| 摘要 |
A method and an apparatus is disclosed for generating speech feature signals characterizing speech signals by generating a set of spectral parameters characterizing speech signal, weighting the set of spectral parameters, and generating speech characterizing feature signals based on the weighted set of spectral samples. If speech signals are received they are blocked into frames and then in combination a linear predictive coding and cepstral recursion analysis is performed on the blocked signals to produce mel-LPC cepstrum feature signals. The mel-LPC cepstrum features are then compared to a recognition model. <IMAGE> |
