Soft decoding method for systematic raptor codes

Abstract

Raptor codes are a class of rateless codes that were originally designed for binary erasure channels. To enhance their performance, multiple soft iterative loops can be employed in the decoder. In this study, the authors propose a method to construct modified Tanner graphs for systematic raptor codes in order to increase the efficiency of soft decoding. A modified Tanner graph is drawn in terms of bit and check nodes for each component code, instead of encoding and information nodes, as in conventional Tanner graphs, for its component Luby transform codes. This modified Tanner graph enables the use of a more simplified soft iterative decoding algorithm and provides improved decoding performance compared with the conventional method. In addition, they propose an efficient joint iterative soft decoding method with a parallel process to speed up the iterative decoding. The simulation results in this study reveal that the modified Tanner graph works better than the conventional Tanner graph for joint iterative decoding, and that the proposed parallel decoding method can contribute to further enhance the performance as well as the decoding speed by reducing the average number of iterations.