Soft-Decision Decoding for Reed-Solomon Codes

A soft-decision decoding for Reed-Solomon code will be proposed in this project. Although RS codes have been adopted in many current standards of communication systems and storage media, only hard-decision decoding is practical. If the soft-decision decoding can be used, the coding gain can be increased about 2-3dB in AWGN channel. The techniques investigated in this project will base on Belief Propagation (BP) decoding. Since RS code is a class of non-binary linear block codes, in order to reduce the computation complexity, the parity-check matrix will be transferred to the corresponding binary image, and the binary BP decoding can be used. However, due to the high-dense binary image, satisfactory decoding results cannot be achieved when the BP decoding is directly applied to RS code. The reason is that the bipartite graph corresponding to the high-dense binary image contains a large number of short cycles, which introduce a fast increasing in the correlation between the decoding messages, and obstruct the decoding convergence. In order to solve this problem, there will be two solutions introduced in this project. The first one is using different binary images. The binary images that have a more sparse mapping will be adopted, the shot cycles in the corresponding bipartite graph can be consequently reduced. The second solution is based on the scheduling of the BP decoding, where the propagation path of decoding messages will be constrained in order to enlarge the equivalent cycle length. With the assistance of dynamic scheduling techniques, the constraints specified for RS code decoding can be qualified, and finally fixed decoding schedules can be arranged. After the combinations of these two solutions are evaluated, the design of a practical soft-decision decoding for RS code can be completed in this project.

Project ID:PB10508-0206
External Project ID:MOST105-2218-E182-002