| 摘要 |
PURPOSE:To shorten the computation time required for pulse retrieval processing by analyzing speech signals in frame unit and predicting pitches in accordance with the intercorrelation obtd. from the results of the analyses, then coding the speech signals. CONSTITUTION:The pitch information of every frame of the speech signals is determined by a 1st means X. The characteristic parameters of the speech signals are analyzed by a 2nd means Y to determine the self-function of the pulse response, then the residual difference signal of the speech signals is determined by a 3rd means Z. The intercorrelation of every frame and the max. value thereof are determined in accordance with the self-function determined by the 2nd means Y. The pulse positions and amplitudes of the prescribed numbers are retrieved in accordance with the results of the means X-Z by a pulse retriever 12 and the pitch prediction is executed at the intercorrelation levels, by which the speech signals are coded. The computation time required for the pulse retrieval processing is thereby shortened. |
