In this talk I will discuss the problem of source separation in the presence of noise and when the number of sources can be larger than the number of sensors. I will highlight approaches under classes of assumptions that make the problem tractable, and focus on one class of assumptions related to sparse representations of signals that leads to effective algorithms. The specific sparseness assumption used in this work is that the maximum number of statistically independent sources active at any time and frequency point in a mixture of signals is one or small. It allows us to solve the maximum likelihood formulation of the source estimation problem. Solutions obtained are applicable to an arbitrary number of microphones and sources, but work best when the number of sources simultaneously active at any time frequency point is a small fraction of the total number of sources.
Biography: Justinian Rosca received his M.S. and Ph.D. in Computer Science from the University ofB in 1994 and 1997, respectively. Previously, he received his Dipl.-Eng. in Computers and Control Engineering from Polytechnic University of Bucharest. He is currently program manager at Siemens Corporate Research in Princeton NJ, where he heads the Audio, Signal Processing and Wireless Program focusing on statistical signal processing. Justinian's research interests are in the areas of machine learning and signal processing, in particular fundamentals of speech representation and processing, sparse representations, blind signal separation, adaptive principles in stochastic search and optimization, and probabilistic inference. He has given tutorials on stochastic search techniques at the Genetic Programming and Genetic and Evolutionary Computation conferences in 1998 and 1999. Within Siemens he has been contributing to next generation technology for hearing aids, mobile and WLAN phones. Justinian has more that three dozen patent applications, and more than 75 reviewed publications. He co-authored a book on solved problems in higher mathematics. He is presently on the editorial board of the Journal for Genetic Programming and Evolvable Hardware and serves on conference committees in the areas of machine learning and signal processing. He organized the Sparse Representations in Signal Processing workshop at NIPS 2003 and is the Program Chair of the ICA/BSS 2006 Conference.