Blind Source Separation Using Time-Frequency Masking

Abstract

In blind source separation (BSS), multiple mixtures acquired by an array of sensors are processed in order to recover the initial multiple source signals. While a variety of Independent Component Analysis (ICA)-based techniques are being used, in this paper we used a newly proposed method: The Degenerate Unmixing and Estimation Technique (DUET). The method applies when sources are W-disjoint orthogonal; that is, when the time-frequency representations, of any two signals in the mixtures are disjoint sets. The method uses an online algorithm to perform gradient search for the mixing parameters, and simultaneously construct binary time-frequency masks that are used to partition one of the mixtures to recover the original source signals. Previous studies have demonstrated the robustness of the method. However, the investigation in this paper reveals significant drawbacks associated with the technique which should be addressed in the future.