A novel decoding method for Linear Index Joint Source-Channel Coding schemes

It is a known fact that fixed-to-variable length coding schemes are not robust to post-channel decoding errors and even a single error after channel decoding can cause desynchronization of transmitter and receiver thus making the rest of the decoded bitstream unusable. One of the method to avoid the catastrophic effect of such schemes e.g. Huffman and Arithmetic coding, is to use the linear data compression instead. Generally, it is difficult to compress the source up to its entropy using linear compression schemes thus considered infeasible in the context of conventional Separated Source-Channel Coding communication system. However, a Linear Index Coding scheme was proposed in the literature using Joint Source-Channel Coding which can compress the source data up to its entropy using channel codes. The scheme directly encodes the source quantization indices to the channel codewords. The compression is achieved by assigning the data budget to each quantized bitplane depending on its entropy conditioned on previous bitplanes while only transmitting the coded bits. Decoding is performed in a sequential manner where each decoded bitplane provides the a priori probabilities to the subsequent bitplane for the systematic bit positions. The scheme is robust to post-channel decoding errors but results in lower bandwidth efficiency. In this paper, we proposes a novel decoding method for the above mentioned Joint Source-Channel Coding scheme. The proposed scheme decodes each bitplane in a cooperative and iterative manner. At each decoding iteration, only the symbols recovered with high reliability are used to provide likelihoods to the subsequent bitplanes. The proposed scheme not only improves the bandwidth efficiency of the transmitted data but also shows increased robustness under channel mismatch conditions.