摘要 |
PURPOSE:To easily reflect the dynamic characteristics of voices in output symbols by comparing the characteristic vectors of the input voices and the characteristic vectors based on stored predictors and specifying the predictor from the plural predictors. CONSTITUTION:The characteristic vectors of the input voices from a characteristic holding part 43 are supplied to a characteristic predicting section 42, by which the characteristic vectors complying with the plural predictors selected from a predictor memory section 41 are predicted. The errors between the predicted characteristic vectors and the inputted characteristic vectors are calculated in an error calculating section 44 of the error calculating section and a deciding section 45 specifies the one predictor corresponding to the predicted characteristic vector to minimize the errors. The voice coding device which easily reflects the dynamic characteristics of the voices in the output symbols is obtd. in this way. |