The
research work presented in this talk is on the topic of blind
separation of the mixed acoustic signals in the presence of noise.
The term "blind separation" means that the separation
is done without any prior knowledge of the original signals. To
this end, we proposed a new ICA algorithm, the subband-based ICA
algorithm. It is an extension of the traditional Independent Component
Analysis (ICA). The experimental results demonstrated the effectiveness
and robustness of the new algorithm.
work
with: P.S. Krishnaprasad and S. Shamma
|