摘要 |
The invention claims a method and an appropriate acoustic signal processing device for estimating a set of linear predictive coding coefficients (¸ s,k ) of a microphone signal (x(k)) using minimum mean-square error estimation with a codebook comprising several predetermined sets ¸ s j of linear predictive coding coefficients. The method comprises the steps:
- determining (102) sums p �¢ ¸ ^ s , k - 1 | ¸ s i of weighted w s , k - 1 j backward transition probabilities ( b ij ) describing the transition probabilities between said predetermined sets ¸ s j of linear predictive coding coefficients, whereas said backward transition probabilities ( b ij ) are obtained from signal training data by mapping said signal training data to one set ¸ s j of said codebook and by determining relative frequencies of transitions between two said sets ¸ s j of said codebook. Modelling the "memory" of the codebook according to the invention has the advantage that the accuracy of estimating linear predictive coding coefficients is increased considerably also for speech components.
|