VOICE ENCODER ADOPTING ANALYSYS-SYNTHESIS TECHNIQUE BASED ON PULSE EXCITATION

摘要

PURPOSE: To use a voice encoder, where an analysis-syntnesis technique is adopted, with a satisfactory performance. ;CONSTITUTION: In an analysis-synthesis encoder, an original voice signal is shifted by a short time so that it is consistent with an expected signal, which is encoded with a replica made by a long-time synthesis filter, with respect to time. This shift is determined in each subframe by through search in the range of such possible values that energy of an error signal is minimized. Once the optimum shift is determined, an optimum excitation is searched. This excitation has a very small number of pulses arranged in a decision theoretic constitution and is selected in a code book including words all of which are obtained from a limited number of keywords. The decision theoretic constitution realizes quick search of the optimum excitation without storing the code book neither actually executing the synthesizing filter operation of candidate excitations.;COPYRIGHT: (C)1995,JPO