A parallel APP decoding algorithm for accelerating decoding throughput of turbo codes

Abstract

We propose a parallel a posteriori probability (APP) decoding algorithm for increasing the decoding throughput of turbo codes. The parallel decoding algorithm divides a noisy codeword into multiple sub-blocks which are decoded in parallel. The authors in [1] proposed utilizing the forward and backward variables computed by neighbor sub-blocks in the previous iteration as the initial boundary conditions for each sub-block in the current iteration. We extend this algorithm by organizing sub-blocks into odd and even numbered groups, which perform recursions in unison but in the opposite directions. This enables more efficient utilization of the boundary conditions. We show by means of numerical simulations that our proposed algorithm further improves the decoding throughput of turbo codes in highly parallel decoding scenarios.