Anechoic phase estimation from reverberant signals

Abstract

Most dereverberation methods aim to reconstruct the anechoic magnitude spectrogram, given a reverberant signal. Regardless of the method, the dereverberated signal is systematically synthesized with the reverberant phase. This corrupted phase reintroduces reverberation and distortion in the signal. This is why we intend to also reconstruct the anechoic phase, given a reverberant signal. Before processing speech signals, we propose in this paper a method for estimating the anechoic phase of reverberant chirp signals. Our method presents an accurate estimation of the instantaneous phase and improves objective measures of dereverberation.

Amplitude and phase dereverberation of monocomponent signals

“…We compare our approach with a state-of-the-art suppression method [17], which focuses only on the magnitude of the STFT and ignores the phase information. Our previous work [11] is not included in the benchmark, as it can only deal with linear frequency modulation.…”

“…We know from [10] that phase estimation improves the predicted speech quality and we proposed in [11] a dereverberation method that uses phase information to improve the dereverberation performance. However, this method is restricted to linear chirp signals.…”

Amplitude and phase dereverberation of harmonic signals

“…This is the main drawback of these methods, because using this corrupted phase reintroduces reverberation and distortion in the signal, as shown in [5]. We know from [6] that phase estimation improves the predicted speech quality and we proposed in [7] a dereverberation method that uses phase information to improve the dereverberation performance. However, this method was restricted to linear chirp signals, assuming the signal amplitude to be known, and it focused on the estimation of the anechoic phase.…”