Belief Propagation Decoding Of Polar Codes Using Intelligent Post-processing

Polarization code is currently the only coding technology proven to reach the Shannon limit in theory.Due to its excellent coding performance and low complexity,it has extremely high research value.Among the commonly used decoding algorithms for polar codes,Belief Propogation(BP)decoding algorithms have the characteristics of low complexity and can be calculated in parallel,so they have attracted the attention of the industry.However,there is still a certain gap between the performance of the BP decoding algorithm and the SCL decoding algorithm.The study found that if the information bits with higher probability of translation error are reversed after BP decoding,the decoding performance will be improved,so it is very important to choose the appropriate information bit.Deep learning has excellent learning ability,and the neural network can process new data after training.Therefore,the deep neural network is used for the selection of effective information bits,which can effectively improve the accuracy of selection and improve decoding performance.This article mainly studies the BP decoding algorithm based on intelligent information post-processing,and proposes a BP decoding algorithm based on parity post-processing.The specific work is as follows:The thesis firstly proposes a BP decoding algorithm based on deep post-processing intelligent post-processing,studies the coding and decoding principles of polarization codes,analyzes the physical parameters that affect the performance of BP decoding of polarization codes,and extracts appropriate features.The deep neural network classifier classifies the output data of BP decoding into correct bits(that is,the classifier thinks that the bit is probably the correct translation bit)and wrong bits(that is,the classifier thinks the bit is probably the wrong translation bit).From the category of erroneous bits,select the most likely erroneous bits as additional frozen bit information to correct residual errors in the BP decoding iteration process and improve the decoding performance of the BP decoding algorithm.Compared to the BP decoding algorithm,which has a gain of 0.4db,the computational complexity and energy consumption have increased by 22.83%..The paper proposes a BP decoding algorithm based on parity post-processing.Through statistical analysis of the reliability of the information bits,the rule that the probability of decoding errors presents a stepwise decline is obtained,and parity check codes are used to perform unreliable bits.Error correction is used as additional frozen bit information.Compared with the original BP decoding algorithm,the algorithm has a gain of 0.4db at high signal-to-noise ratio,and the performance reaches the SCL(L=2)algorithm.