Overview
Researchers at the University of Maryland have developed a content-based audio classification algorithm based on novel multiscale spectro-temporal modulation features inspired by cortical processing.A potential use for the classification system is to discriminate speech from non-speech. Non-speech, for example, could consist of animal vocalizations, music, or environmental sounds. In head-to-head comparisons with two other state-of the-art approaches to discriminating between speech and non-speech, the multiscale spectro-temporal system performed significantly better.
These algorithms also have applications in audio and data retrieval, archival management, modern human-computer interfaces, and in the entertainment and security industries. The researchers are also working on developing algorithms to enhance speech in noisy environments using an auditory model.
For more information please contact 301-405-3947, E-mail: [email protected]
Contact Info
UM Ventures
0134 Lee Building
7809 Regents Drive
College Park, MD 20742
Email: [email protected]
Phone: (301) 405-3947 | Fax: (301) 314-9502