Reduced-Search Source Decoders for Iterative Source-Channel Decoding

Abstract:
Iterative source-channel decoding (ISCD) exploits the residual redundancy of source codec parameters by using the Turbo principle. However, ISCD might require more computational complexity than available as the utilized soft decision source decoder (SDSD) can be computationally quite expensive. In this paper we propose a reduced-search SDSD, based on the M-algorithm known from channel decoding, which considerably reduces the complexity of the receiver. Furthermore, we show that by slightly modifying the quantization at the transmitter, the complexity can be further reduced without noticeable performance losses. Complexity figures are given for all approaches as well as a simulation example showing the performance of the complexity-reduced SDSD in an ISCD framework.