| 摘要 |
<p>A method and apparatus (206) for time-scale modification of an input signal comprises the steps and a device that performs the steps of digitizing the input signal for analysis, storing the input signal in a memory and then dividing the input signal into time frames (250). The method and device then determines a difference in the short-term frequency spectrum between adjacent time frames (252 and 254) to provide a first speed factor (S1) and the input signal is time-scaled (208) in accordance with the first speed factor. Optionally, a second speed factor (S2) can be determined from the percentage of silence (262) observed within the input signal and the input signal can then be time-scaled in accordance with the first and second speed factors to provide a given quality of reconstructed speech.</p> |
