Iterative source-channel decoding design using distortion based index assignment and joint redundant information

Abstract

This work presents a convolutional coded iterative source-channel decoding (ISCD) to improve the fidelity of joint source-channel coded systems through designing index assignment (IA). A proposed end-to-end distortion based IA generates the joint redundant bits to provide more useful information for source and channel decoders. Specifically, such joint redundant bit not only enhances the error-correcting capacity of source codes but also provides the self-concatenated convolutional codes instead of convolutional codes for ISCD. In contrast to the conventional ISCD, our proposed ISCD employs the self-iterative channel decoding algorithm, and the extrinsic information exchange can be further enhanced through joint redundant bits. With considering joint redundant bits, we propose a modified source decoder with the goal to extract the extrinsic information for the self-iterative channel decoder. Simulation results show the proposed ISCD with our designed IA can further improve the Parameter Signal-to-Noise ratio performance.