My current research
focuses on the following:
Sensing/Sampling. In the classical setting, a measurement setup
remains fixed throughout the process of data acquisition. In adaptive
sensing, at every time step, the measurement setup is altered based on
to overall maximize the information of interest. Some of the active
areas of application of this concept
appear in mine detection, see through the wall, SAR imaging, and target
- Manifold Learning.
Manifolds offer the capability to describe high dimensional data using
a low dimensional representation. Dimensionality
reduction of high-dimensional data that lies on a manifold allows
visualization of the data, reduction in computational complexity of
data processing, and the capability of intrinsic data processing. Areas
of application include: medical diagnosis, target recognition, analysis
of internet data, and sensor networks.
- Sparse Representations
for Signal Processing. We are interested in investigating data
sparse according to some basis or dictionary. In other words, the data
can be represented using only a small number of basis/dictionary
elements. Image compression methods, which are based
on vector quantization, demonstrate that an image can be represented in
a sparse fashion through fixed bases, e.g., discrete cosine transform
(DCT) and wavelets. Sparse reconstruction takes advantage of the signal
sparsity in reproducing the signal from partial and/or noisy
observations. Areas of application include: electromagnetic
imaging, molecular imaging, and sensor/waveform selection.
Last updated December 14, 2007.