Wavelet ridge track interpretation in terms of formants

This paper proposes two new approaches for formant tracking using Fourier and wavelet ridges. The speech signal is decomposed into Time-Frequency representations issued from windowed Fourier transform and wavelet transform. Formant tracking is achieved by exploring ridges from time-frequency representation and imposing continuity constraints on formant trajectories. These approaches are validated by their application on synthesized vowels. The formant tracking shows a reliable estimation in comparison with the values given in synthesis. In the same way, the two analyses are applied on natural voiced speech. All the results are compared to two traditional methods using some type of LPC analysis. Fourier ridge detection analysis is shown to be a useful tool for formant tracking compared to traditional methods.