Research On High Performance Decoding Algorithm Of Polar Codes

The fifth generation of mobile network(5G)is a research hotspot in the field of communication in recent years.Both academia and industry are striving to promote the development of 5G.Channel coding technology is one of the core technologies in the field of wireless communication,which is an important technical means to against transmission interference.Its performance is directly related to network coverage and user transmission rate.As the first advanced channel coding scheme that can theoretically reaches the Shannon limit,Polar code was proposed by E.Arikan at the 2008 International Conference on Information Theory and became the short code standard of control channel in the scenario of 5G Enhanced Mobile Broadband(e MBB)in 2016.However,as a new coding scheme that was born only for more than ten years,Polar code has a lot of shortcomings in practical application.First of all,for the short code length of the Polar code due to the incomplete channel polarization,many sub-channels can not reach the standard of perfect channel,and the reliability is poor.Secondly,the decoding scheme based on the Successive Cancellation(SC)algorithm has the problem of error propagation,which greatly affects the transmission performance.In order to solve the problem of poor performance of Polar code in medium and short code length,this dissertation studies the error correction capability of cyclic redundancy check(CRC)code based on the CRC-Polar cascade structure adopted by 5G standardization and proposes two kinds of error correction decoding schemes,named error-correcting table based segmented CRC error correcting aided SC list(ET-SCC-SCL)decoding and deep learning based SCC-SCL(DL-SCC-SCL)decoding respectively.Aiming at the problem of poor performance of Polar code in short and medium code length,this dissertation proposes an ET-SCC-SCL decoding algorithm based on error correction table.The information block is segmented according to the error probability of the sub-channels reasonably and a CRC detector is attached to the tail respectively,so as to complete error detection and correction in a timely manner.The sub-information block is decoded and verified in the form of segmentation decoding,and the sub-information block that cannot pass the verification is corrected by using the CRC error correction method based on the error correction table.Through complexity analysis and simulation,the effectiveness of the decoding scheme is verified,which can greatly improve the transmission performance of Polar code.In order to further improve the error correction performance of the decoder,this dissertation proposes a DL-SCC-SCL decoding algorithm based on deep learning method.Long Short-Term Memory(LSTM)network is used to replace the error correction table to complete the error correction process.The error pattern is jointly identified by the log likelihood ratio(LLR)generated during decoding and the syndrome generated during CRC check.By introducing the neural network and new auxiliary information,the limitation of the size of the error correction table is overcome.Complexity analysis and simulation results show that the DL-SCC-SCL algorithm can obtain higher decoding performance gain and significantly improve the transmission performance of Polar code in short to medium code length.