摘要 |
PURPOSE: To improve speech quality by varying prediction coefficients, calculated by frames, in the frames sequentially by arbitrary step width so that residual signal power is minimized. CONSTITUTION: Frame length is set long (e.g. 30-40msec) and the bit distribution quantity of prediction coefficients is increased; and the resulting decrease in time resolution is prevented by varying the calculated prediction coefficients in the frames sequentially in the optimization direction. This means initially set a predictor 15 to the prediction coefficient calculated at the head of a frame, and then the prediction coefficients operate so as to minimize momentary residual electric power, so that a prediction gain increases. Consequently, the speech quality can be improved. |