MULTI-PULSE EXCITED LINEAR PREDICTIVE SPEECH CODER

摘要

An LPC-synthesizer (1) produces a synthetic speech signal (s(n)) of which the difference (2) from the reference speech signal (s(n)) is perceptually weighted (4). In response to the weighted error signal (e(n), epsilon (n)) error minimizing means (5) control the multi-pulse excitation signal generator (6). The error minimizing procedure is accelerated by affecting the minimizing operation only in the region of the maximum of an auxiliary function (Mk(n) which is a measure of the energy of the weighted error signal.