Department of Computer Science and Engineering

Regional Campus Faculty Candidate

Button Binaural Sound Segregation in Multisource and Reverberant Environments

Nicoleta Roman
Department of Mathematics
The Ohio State University, Lima

Aug 3 2009 10:30AM
280 Dreese Labs
All interested parties are welcome.
Refreshments will be served prior to the talk.

Abstract:
Sound separation in a noisy reverberant environment is an important and unsolved problem in many applications including automatic speech recognition (ASR), speaker recognition, audio information retrieval, and intelligent hearing aids. While machine separation remains a great challenge, the human auditory system shows a remarkable ability for sound separation. This perceptual ability has motivated the emerging field of computational auditory scene analysis (CASA), which aims to build sound separation systems based on principles of human hearing. A main computational goal of CASA systems is an ideal time-frequency (T-F) binary mask, which selects the target if it is stronger than the interference in a local T-F unit. In this talk I will present binaural CASA approaches that attempt to estimate this binary mask based on the location information of individual sources. Listening tests as well as systematic evaluations in terms of speech recognition performance have shown very promising results with multiple interfering speakers as well as room reverberation.

Bio:
Nicoleta Roman received the B.S. and M.S. degrees in Computer Science from the University of Bucharest, Bucharest, Romania, in 1996 and 1997, respectively, and the Ph.D. degree in Computer Science and Engineering from The Ohio State University, Columbus, in 2005. Since 2005, she has been a Visiting Assistant Professor in the Department of Mathematics, The Ohio State University, Lima. Her research interests include computational auditory scene analysis, binaural processing, and machine learning.

Host: Srini Parthasarathy

OSU logo