JOINT OPTIMIZATION AND MODEL PARAMETER IN PARAMETRIC SPEECH CODER

摘要

PROBLEM TO BE SOLVED: To provide a more accurate voicing model. SOLUTION: The optimization of a speech synthetic system which optimizes a synthesizing filter minimizes the synthesis error between an original speech sample and a synthesized speech sample. Gradient retrieval algorithm in a solution area also assists the minimization of the synthesis error. The gradient retrieval algorithm of a speech encoding system computes the gradient of a speech synthesis polynomial by using the contribution of a decomposition coefficient. The contribution of the decomposition is computed again by subsequent repetition.