TIME-FREQUENCY CONSTRAINTS FOR PHASE ESTIMATION IN SINGLE-CHANNEL SPEECH ENHANCEMENT

Previous single-channel speech enhancement algorithms often em-ploy noisy phase while reconstructing the enhanced signal. In this paper, we propose novel phase estimation methods by employing several temporal and spectral constraints imposed on the phase spec-trum of speech signal. We pose the phase estimation problem as es-timating the unknown clean speech phase at sinusoids observed in additive noise. To resolve the ambiguity in phase estimation prob-lem, we introduce individual time-frequency constraints: group de-lay deviation, instantaneous frequency deviation, and relative phase shift. Through extensive simulations, the effectiveness of the pro-posed phase estimation methods in single-channel speech enhance-ment is demonstrated. Employing the estimated phase for signal re-construction in medium-to-high SNRs leads to consistent improve-ment in perceived quality compared to when noisy phase is used. Index Terms — Phase estimation, single-channel speech en-hancement, time-frequency constraints, perceived speech quality. 1