Output details
13 - Electrical and Electronic Engineering, Metallurgy and Materials
University of Surrey
Video-Aided Model-Based Source Separation in Real Reverberant Rooms
This paper introduces a technique to overcome challenges for robust voice separation from multiple speakers in complex scenes, as the changing dynamics within a natural environment pose big challenges for robust voice separation. This solution is to integrate visual modalities for robust tracking of the speakers and identify unmixing filters for voice separation of moving speakers. Currently sponsored by EPSRC, the technique has a future in next-generation healthcare technology, surveillance and teleconference systems. The Centre for Defence Enterprise of the Ministry of Defence is interested in this technology for surveillance.