Finite—State Vector Quantization

Abstract

Like a predictive vector quantizer, a finite-state vector quantizer is a vector quantizer with memory. In general, a coding system with memory means that the index or channel symbol generated by the encoder depends not only on the current input vector but also on prior input vectors; correspondingly, the current reproduction produced by the decoder depends not only on the current channel symbol but also on the past history of the channel symbols received by the decoder. Both finite-state and predictive vector quantization can be considered as special cases of a more general class of VQ systems with memory called recursive vector quantization or feedback vector quantization [201], [202]. Recursive vector quantization is the generalization to vectors of the scalar recursive quantizers considered in [111], [236], [133]. It is instructive to begin with this more general class and then specialize it to the finite-state case.