Columbia EE : LabROSA :
[speechsep image]

Separating Speech from Speech Noise


The task of separating speech in complex acoustic environments -- such as a single voice in a cocktail party -- is an extremely difficult challenge. Many speech enhancement or separation techniques cannot accommodate the situation when both target and interference have the same properties, because both are speech. This project is concerned with applying some novel models -- using Computational Auditory Scene Analysis (CASA) and trained models of the speech signal -- to see how well speech can be separated. In particular, our goal is to provide separations that are demonstrably of benefit to human listeners, hence our collaboration with perceptual experimentalists at EBIRE and Boston University.

Partners

Resources

Project Reports

Theses

Related Publications

2009

2008

2007

Acknowledgment

This material is based in part upon work supported by the National Science Foundation under Grant No. IIS-05-35168. Any opinions, findings and conclusions or recomendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation (NSF).


Valid HTML 4.0! Last updated: $Date: 2005/08/09 03:26:12 $
Dan Ellis <dpwe@ee.columbia.edu>