The Study of Hybrid Decoders of Low-Density Parity-Check Codes

The algorithms for decoding low-density parity-check (LDPC) codes can be classified into the belief propagation (BP) and the bit-flipping (BF) algorithms. The class of BP algorithms involves the sum-product (SP) and the min-sum (MS) decoding algorithm; and the class of BF algorithms involves IMWBF, MWBF, WBF and BF decoding algorithms. In general, the class of BP algorithms have excellent BER performance, however they have high decoding complexity due to the operation of iterative decoding. On the other hand, the BF algorithms have low complexity but are worse than BP algorithms in BER performance. To make a trade-off between BER performance and system complexity, hybrid schemes (or called bootstrap decoding scheme) [20] have been proposed. In the first decoding iteration , these hybrid schemes process a LDPC codeword using a BP algorithm to enhance the reliability of symbols in the codeword, then decode the codeword with BF iterative decoding. However, the corresponding hardware circuit of these hybrid schemes should be complex because two different decoding algorithms need to be implemented in one decoder. In this project, a new decoding algorithm will be developed based on AWGN channel. This decoding algorithm can be programmed, by setting parameters, to execute SP, MS, IMWBF, MWBF, WBF, BF and GWBF decoding algorithms; where GWBF is deduced from the new algorithm. The GWBF can be shown that its BER performance is better than that of IMWBF. Since the new decoding algorithm can be programmed to execute BP or BF decoding in one hybrid decoder, its hardware complexity should be very low.

Project ID:PB10108-2812
External Project ID:NSC101-2221-E182-049