Abstract
In the vector quantization(VQ) methods, computation and memory amounts increase exponentially according to vector dimension and coding rate. As a fast VQ method utilizing geometrical structure of the reflection group, called RGVQ, has been proposed. It is known that, for uniformly distributed vectors on unit sphere, coding performance of the RGVQ is almost the same as one of the nearest neighbor search VQ, within practical computational complexity. In this paper, the RGVQ is applied to the waveform quantizer of the variance normalizing VQ for the Gaussian vector source.
