Research On Polar Codes Decoding Algorithm

As a new type of channel code,polar codes have been proved to be the only one which can achieve the capacity of the channel. The basic principle of polar codes is channel polarization phenomenon. Use one binary input discrete memoryless channel for several times to make them independent with each other. After performing channel combing and channel splitting, these independent channels will turn into a set of successive uses of synthesized binary input channels called polarized channels. If the number of channel is large enough, channel polarization phenomenon will be obvious. It seems to be that the channel capacity of some polarized channels tends to 1, while others tend to zero, which enable achieve the channel capacity. Compared with widely used channel codes like LDPC and Turbo codes, polar codes have several advantages such as: obvious structural characteristics,unnecessary randomization process, and the good asymptotical performance. So polar codes have great value to be researched, and it will have great application in the future.As the code length is large enough, polar codes will achieve channel capacity, however, the performance of limited code length is not ideal. Therefore, the research of polar codes is mainly inclined to improve decoding performance. SCL decoding algorithm and SCL-CRC decoding algorithm largely improve the performance of the polar codes, however, both the two methods have very high decoding computational complexity and occupy large storage spaces.The paper proposes two decoding methods to solve the problems above, improved SCL-CRC decoding method and multi-CRC aided SCL decoding algorithm, both of them make use of the distribution of the reliability of polarized channels. The improved SCL- CRC decoding method perform SCL-CRC on channels mostly have low reliability, while performing SC or SCL(with small L) on channels mostly have high reliability. The multi-CRC aided SCL decoding algorithm put information into several parts, each part can be protected by CRC, and the retained paths through CRC could be more than one. The two methods can save the storage spaces, reduce decoding computational complexity, as well as is more convenient in the HARQ applications. As long as the decoding of the first part make mistakes, the system can draw a conclusion that the whole frame decoding is error, thus unnecessary calculation can be avoided.There is another way to improve the performance of polar codes by changing the structure of polar code. The paper put forward concatenating polar codes with other code which has strong error-correcting capacity to large the minimum distance of polar codes. The paper mainly discusses concatenating polar codes with RS code as well as the improved methods to boost the decoding performance.