Blind Source Separation in Audio Processing Application Based on Direction of Arrivals

The proposed work deals with the development of signal processing algorithms to solve the source separation problem. The sound signals recorded in a real environment are not instantaneous but convolutive mixtures, because of the delay and the reflections. In order to separate these signals, it is effective to transform the signals into time-frequency domain. The difficulty in this approach is that the original signals are overlapping at the output stage i.e. suppose signal S1 and S2 are mixed in M1 and M2 mixtures, the mixed components of S1 and S2 will present in both M1 and M2 and while separating from these mixtures as Y1 and Y2 do not match S1 and S2 respectively in the sense components of S1 present in Y2 and vice-versa. This leads to ambiguity in separation and also BSS methods fail in the performance analysis because of permutation problem. This work proposes a novel approach to overcome this problem based on the direction of arrivals of the sources. In this work, clustering based algorithm is used to estimate unmixing matrix. Proposed algorithm is simulated using the MATLAB and the performance is validated in terms of source to distortion ratio (SDR), source to interference ratio (SIR) and source to artifact ratio (SAR) values. Simulation results show the superior performance of the proposed methods compared to other methods.