Abstract: A joint source-channel decoding scheme implemented by simultaneously decoding and updating parameters based on HMM is proposed.First, the source parameter used in the joint source-channel decoding is updated by the a priori knowledge obtained through the training sequence and the received noisy information sequence.The joint source-channel decoding is then employed.These two steps are implemented iteratively.The iterative updating algorithm improves precision of the probability structure of the received sequence and performance of the joint source channel decoding.The degree of parameter improvement of the source model can be obtained from the perspective of the information theory.

Key words: hidden Markov model, Baum-Welch algorithm, discrimination information, joint source-channel coding/decoding, iterative