PROJECT TITLE :

Sparsity And Low-Rank Amplitude Based Blind Source Separation - 2017

ABSTRACT:

This paper presents a new technique for blind supply separation drawback in reverberant environments with additional sources than microphones. Primarily based on the sparsity property within the time-frequency domain and also the low-rank assumption of the spectrogram of the source, the STRAUSS (SparsiTy and low-Rank AmplitUde based mostly Source Separation) method is developed. Numerical evaluations show that the proposed technique outperforms the existing multichannel NMF approaches, while it is completely based mostly on amplitude info.