Channel optimization of predictive VQ for spectrum coding
J. Linden, Jan Skoglund · 2002
In this paper, design algorithms for channel optimized predictive vector quantizers are presented. Traditional methods for memoryless quantizers can be applied directly to the prediction error codebook. However, the predictor must be redesigned to obtain good performance. We present a new and more general method that simultaneously optimizes the predictor and the codebook. We also show that this design method can be used to obtain index assignments for standard vector quantizers that are advantageous to what is obtained by post process index assignment algorithms.