Contrast functions for blind source separation based on time frequency distributions
Adel Belouchrani, Karim Abed‐Meraim · 2004
This paper introduces new source separation techniques exploiting the time frequency signatures of the source signals. The proposed approach relies on time frequency distributions (TFD). Two TFD-based contrast functions are presented for non stationary sources. Iterative algorithms using the relative gradient technique are used to optimize the proposed contrast functions and perform the source separation.