| 摘要 |
A synthesis filter is disclosed which models the effect of the fundamental frequency of speech for digital speech coders operating on the analysis-by-synthesis principle. High fundamental frequencies having a period shorter than the corresponding cycle length of the frame employed in the analysis-by-synthesis method are optimally encoded. The filter is constructed of a number of parallel, separately updatable synthesis-memory blocks. When analysis delays shorter than the analysis frame are used, a portion of a signal that was stored in memory several frames earlier is selected and scaled to approximate the missing portion of the analysis frame using the available portion of the analysis frame. |
