Channel optimized predictive vector quantization
J. Linden · IEEE Transactions on Speech and Audio Processing · 2000
The paper investigates how channel optimization techniques can be applied to predictive vector quantizers. In particular, an efficient encoder search procedure and two design methods are derived. The design methods proposed here, one sample iterative and one block iterative, simultaneously optimize the predictor and the codebook. Extensive simulations show the advantage of this quantization method compared to other memory based quantization schemes as well as memoryless VQ. We also demonstrate that the design methods can be used to obtain index assignments that are advantageous to that obtained by post process index assignment algorithms.