Research On Coding And Decoding Method Of Repeat Accumulate Code

As one member of error correcting codes that almost reach the Shannon limit,Repeat Accumulate (RA) code is a common subset of both Turbo code and LDPC (LowDensity Parity Check) code, and yet it enjoys a significant advantage of codingsimplicity over Turbo code and low decoding complexity over LDPC codes. Due to itsexcellent performance and low coding-decoding complexity, RA code attracts numerousresearchers and it has already been an important candidate of error correcting codeschemes for future communication systems.RA code,including RA code and irregular RA code, as its relation to Turbo codeand LDPC code, it can be decoded either by the BCJR algorithm or the BP algorithm.The BP decoding method is first discussed in this paper. As the standard BP algorithmis too complex, some improved algorithms are introduced, which includes the min-sumalgorithm, the normalization algorithm, the migration algorithm, and the broken-linealgorithm. Through these algorithms, decoding complexity is reduced and meanwhilelittle performance loss is brought in. In the second part, the BCJR algorithm of RA codeis introduced, and a comparison of BP and BCJR method is made. It is found that shortloops in one code word could lead to negative effects on decoding performance of theBP algorithm. In light of the loop resolution methods of LDPC codes, it is necessary tomake some modification in the encoding process in order to improve the performance ofthe BP algorithm. By introducing interleavers which are fixed rather than random, andcoding RA in multiple fields instead of two, small or medium-sized loops of RA codecan be eliminated observly, and big improvement of BP decoding performance could beachieved.