SC-LDPC Code With Nonuniform Degree Distribution Optimized by Using Genetic Algorithm

We propose a class of spatially-coupled low-density parity-check (SC-LDPC) codes consisting of a finite-length chain of component LDPC matrices which are assigned with different degree distributions, i.e., the strength of spatial coupling is nonuniform within an SC-LDPC code. For this class of codes, we use a modified density evolution algorithm to calculate the belief-propagation (BP) threshold on the binary erasure channel (BEC). We then optimize the degree distributions by a genetic algorithm where the BP-threshold is considered as the fitness value. The resultant codes were confirmed to exhibit improved decoding performance compared with SC-LDPC codes having uniform degree distributions.