A Time-Frequency Algorithm for Noisy BSS Model

In most practical blind source separation (BSS) applications, the observations contain additive source noise that limits the performances of most existing BSS algorithms. In this paper, we propose a new BSS approach exploiting the difference in the time-frequency (t-f) signatures of these sources to be separated. The approach uses smooth pseudo Wigner-Ville distribution (SPWVD) to obtain t-f distribution, then localizes the signal energy by Hough transform and selects a set of spatial t-f points based on the dominant eigenvalue of SPWVD of observations. Finally, numerical performance simulations are provided highlighting its effectiveness.