Joint source/channel soft Viterbi decoding

The design criterion of a joint source/channel soft decision decoder is developed with both binary convolutional codes and nonbinary dual-k codes. The performance of the modified decoder is evaluated and compared to the performance of the decoder proposed by Sayood et al. (1994) by Monte Carlo simulations. As expected, the soft decision system outperforms the hard decision system simulated by Sayood (1994). In the context of decoding 2 and 3-bit DPCM compressed images, we show the advantage of using the nonbinary dual-k code, which matches the alphabets of the source encoder output, over binary convolutional codes. Furthermore, simulation results verify that a coding gain is achieved by exploiting higher residual redundancy levels at the source encoder output.