Estimation of underdetermined mixingmatrix with unknown number of overlapped sources in short-time Fourier transform domain

The estimation of the mixing matrix as well as the number of sources in blind source separation are two challenging problems. This paper proposes an effective estimation method to solve these two problems for underdetermined blind separation of overlapped sources in short-time Fourier transform (STFT) domain. Our study considers the blind estimation of the mixing matrix based on subspace projection as well as clustering methods, and the number of sources can be therefore estimated by counting the columns of the estimated mixing matrix. The proposed estimation method is noise-robust and suitable for the sources whose spectral contents are highly overlapped in STFT domain. Numerical results on speech sources are presented to illustrate the effectiveness and robustness of the proposed method.