Sixth European Conference on Speech Communication and Technology

We present a novel efficient algorithm for timescale modification (TSM) of speech which gives output quality equal to that of a conventional TSM algorithm, but having computational load an order of magnitude less. The algorithm presented uses a fixed length rectangular stepping window and a simple peak alignment criterion to track the local natural scaling factor and adapt the window step size. The desired TSM factor is realised by the appropriate number of applications of the constantly varying local natural scaling factor. The local natural scaling factor estimate is updated at subpitch period intervals giving accurate pitch tracking and high quality in the output scaled signal.
Full Paper (PDF)
Acoustic Example #1
Acoustic Example #2
Acoustic Example #3
Acoustic Example #4
Acoustic Example #5
Bibliographic reference. Lawlor, B. / Fagan, A. D. (1999): "A novel high quality efficient algorithm for timescale modification of speech", In EUROSPEECH'99, 27852788.