System For Separating Multiple Acoustic Sources

DESCRIPTION (provided by applicant): We propose to develop a system for isolating and extracting one or more individual acoustic sources from the output of microphones responding to multiple simultaneous sources. Our system would make it possible to "clean up" live signals in real time by suppressing background noise and competing foreground sources. The proposed system will employ multiple microphones and algorithms to extract individual acoustic sources from noisy environments. A distinguishing characteristic of our system is that some microphones are used to "tag" known acoustic sources. Tagging microphones are attached to significant target or masking sources that are identified to the system. Other microphones are used to pick up background noise and remote (untagged) target or masking sources. The proposed system would be useful to researchers who need to create high-fidelity low-noise recordings in noisy environments, and who are not audio engineers. Our system would allow a user to tag the most prominent acoustic sources, record the entire "acoustic scene", and extract the desired acoustic source signals. The second important use of our system would be as an assistive listening device for persons with mild to moderate hearing loss, allowing them to function effectively in noisy social situations such as meetings, restaurants, and conferences.