Joint source-channel decoding of variable-length encoded sources

In systems that employ source coding, imperfect compression may leave some redundancy in the data at the input to the channel encoder. This paper presents a new method for exploiting that redundancy via joint source-channel maximum a posteriori (MAP) decoding. The technique may be applied to systems that employ a variable-length source code in conjunction with a channel code; the decoder uses the residual redundancy that remains after compression to enhance robustness. The proposed technique (applicable to both memoryless and Markov sources) combines the source model, the source decoder and the channel decoder to construct a joint decoder with a conventional Viterbi structure. Simulation results comparing the performance of this approach with the conventional tandem decoding approach are presented.