A finite-state vector quantization correlated encoding （FSVQ-CEC） strategy is presented in this paper. The rate-distortion performance of FSVQ-CEC for linear-correlated Gaussian observations is demonstrated

and is compared with the optimum performance theoretically attainable of correlated encoding and the performance by simple encoding which fails to capitalize on the correlation between sources. The FSVQ-CEC scheme is applied to an analysis-synthesis adaptive predictive speech encoding system so as to exploit the redundancy between vocal tract codevectors and excitational codevectors.