Current Research Interests
My research interests are in the general field of digital signal
processing, particularly in multirate systems, filter banks,
transforms, wavelets, and their applications in signal analysis,
sparse representation, compression, processing, and communications.
My research group's focus
is on providing efficiency, versatility, flexibility, adaptivity,
and scalability in all aspects of a modern state-of-the-art digital
signal processing/communication systems. Although practicality is high
on our list, the starting point has always been developing fundamental
tools for signal processing with a particular emphasis on understanding
the theoretical foundation.
Since 2006, my research interest has switched to the emerging area of
compressed sensing (CS). Here, one seeks to acquire a small number of
digital measurements (sampling) of a sparse signal directly in
``compressed format'' (often achieved by a series of random linear
projections) and rely on sophisticated reconstruction algorithms
such as basis pursuit or matching pursuit to recover the signal
from these measurements. To date, work on compressed sensing has mainly
explored various theoretical aspects and mathematical explanations
of the framework. My group focuses on developing practical large-scale
compressed sensing algorithms with fast, efficient, and hardware-friendly
implementations. Another area of interest is novel applications of
compressed sensing to natural speech, audio, image, and especially
video sequences.
A few current on-going research projects in my group include:
- applying CS principle to fast video sensing: particularly,
video frames are intra-sampled by a CS encoder while they are
inter-decoded at the decoder side, exploiting temporal correlation
among frames
- fast and practical recovery of video data (denoising,
super-resolution, multi-view...) via a simple,
yet very effective, inter-frame sparsity model
- compressed sensing over packet-switched networks:
optimal algorithms for sending CS measurements and reconstructing signals
over practical networks; this effort includes progressive CS, sequential CS,
and multiple description CS
- robust multimedia transmission via compressed sensing:
taking advantage of error sparsity for robust recovery of
multimedia
- robust face and pattern recognition via compressed sensing:
again taking advantage of error sparsity and signal inter-correlation
- exploring the addition of local CS measurements
and side information to improve
reconstruction performance on fast time-varying and sparsity-varying signals
- applying CS principle to fast SVD approximation and rank
minimization of huge matrices and applications in fast database search.
Students
Alumni
- Lijie Liu,
Ph.D. 12/2008.
Dissertation: On Filter Bank and Transform Design with the Lifting Scheme
Current Position: in MBA program,
Rotman School of Management, University of Toronto,
ON , M5S 3E6, Canada.
- Carlos L. Salazar,
Ph.D. 05/2007.
Dissertation: Computationally Scalable Spatial Resizing of DCT Domain Compressed Images and Video
Current Position: Researcher,
National Security Agency (NSA),
Fort Meade, MD, USA.
- Wei Dai,
Ph.D. 05/2005.
Dissertation: Adaptive Block-based Decomposition with Pre-/Post-Filtering
Current Position: Member of Technical Staff,
FastVDO LLC,
Columbia, MD, USA.
- Jie Liang, Ph.D. 12/2003.
Dissertation: Filter Bank Design for Image/Video Compression and Digital Communications
Current Position: Assistant Professor, School of Engineering Science,
Simon Fraser University, Burnaby, BC, Canada.
- Chengjie Tu, Ph.D. 12/2003.
Dissertation: Pre-/Post-Filtering and Context Modeling for DCT Based Block Coding Systems
Current Position: Software Design Engineer,
Microsoft Corporation,
Redmond, WA, USA.