An Automatic Pitch-Marking Method using Wavelet Transform

This paper describes a new automatic pitch-marking method using wavelet transform. This method detects discontinuity in the speech waveform which occurs at the glottal closure instant (GCI). A time domain prosodic modification technique requires an appropriate determination of the synthesis pitch-marks. We evaluated the performance of the newly developed pitch-marking method by using our internal speech databases with an electroglottograph signal. We achieved 96 percent detection accuracy on the performance evaluation. We confirmed that the proposed pitch-marking method is suitable for waveform concatenation-based synthesis through a listening test using pitch modified speech.

By: Masaharu Sakamoto, Takashi Saitoh

Published in: ICSLP2000 6th International conference on Spoken Language Processing, unknown, vol.3, p.650-653 in 2000

Please obtain a copy of this paper from your local library. IBM cannot distribute this paper externally.

Questions about this service can be mailed to reports@us.ibm.com .