Harmonic-Temporal Clustering of Speech for Single and Multiple F0 Contour Estimation in Noisy Environments

We present in this paper a novel F₀ contour estimation method based on a parametric description of the wavelet power spectrum of speech that accounts for its structure simultaneously in time and frequency directions. We model the speech spectrum as a sequence of spectral clusters governed by a smooth common F₀ contour expressed as a spline curve. The harmonic and temporal structure of these clusters and their common F₀ contour are estimated simultaneously. Through experimental comparisons with existing methods, we show that our algorithm is competitive on clean single-speaker speech, and that it outperforms existing methods both in the presence of noise and for the estimation of multiple F₀ contours of cochannel concurrent speech.