Abstract:

To overcome the drawback of the extended min-sum algorithm (EMS), an improved EMS decoding algorithm is proposed, which is designed for LDPC over GF(q). The algorithm adaptively chooses the rank of FHT in each iteration step according to the average variance of bit nodes’ probability pairs. It differs from the original EMS algorithm in which the rank of FHT for each iteration step is a constant. Moreover, for oscillating variable nodes, the message of the previous iteration is added to the current message in the iterative procedure to reduce loss of performance and effects of oscillation. Experimental results show that the proposed algorithm can achieve better performance and converge faster than EMS with the same decoding complexity.

Key words: low density parity check codes, sum-product algorithm, fast Hadamard transform, extended minsum algorithm(EMS), Galois field