Paris Smargdis (U. of Illinois) - Models for Mixed Sounds
Abstract: Dealing with superimposed signals is one of the most fun and challenging problems we find in audio and speech processing. In this talk I'll present a principled approach to this problem by taking advantage of the specific geometry that mixed signals exhibit. Using that representation, and circumventing obvious combinatorially-hard solutions, I'll introduce some probabilistic algorithms that can help in efficiently solving a variety of problems involving mixed signals. Emphasis will be placed on the idea that source separation is not always the most desirable goal and I'll show how what we have learned about source separation can be recast for use in much more practical problems.
Biography: Paris Smaragdis is an assistant professor in the Computer Science and the Electrical and Computer Science departments at the University of Illinois at Urbana-Champaign. He completed his graduate and postdoctoral studies at MIT, where he conducted research on computational perception and audio processing. Prior to the University of Illinois he was a senior research scientist at Adobe Systems and a research scientist at Mitsubishi Electric Research Labs, during which time he was selected by the MIT Technology Review as one of the top 35 young innovators of 2006 for his work on machine listening. Paris' research interests lie in the intersection of machine learning and signal processing and he loves to scheme against old and established approaches.