Research On Joint Source And Channel Polarization Coding Theory And Key Technologies

Polar codes is a constructive coding scheme for asymptotic capacity achieving based on the idea of channel polarization,which has been chosen as the coding scheme for the control channel of 5G enhanced mobile bandwidth scenarios due to its excellent performance at short codes.The polarization idea can also be applied to source coding,called source polarization,and the source coding rate based on the source polarization theory can reach the source entropy.Combining source polarization theory and channel polarization theory,joint source and channel polarization coding has become an important research direction.However,the theory of joint source and channel polarization coding and the analysis method of the performance of the coding/decoding algorithm need further research,and the research of related joint source channel polarization decoding algorithm is still in the initial stage,and the basic design theory and coding/decoding method of joint source and channel polarization coding for distributed scenarios still has a large gap.In view of this,this dissertation investigates the design theory and coding/decoding method of joint source and channel polarization and its application in distributed scenarios to improve the overall performance and pave the way for the application of polar codes in future wireless communication systems.In this dissertation,we first study the theoretical basis of the joint source and channel polarization coding and establish a complete framework of joint source and channel polarization coding.Then,based on this framework,this dissertation utilizes the factor graph representation of polar codes,proposes the joint source and channel belief propagation decoding algorithm,designs the external information transfer chart analysis method under biased sources,and further designs the joint source and channel polarization coding/decoding method under distributed scenarios.Specifically,this dissertation includes innovative work in the following four areas.First,combining source polarization theory and channel polarization theory,the joint source and channel polarization coding method is investigated,the theoretical framework of joint source and channel polarization coding is constructed,and the double polar codes method is proposed.Based on the theoretical framework of joint source and channel polarization coding,this dissertation first proves the upper bound of error probability for the double polar codes,then proves that the end-to-end mutual information of the joint source and channel polarization coding system can reach the source entropy,and also gives the upper and lower bounds of the end-to-end polarization mutual information.Based on the type of channel polar codes used in double polar codes method,double polar codes can be divided into two categories.The first category is the double polar codes(D-PC)method consisting of source polar code and non-systematic channel polar code;the second category is the system double polar codes(SDPC)method consisting of source polar code and channel systematic polar code.For the D-PC method,this paper proves the sufficient conditions for the optimal mapping between the source polar code and the channel polar code,and based on this,the optimal mapping construction algorithm is proposed.In addition,this dissertation designs an error set construction method for source polar codes in order to avoid residual errors in source polar codes with finite code length,which can effectively eliminate the error floor phenomenon in joint source and channel decoding.Second,the design of the joint source and channel decoding method based on the belief propagation(BP)algorithm is optimized so that the channel polar code in the double polar codes can make full use of the residual redundancy of the source polar code.In the TL-BP decoding algorithm,the source BP decoder and the channel BP decoder interact with each other to improve the overall system performance.The simulation results show that the SD-PC with TL-BP algorithm outperforms the existing joint source and channel polarization coding method.Then,by extending the factor graph representation of polar codes to double polar codes,the joint source and channel factor graphs of D-PC and SDPC,called joint factor graph(J-FG)and systematic joint factor graph(SJ-FG),are proposed in this dissertation,respectively.Based on the J-FG and SJ-FG,the joint belief propagation(J-BP)algorithm and the systematic joint belief propagation(SJ-BP)algorithm are proposed in this dissertation.The simulation results show that both D-PC with J-BP algorithm and SD-PC with SJ-BP algorithm can achieve significant performance gains compared with existing joint source and channel polarization coding methods.In addition,a theoretical performance analysis tool for joint decoding of double polar codes is proposed in this dissertation,called the biased external information transfer(B-EXIT)method,which can effectively analyze the convergence performance of joint decoding of double polar codes under non-uniform sources.Third,a joint successive cancellation list(J-SCL)decoding algorithm is proposed in this dissertation.In this dissertation,,based on the structural features of the D-PC method,the trellis diagram of source polar code and that of channel polar code are fused into a composite trellis diagram,in which there are three types of leaf nodes,namely,the joint source-channel(JSC)node,the low entropy node and the frozen node.The JSC node is both the information bit of the channel polar code and the high entropy bit of the source polar code.Based on the composite trellis diagram of D-PC,this dissertation proposes a method to construct the index set of JSC nodes and low entropy nodes in the composite trellis diagram.Using the index set of JSC nodes and low-entropy nodes,this dissertation proposes a J-SCL decoding algorithm for D-PC,which combines the L paths of the channel SCL decoder and the L paths of the source SCL decoder into a unified L joint decoding path,and the joint path metric is updated considering the output of both source and channel decoding.The simulation results show that the J-SCL decoding algorithm can effectively reduce the error level without transmitting the error set of the source compared with the BP decoder-based joint decoding algorithm.Fourth,for the distributed coding,this dissertation designs a distributed joint source and channel coding(DJSCC)method,called triple polar code(T-PC)method,which can effectively exploit the correlation between sources and the residual redundancy in source codes in the DJSC scenario.The TPC method consists of a systematic polar code and D-PC.Based on the T-PC method,this dissertation further designs a distributed joint source and channel decoding scheme,which consists of a cyclic redundancy check assisted successive cancellation list(CA-SCL)decoding algorithm and a J-SCL decoding algorithm.In this distributed joint decoding algorithm,the hard judgment result of the CA-SCL decoder is used as side information in the J-SCL decoding process based on the correlation between the sources.At the same time,the hard decision result of the J-SCL decoder is used as side information in the decoding process of CA-SCL.The simulation results show that the T-PC method achieves better performance than the existing DJSCC method based on systematic polar codes in distributed coding.In this dissertation,the theory and key techniques of joint source and channel polarization coding are studied,and the joint source and channel polarization coding method is extended in distributed scenarios.In the coding method,the D-PC method and SD-PC method are proposed,and the construction of the optimal mapping in the D-PC method is also given.In terms of decoding,two types of joint source channel decoding algorithms based on the BP and SCL algorithms are proposed in this dissertation.In the distributed coding,this dissertation investigates the distributed joint source and channel polar coding method.This dissertation improves the basic theory of joint source-channel polarization coding/decoding,improves the overall performance of the JSCC system,and provides guidance for the application of polar codes in future infinite communication systems.