UC San Diego Professor Receives Signal Processing Society Technical Achievement Award
San Diego and New Orleans, March 6, 2017 — University of California San Diego professor Bhaskar Rao will receive the 2016 Technical Achievement Award this week from the IEEE Signal Processing Society (SPS). The formal ceremony takes place in New Orleans during the 42nd IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP 2017).
In announcing the award, the society cited Rao’s “fundamental contributions to array processing and sparsity-based signal processing.” The award is one of the society’s most prestigious honors.
With publications in the area dating back to 1992, Rao was among the first to recognize that the concept of sparsity is fundamental – and that the underlying mathematical principles are important to the signal processing community. He co-organized the first special session on sparsity at ICASSP 1998 titled "SPEC-DSP: Signal Processing with Sparseness Constraint.”
“The problem of sparse signal recovery has recently received much attention with the development of compressed sensing and results providing insights into the robustness of l1 based recovery methods,” said Rao. “But the problem of computing sparse solutions to an underdetermined linear system of equations has a much longer history, and our work in this context began in the early 1990s, dealing with the problem of biomagnetic imaging.”
Rao – who has been a Fellow of IEEE since 2000 – is one of two engineers to pick up SPS Technical Achievement Awards at ICASSP 2017. The other is Jelena Kovačević of Carnegie Mellon University. The society cited Kovačević for her “contributions to the theory and practice of signal representations.”
While Rao was elected an IEEE Fellow for his work on the statistical analysis of subspace algorithms for harmonic retrieval, for the past two decades he has been a pioneer in the area of sparsity and its role in signal processing.
"This award recognizes Bhaskar's exemplary contributions to signal processing in every decade,” observed fellow ECE professor Piya Pal. “They include the first definitive analysis of polynomial rooting algorithms for harmonic retrieval in the eighties, and extensive development and analysis of Bayesian frameworks for sparse signal processing in the last decade."
Rao began developing novel measures and algorithms for sparse modeling. In a landmark paper* from 1997, his student Irina F. Gorodnitsky and he developed FOCUSS, a re-weighted l2 norm minimization algorithm for sparse signal reconstruction from limited data.
Some of Rao’s significant contributions resulted from collaborations with other ECE faculty members, including Ken Kreutz-Delgado. The two published a paper** with his student Shane F. Cotter in July 2005 in IEEE Transactions on Signal Processing on sparse solutions to linear inverse problems with multiple measurement vectors. “That paper represents a significant generalization of the sparsity framework,” said Kreutz-Delgado. “It extended the framework to deal with multiple measurements vectors – greatly enhancing the application domain.”
In 2011 Rao collaborated with fellow ECE professor Young-Han Kim and then-student Yuzhe Jin (Ph.D. ’11) on a paper*** in information theory about limits on support recovery of sparse signals. According to Kim, the paper was important for two reasons. “It provides a completely new way of thinking about sparse signal processing in the framework of multiple access communications," said Kim. "This novel framework, in turn, establishes a tight performance bounds on the limit of sparse signal recovery."
His student David P. Wipf and Rao also developed a leading algorithm for sparse signal recovery based on sparse Bayesian learning that remains in widespread use today. “That paper**** in 2004 and follow-up work in 2007 with Wipf and 2011 with Zhilin Zhang represent a seminal contribution to algorithmic work,” said ECE professor and chair Truong Nguyen.
Professor Rao is also at ICASSP 2017 because he is a co-author of three papers to be presented at the conference with Ph.D. students and/or fellow UC San Diego faculty co-authors. The papers are scheduled for presentation on Tuesday, Wednesday and Thursday, respectively. Rao is senior author on all three:
- Multimodal Sparse Bayesian Dictionary Learning Applied to Multimodal Data Classification, co-authored with Igor Fedorov and ECE Prof. Truong Q. Nguyen. First-author ECE Ph.D. student Fedorov (who expects to complete his doctorate in 2017) is a member of Rao’s Digital Signal Processing Lab as well as Nguyen’s Video Processing Lab. The paper is part of the machine learning for signal processing track.
- Sparsity Regularized Principal Component Pursuit, co-authored with ECE Prof. Pamela Cosman and her Ph.D. student, Jing Liu. The paper is scheduled during the conference’s signal processing theory and methods track.
- Multivariate Scale Mixtures for Joint Sparse Regularization in Multi-Task Learning, joint with Ritwik Giri, in the track on machine learning for signal processing. ECE alumnus Giri (Ph.D. ’16) graduated with a dissertation on Bayesian sparse signal recovery. (At right: Giri's final defense of his dissertation, introduced by his advisor, Bhaskar Rao.)
Immediately after receiving his Ph.D. in electrical engineering from the University of Southern California in 1983, Rao joined the ECE faculty as an assistant professor in the Jacobs School of Engineering. In 1989 he joined the ECE faculty as an assistant professor. He advanced to full professor in 1995.
Since 2008, Rao has held the Ericsson Endowed Chair in Wireless Access Networks. He is also an active participant in the Center for Wireless Communications (and its director from 2008 to 2011), as well as the California Institute for Telecommunications and Information Technology (Calit2) and its UC San Diego division, the Qualcomm Institute.