A COMPARISON OF A WAVELET TRANSFORM EVENT DETECTION PITCH DETECTOR WITH CLASSICAL PITCH DETECTORS
Shubha L. Kadambe, G.F. Boudreaux-Bartels · 2005
Classical cepstral or auto-correlation based pitch detectors[l-5] developed hitherto have the following disadvantages: 1) they are not sensitive to non-stationary pitch variations since they estimate the average pitch period over a given segment of a speech signal, 2) they are not suitable for both low pitched and high pitched speakers since the analysis frame length is fixed and 3) they are not robust to noise. In this paper, we describe an event based pitch dctector which overcomes the above mentioned problems. We apply the Dyadic Wavelet Transform(D,1;TiT) to detect glottal closure, which we define here as an event, and estimate the pitch period by measuring the time interval between two such cvents. The performance of this event based pitch detector is compared with that of non-cvcnt based pitch det.ectors which use Ccpstrum and Autocorrelation[1,2] metliods to estimate t.he pitch period.