding is to design specific algebraic algorithms to perform tasks such as code parameter identification when the channel coding type(such as error correction code type,interleaving type,etc.)is known.However,channel coding usually consists of one or more of error correction coding,interleaving,and scrambling codes.In non-cooperative communication systems,the channel coding type at the receiver is usually unknown.Therefore,before performing parameter identification,it is necessary to first identify the channel coding structure and type.Based on deep learning technology,this thesis proposes end-to-end solutions to problems such as channel coding structure identification,error correction code type and interleaving type identification,and descrambling.Firstly,this thesis proposes basic convolutional neural network,network based on Inception structure,and network based on residual structure to identify the channel coding structure of signals.By comparing the test results of the three networks,it is found that the Inception network model is the best network,with a recognition accuracy of 99%.Secondly,considering the correlation between different signal samples,a multi task learning network based on the Inception structure is proposed.For signals using error correction coding and interleaving coding,both error correction coding type recognition and interleaving type recognition tasks can achieve an accuracy rate of over 90%.Finally,this thesis proposes an autoencoder to descramble the signal,using the scrambled signal samples identified by the channel coding structure recognition network as a dataset.When the signal-to-noise ratio reaches 10 d B,the bit error rate is less than 0.001.In summary,the research work in this article fully demonstrates the feasibility and effectiveness of deep learning in channel coding structure identification and signal descrambling tasks.Compared with traditional methods,the methods proposed in this article have a wide range of applicability and are more intelligent in operation.