April 9, 2007
Speaker: Michael S. Pedersen, Oticon Research Center, Denmark
This wednesday, April 11th, Michael Syskind Pedersen of Oticon in Denmark (one of the world's biggest hearing aid companies) will be visiting to give a seminar. Michael has developed some very clever and practical audio source separation techniques that can recover multiple voices from just two microphone channels. I believe he has some very nice and insightful demonstrations. Please join us.
The human ability to pay attention to a single sound in the presence of many simultaneous talkers is known as the cocktail party phenomenon. Separation of speech mixtures, often referred to as the cocktail party problem, has been studied for decades. Especially hearing impaired people find it difficult to cope with situations when many people speak simultaneously. Today's modern hearing aids therefore attempt to enhance speech in noisy situations, either by use of a single microphone recording or multiple microphone signals.
In many source separation tasks, the separation method is limited by the assumption of at least as many sensors as sources. A novel method for underdetermined blind source separation using an instantaneous mixing model which assumes closely spaced microphones is presented. Two source separation techniques have been combined, independent component analysis (ICA) and binary time-frequency masking. By estimating binary masks from the outputs of an ICA algorithm, it is possible in an iterative way to extract basis speech signals from an instantaneous mixture as well as a convolutive mixture. In principle an arbitrary number of mixed speech signals can be separated using only two microphones.
Michael Syskind Pedersen received the M.S. degree in 2003 from the Technical University of Denmark (DTU). In 2006 he received a Ph.D. degree from the Section for Intelligent Signal Processing at the Department of Mathematical Modelling, DTU. In 2000 he was an exchange student at New Jersey Institute of Technology, Newark, NJ. In 2005 he was visiting Professor DeLiang Wang at the Department of Computer Science and Engineering at The Ohio State University, Columbus, OH. Since 2001 Michael has been employed with the hearing aid company Oticon.