Analysis by synthesis adaptive predictive correlated speech coding

A correlated encoding strategy called the finite-state vector quantizer for correlated encoding (FSVQ-CEC) for the analysis-by-synthesis adaptive predictive speech coding (ASAPC) speech system is developed, which exploits the redundancy between vocal tract information and excitation information. Coding efficiencies have been obtained by taking advantage of the correlation among the speech samples or frames. The correlated characteristics between vocal tract parameters and excitational ones of these coders are first described and the correlated encoding strategies are presented. To demonstrate the simulation results and effectiveness of FSVQ-CEC, the authors apply FSVQ-CEC to linear-Gaussian observation and compare its rate-distortion performance with a theoretical bound