Time-Frequency Masking for Blind Source Separation with Preserved Spatial Cues

Abstract

In this paper, we address the problem of speech source separation by relying on time-frequency binary masks to segregate binaural mixtures. We describe an algorithm which can tackle reverberant mixtures and can extract the original sources while preserving their original spatial locations. The performance of the proposed algorithm is evaluated objectively and subjectively, by assessing the estimated interaural time differences versus their theoretical values and by testing for localization acuity in normal-hearing listeners for different spatial locations in a reverberant room. Experimental results indicate that the proposed algorithm is capable of preserving the spatial information of the recovered source signals while keeping the signal-to-distortion and signal-to-interference ratios high.

Phoneme-Based Distribution Regularization for Speech Enhancement

“…To address this problem, speech enhancement is usually performed to separate the clean speech from background noises with reverberation. Thanks to the advancement of computational auditory scene analysis [3,4], speech enhancement has been viewed as a supervised learning problem. Existing speech enhancement methods mainly learn a deep-learning-based neural network, which can effectively model the relationship between a noisecorrupted signal and the clean speech.…”

Monaural Speech Enhancement with Dilated Convolutions

“…For decades, monaural speech separation has been studied for speech processing applications. Inspired by Time-Frequency (T-F) masking based on computational auditory scene analysis (CASA) [1,2], monaural speech separation has been recently treated as a supervised learning problem. The choice of training model, acoustic features, and training target are all important factors for accurate supervised speech separation.…”