Research On Encoding And Decoding Of Rate-compatible Punctured Polar Codes

In order to enhance the reliability and anti-interference of communication system,channel coding came into being.After the formation of channel coding theory,after decades of development and innovation of channel coding technology,various coding methods emerge one after another.In recent ten years,polar code has become a research hotspot in the field of communication.Researchers in the industry have carried out a lot of depth and breadth oriented research on the coding algorithm,decoding algorithm and rate-compatible coding scheme of polar code.Considering that the code length of polar code is limited,in the actual communication system,due to the limitations of channel conditions and other factors,the system requires that the code length and code rate of polar code have certain variability and flexibility,so that the encoded data can match the transmission channel conditions.Taking this as the starting point,this thesis studies the encoding and decoding scheme of rate-compatible polar code,the polar code rate is adapted according to the actual communication system,and the bit error performance of the ratecompatible polar code in communication transmission is improved.This paper takes the rate-compatible polar code as the research direction,and focuses on the rate-compatible polar code encoding and decoding scheme.The decoding performance of rate-compatible polar codes;the rate-compatible polar code coding scheme based on the general partial order method to reduce the complexity of the coding scheme;the rate-compatible polar code decoding scheme based on deep learning to improve the decoding performance of the decoder.The main work of this paper is as follows:(1)A rate-compatible polar code coding scheme based on auxiliary matrix optimization is proposed.Firstly,the connection is established by using the correlation between sub-channels,the sequence of polar code sub-channels is obtained,and bit-reverse permutation algorithm is carried out to achieve the even punctured bits.Secondly,using the inverted reliability sequence of polarization sub-channels,auxiliary matrices of different sizes are generated,and these auxiliary matrices are screened and simulated to obtain the general optimal auxiliary matrix.The optimal auxiliary matrix is applied to the shortening mode algorithm and puncturing mode algorithm to improve its performance advantages.The simulation results show that the proposed coding scheme can flexibly select the punctured position,improve the accuracy of punctured position,and improve the performance of rate-compatible polar code.Compared with several common rate-compatible polar codes,the proposed polar code with auxiliary matrix optimization can obtain a performance gain of about 0.05db-0.5db when BER is 10-3.(2)A rate-compatible polar code encoding scheme based on universal partial sequence method is presented.First,the reliability of the subchannel in the polar code is determined by using UPO rules.Secondly,the sub-channels determined by UPO rules can not be used for simulation analysis under optical channel to obtain the determined sub-channel reliability ordering.Then,by introducing the polarization weight formula,the general polar code sub-channel reliability calculation formula under the optical channel is established.Based on this,the rate-compatible encoding scheme under the optical channel is designed.The simulation results show that the proposed polar code optimization scheme can significantly reduce the system complexity and achieve similar performance as the polar code without the optimization scheme.(3)A rate-compatible polar code decoding scheme based on deep learning is presented.The optimized scheme uses CNN model as the basic structure of rate-compatible polar code decoder.The LLR values of the received sequence after decoding of the punctured polar codes are input into the decoder for training.Unlike conventional polarization LLR information,there is a large amount of zero data in the punctured polar code LLR information.By adjusting the loss function,activation function,number of layers of convolution network and number of convolution cores in the neural network,the decoding performance of the neural network is compared under each network parameter.The simulation results show that the three-layer convolution network has 128-64-32 convolution cores and RMSE loss function,respectively.When the ReLU activation function is modified,the rate-compatible polar code decoder has the best performance.When BER is 10-4,the CNN punctured decoder proposed in this thesis is superior to the traditional punctured decoder with a performance gain of about 0.2dB.