Fast vector quantisation using cache codebook

A vector quantiser is described in which the best representative of the input vector is searched for among the most recently used codewords, and is stored in a cache codebook. More than 60% reduction of the coding time can be achieved with less than 1% decrease of SNR.