Simple pitch-dependent algorithm for high-quality speech-rate changing

A time-domain algorithm for speeding up or slowing down speech is described. Present cut-and-splice methods preserve pitch contours and spectral shape, but produce buzzy speech, even if care is taken to blend or match edges of segments. The new method is a variant of cut-and-slice; it consists of a crude pitch detector followed by an algorithm that discards (or repeats) pieces of speech equal in length to a pitch period, then blends together the ends of the pieces that are retained. The method is easily implementable in real time, and is effective for noisy speech. The method is not pitch-synchronous. Rate changes from 0.5 to 2.0 sound natural; slowdown to 0.1 is intelligible. The method has been used (experimentally) in conjunction with silence removal to compress speech 4/1 and then reexpand to a signal that is perceptually almost identical to the original.