Improved Soft Decoding of Reed-Solomon Codes on Gilbert-Elliott Channels

摘要

It was shown by Guruswami and Rudra that Reed-Solomon codes can be list decoded to recover from phased burst errors (i.e. errors occurring within fixed regular intervals) up to the information-theoretic limit and, in particular, beyond the Guruswami-Sudan bound. In this paper, we present evidence that the algorithm developed by Guruswami and Rudra can also give improvement for more "irregular" burst errors. We develop a low-complexity multiplicity assignment scheme for soft decoding of Reed-Solomon (RS) codes. Specifically, we present simulation results where such soft decoding of RS codes outperforms the existing soft decision decoding algorithms of Koetter and Vardy as well as the algorithm of Das and Vardy on Gilbert-Elliott channels (under QAM and BPSK modulations) for channels that are more bursty. We also present a theoretical result that shows that for certain Gilbert-Elliott channels, with high probability of errors, the output list size for list decoding RS codes is one.