This is the website demo of the proposed source separation method submitted to ICASSP 2023.
REAL-TIME MULTICHANNEL SPEECH SEPARATION AND ENHANCEMENT USING A BEAMSPACE-DOMAIN-BASED LIGHTWEIGHT CNN
Marco Olivieri(1), Luca Comanducci(1), Mirco Pezzoli(1), Fabio Antonacci(1), Augusto Sarti(1),
Davide Balsarri(2), Luca Menescardi(2), Michele Buccoli(2), Simone Pecorino(2), Antonio Grosso(2)
(1): Dipartimento di Elettronica, Informazione e Bioingegneria - Politecnico di Milano, Milan, Italy
(2): BdSound S.r.l., Milan, Italy