An Efficient MMSE Decoder for Differential Source Coders with Adaptive Quantizer: A Joint Source Channel Coding Perspective

A minimum mean square error (MMSE) source decoder for the reconstruction of a signal, encoded using a differential source code with an adaptive quantizer, and transmitted over a noisy channel is proposed. One group of the previous related works on MMSE-based joint source channel decoding aim to reconstruct the codec parameter rather than the original signal, which is suboptimal. The other group that aim to reconstruct the original signal have not considered the adaptive differential encoder. The proposed solution results in a two stage MMSE decoder. The decoder benefits from residual redundancy at the output of the source encoder modeled by a ¿-order Markov model. The results demonstrate the effectiveness of the proposed scheme.