H∞ optimal signal predictive quantization subject to channel loss

This paper investigates signal predictive quantization subject to channel loss. It uses the analysis and design tools of H_∞ optimal filtering of discrete-time Markov jump linear system to obtain an optimal predictor for a signal predictive quantization system with channel loss. A linear matrix inequality (LMI) based design method is derived to design the predictor that minimizes the variance of the reconstruction error and removes the unrealistic assumption on quantizer model in the present literature. Simulation results are presented to show the advantages of the obtained design method.