Weighted nonbinary repeat-accumulate codes

Repeat-accumulate (RA) codes are random-like codes having remarkably good performance over an additive white Gaussian noise (AWGN) channel, like turbo and low-density parity-check (LDPC) codes. In this correspondence, we introduce an ensemble of random codes called "weighted nonbinary repeat-accumulate (WNRA) codes" whose encoder consists of a nonbinary repeater, a weighter, a pseudorandom symbol interleaver, and an accumulator over a finite field GF(q). They can be decoded in a simple way by applying the sum-product algorithm to their factor graphs over GF(q). Simulation results show that WNRA codes with proper weighting values over GF(4) or GF(8) are superior to binary RA codes on AWGN channels.