EUROSPEECH 2003 - INTERSPEECH 2003 8th European Conference on Speech Communication and Technology Geneva, Switzerland September 1-4, 2003

Dominance Spectrum Based V/UV Classification and F_0 Estimation

Tomohiro Nakatani (1), Toshio Irino (2), Parham Zolfaghari (1)

(1) NTT Corporation, Japan
(2) Wakayama University, Japan

This paper presents a new method for robust voiced/unvoiced segment (V/UV) classification and accurate fundamental frequency (F_0) estimation in a noisy environment. For this purpose, we introduce the degree of dominance and dominance spectrum that are defined by instantaneous frequency. The degree of dominance allows us to evaluate the magnitude of individual harmonic components of speech signals relative to the background noise. The V/UV segments are robustly classified based on the capability of the dominance spectrum to extract the regularity in the harmonic structure. F_0 is accurately determined based on fixed points corresponding to dominant harmonic components easily selected from the dominance spectrum. Experimental results show that the present method is better than the existing methods in terms of gross and fine F_0 errors, and V/UV correct rates in the presence of background white and babble noise.

Full Paper

Bibliographic reference.  Nakatani, Tomohiro / Irino, Toshio / Zolfaghari, Parham (2003): "Dominance spectrum based v/UV classification and f_0 estimation", In EUROSPEECH-2003, 2313-2316.