EUROSPEECH 2001 Scandinavia
7th European Conference on Speech Communication and Technology

Aalborg, Denmark
September 3-7, 2001


Formant Estimation using Gammachirp Filterbank

Kas Ouni, Zied Lachiri, Noureddine Ellouze

LSTS, ENIT, Tunisia

This paper proposes a new method for formants estimation using a decomposition of speech signal in Gammachirp functions base. It is a spectral analysis method performed by a gammachirp filterbank. A similar approach to the modified spectrum estimation, which allows a smooth and an average spectrum is adopted. In fact, instead of using an uniform window commonly used in short-time Fourier analysis, a bank of gammachirp filters is applied on the signal. A temporal average of the estimated spectra is then applied to obtain one spectrum highlighting the formants structure. This approach is validated by its application on synthesized vowels. The formants are detected with good estimation in comparison with the values given in synthesis. In the same way, this analysis is applied on natural vowels. All the results are compared to three traditional methods, LPC, cepstral and spectral ones and also to a same analysis given by a gammatone filterbank. The tracking of formants shows that this method, which based on gammachirp filters, gives a correct estimation of the formants compared to traditional methods.

Full Paper

Bibliographic reference.  Ouni, Kas / Lachiri, Zied / Ellouze, Noureddine (2001): "Formant estimation using gammachirp filterbank", In EUROSPEECH-2001, 2471-2474.