| 摘要 |
The method consists in dividing the speech signal into packets of a specified number of frames of constant duration, by allocating (4) to each frame a weighting as a function of the mean power of the speech signal in the frame, in determining for each frame the corresponding coefficients of the predictive filter, by taking those already determined (5) in the neighbouring frames, if its weighting is similar to at least one of the neighbouring frames, or by calculating them in isolation (6) or by interpolation (7) from the coefficients of the neighbouring filters, in the other cases. Application: vocoders. <IMAGE> |
