On blind separation of nonstationary signals
L.A. Cirillo, Abdelhak M. Zoubir · 2006
In this paper we consider a time-frequency based approach to blind separation of nonstationary signals. In particular, we propose a time-frequency ‘point selection’ algorithm based on multiple hypothesis testing, which allows automatic selection of auto- or cross-source locations on the time-frequency plane. The chosen points are used via a joint diagonalization and off-diagonalization algorithm to perform source separation. A performance comparison of the proposed and existing approaches is provided.