The three broad areas of emphasis in the department are communications, signal
processing, and photonics. The interest of the faculty and their current research
falls generally into one of these areas.
Communications
Research in communications is in the areas of antenna design, wireless communications,
computer and communication networks and network and data security. Research
in antennas is focused on developing flat-panel antennas for US DBS satellite
communication systems, and on small and efficient antenna designs for mobile
communication applications. Another topic is development of software for
design and analysis of antennas based on the finite element method (FEM)
for use with large arrays.
In wireless communications, algorithms are being developed for high speed
communications under the IEEE 802.11a standard. Efforts are focused on developing
technologies that will support indoor and outdoor wireless connections exceeding
100Mbits/second. Another area of research is in power adaptation techniques
for wireless multimedia. Focusing on the design of power efficient transmission
policies for wireless multimedia, this research explores the tradeoff between
transmission power, capacity and communication latency. Minimizing the transmission
power is very critical for many reasons. First, lower transmit power translates
directly into longer battery life. Second, lower transmit power implies lower
interference to other devices operating in the same area and consequently higher
capacities in terms of data rates or number of users supported. The new policies
can result in substantial savings in transmission power (up to 60%) for small
increases in packet delay.
Research in communication networks is aimed at developing model classification
algorithms to automatically fit network traffic to stochastic models to enable
service providers to automatically adapt their traffic control to changing
network conditions. New real-time traffic measurement capabilities for routers
are being developed by defining a "traffic profile" to capture the most important
aspects of traffic flows in a compact mathematical form. These traffic metering
capabilities will enable routers to provide traffic measurement data for a
variety of applications such as traffic control, network planning, security,
and accounting. Finally new models to analyze the cost/benefit trade-offs for
mobile code are being developed by applying risk analysis used in other fields
to quantify the security risks for mobile code. Another project is concerned
with accurate modeling and prediction of network traffic for traffic management
for optimal allocation of available resources. In this work, intelligent-system
based models such as artificial neural networks and fuzzy logic investigated
for their to characterize high-speed traffic at different time scales.
A new scheme for robust coding of video over noisy channels and transmission
to reduce error propagation is being investigated by using a tree structure
by classifying frames into three different categories. The structure introduces
different levels of priority to the frames being coded. Different levels of
error protection can be assigned to packets from different types of frames.
It is expected that a combination of data partitioning with the proposed scheme
will prove beneficial.
Signal Processing
Research in signal processing covers a wide range of application areas. Blind
source separation seeks to undo the acoustical mixing caused by room acoustics
in multichannel recordings of acoustic events without knowledge of the sound
sources, the room acoustics, or the physical arrangement of sources or sensors
in the room. Applications include teleconferencing, audio recording, and
surveillance. A related problem is that has applications in hearing aids
is the removal of echo and reverberation caused by room acoustics that makes
speech unintelligible.
In biomedical signal processing, human vision models are being developed
which link visual acuity to eye motion. Blind system identification algorithms
that are robust and provide a substantial improvement in the resolution and
quality of medical ultrasound images are being investigated. Another area of
interest is the use of subspace methods for speech enhancement and coding.
Other projects explore the use neural networks, fuzzy approaches and genetic
algorithms to problems of interest. New approaches to short-term power system
load forecasting in a deregulated and price-sensitive environment are being
developed for a real-time pricing scenario where energy prices could change
on an hourly basis. A two-stage load forecaster has been developed with the
first stage consisting of a price-insensitive neural network forecaster known
as ANNSTLF and a second stage fuzzy logic based module which transforms the
price-insensitive forecast of ANNSTLF to a price sensitive load forecast. A
second project focuses on developing fuzzy logic based pattern classifiers.
A genetic algorithm (GA) based approach is being developed to optimize all
the required fuzzy system parameters. The resulting fuzzy classifier will be
tested for several different classification applications and its performance
will be compared to other types of classifiers.
In many applications such as video browsing, indexing of relevant scenes
in a video sequence is important for their efficient retrieval. Such indexing
is most commonly done by identifying scene cuts that represent the boundary
between video shots. Scene cut detection involves the identification of frames
at which the content of the scene is significantly different from that of the
previously retained frames. Traditional shot detection methods are sensitive
to noise and often turn out to be inefficient in dealing with the so-called
gradual scene changes. This research is aimed at developing robust techniques
for identifying scene cuts and gradual scene changes.
Photonics
Research in photonics is aimed at fabrication of an efficient monolithic laser
source and fiber-like waveguide for use in wavelength division multiplexed
(WDM) systems. The scheme for coupling light from a laser to a fiber allows
for lower fabrication and production costs by obtaining high coupling efficiencies
of light from lasers to glass waveguides. Another research project is concerned
with determining grating strengths of surface emitting lasers by analyzing
the reflection and transmission characteristics of a quantum-well structure
with a finite length grating. Long wavelength (1310 and 1550 nm) grating-outcoupled
surface-emitting lasers are also being investigated for telecommunications
applications. Other programs are concerned with developing high power (1
to 10 Watts), highly coherent, single-frequency semiconductor lasers; coherent
arrays of high power vertical cavity semiconductor lasers; and radio frequency
gyroscopes designed to operate in the 50 to 100 GHz range of frequencies.
One of the greatest barriers to the insertion of optical interconnections
in systems is cost. An interconnection medium that is immune to misalignments
would radically lower the cost barrier for the insertion of optical interconnections
in systems. Graded index material, developed extensively for reducing modal
dispersion in fiber optics, has a potential application for alignment insensitive
multi-chip optical interconnections as well. Experiments will be conducted
to integrate and evaluate an optical interconnection architecture based on
graded index slab material.
