DSP Group Meeting Schedule

Future Speakers


TIME/PLACE: Friday September 19 at 14:30 in EOW 430
SPEAKER: Andreas Antoniou
TITLE:
ABSTRACT:

Approximate Presentation Order for DSP Group Meetings


NOTE: This order may need to be modified slightly from time to time
due to absences, etc..

Peter Hampton  (Supervisor: Pan Agathoklis)
Wu-Sheng Lu 
Michael McGuire 
Diego Felix  (Supervisor: Michael McGuire)
Mohamed Yasein  (Supervisor: Pan Agathoklis)
Yihai Zhang  (Supervisor: Wu-Sheng Lu and Aaron Gulliver)
Deepali Arora  (Supervisor: Pan Agathoklis)
Eugene Hyun  (Supervisor: Michael McGuire)
Pan Agathoklis 
Ping Wan  (Supervisor: Michael McGuire)
Ze Dong  (Supervisor: Michael Adams)
Najith Liyanage  (Supervisor: Pan Agathoklis)
Michael Adams 
Diego Sorentino  (Supervisor: Andreas Antoniou)
Parameswaran Ramachandran  (Supervisor: Wu-Sheng Lu and Andreas Antoniou)
Ana Maria Sevcenco  (Supervisor: Wu-Sheng Lu)

Past Speakers


TIME/PLACE: Friday August 8 at 14:30 in EOW 430
SPEAKER:
TITLE:
ABSTRACT:
CANCELLED

TIME/PLACE: Thursday July 24 at 15:45 in EOW 430
SPEAKER:
TITLE:
ABSTRACT:
CANCELLED DUE TO LACK OF SPEAKER

TIME/PLACE: Friday July 11 at 14:30 in EOW 430
SPEAKER:
TITLE:
ABSTRACT:
CANCELLED DUE TO LACK OF SPEAKER

TIME/PLACE: Thursday 26 June at 15:45 in EOW 430
SPEAKER:
TITLE:
ABSTRACT:
CANCELLED DUE TO LACK OF SPEAKER

TIME/PLACE: Friday June 13 at 14:30 in EOW 430
SPEAKER: Deepali Arora
TITLE: Addressing the hidden beam problem using a modified beamformer
ABSTRACT:

The advancements in antenna technology and their use in WLANs such as
IEEE 802.11 has shown that the directional antennas are preferable over
omni-directional antennas not only in terms of increased range, better
directivity and substantial reduction in interference but also in large
power savings which is a critical issue in wireless systems. However,
when used in IEEE 802.11 systems that are based on carrier sensing,
directional antennas present their own share of issues such as the
hidden node problem, the hidden beam problem and deafness.

In this seminar a new method to deal with the hidden beam problem in
CSMA based systems will be presented. The method is based on a
modification of the beamformer weights to reduce null depths of an
original beam. The differences in the original and modified beam
patterns are quantified in terms of half power beam width (HPBW) and
directivity which shows that the modified beam yields significant
improvements in null reduction with relatively small changes to
directivity and HPBW. The performance of the modified beam obtained is
compared with that of a original beam in terms of reducing hidden beam
problem when used in non-persistent CSMA systems. The proposed
methodology is simple and effective in solving the hidden beam
problem.

TIME/PLACE: Thursday May 29 at 15:45 in EOW 430
SPEAKER: Yihai Zhang 
TITLE: Blind Polynomial Channel Estimation for OFDM Systems
ABSTRACT:
Orthogonal frequency division multiplexing (OFDM) modulation is widely
used in communication systems to meet the demand for ever increasing
data rates.Characteristics of the transmitted signal can be employed
for blind channel identification. We propose a blind polynomial channel
estimation algorithm using noncircular second-order statistics of the
received OFDM signal. A set of polynomial equations are then formulated
based on the correlation of the received signal.  The solution of these
equations provides an estimation of the channel coefficients. Results
are presented which show that the proposed algorithm provides
performance comparable to LMMSE with lower computational complexity.
The performance is near-optimal for large OFDM systems.

TIME/PLACE: Friday May 16 at 14:30 in EOW 430
SPEAKER: Ana-Maria Sevcenco
TITLE: Medical image segmentation based on adaptive thresholding,
morphological operations and level set methods

ABSTRACT:
Image segmentation of multiple regions plays a critical role in a
variety of image analysis and visualization tasks. Segmentation can be
achieved by several methods such as clustering, histogram-based
methods, edge detection, region growing techniques, level set methods,
and multi-scale segmentation. It is not surprising that performance
enhancement can be achieved by appropriately combining some of these
techniques.  In this talk we present two algorithms for liver
segmentation -- an important task in the development of computer-aided
multi-phase CT diagnosis system for tumor identification in liver.  The
first algorithm is an integrated liver segmentation method based on
adaptive thresholding techniques, morphological operations and edge
detection, where image's local intensity is used as a similarity
measure for region segmentation. The second algorithm employs in
addition a level-set framework for a more accurate segmentation and
region identification. Both algorithms are successfully tested on a set
of 98 real CT scan images. We present some of the experimental results
and a performance analysis to illustrate the efficiency of the proposed
segmentation methods.

TIME/PLACE: Friday May 2 at 14:30 in EOW 430
SPEAKER: Parameswaran Ramachandran
TITLE:
ABSTRACT:

TIME/PLACE: Friday April 18 at 14:30 in EOW 430
SPEAKER: Diego Sorrentino
TITLE: Multiframe Image Super-Resolution Using Quasi-Newton Algorithms
ABSTRACT:

Multiframe image super-resolution algorithms can be used to reconstruct
a high-quality high-resolution image from several warped, blurred,
undersampled, and possibly noisy images acquired from the same scene.

A widely used means of implementing such algorithms is by
optimization-based model inversion. In the past, steepest-descent
methods have been applied. Nonetheless, while easy to implement, these
methods are known for their poor convergence properties and for being
sensitive to numerical ill-conditioning.

In this talk, we show how the multiframe super-resolution problem can
be tackled using powerful quasi-Newton algorithms and provide details on
efficient implementations that are well suited given the large scale nature
of the problem. In addition, we present results of preliminary testing
that indicate a significant improvement in terms of convergence speed
with respect to steepest-descent implementations.

TIME/PLACE: Friday April 4 at 14:30 in EOW 430
SPEAKER: Michael McGuire
TITLE: Bounds on MSE for Non-Linear Estimation Problems
ABSTRACT:

An important issue for parameter estimation problems is to determine if
the desired level of accuracy is achievable with the available
measurements. For this lower bounds on the mean square error (MSE) of
the estimator are often employed.  The most popular lower bound is the
Cramer-Rao lower bound. This bound assumes only the measurements are
used to estimate the parameters of interest.  Prior information on the
parameter values of interest is not included in the bound calculation.
The Cramer-Rao lower bounds also has required regularity conditions
which are often violated in non-linear estimation problems.

In this talk, we will discuss more general lower bounds on the MSE such
as the Bayesian Cramer-Rao, Weinstein-Weiss and extended Ziv-Zakai
lower bounds.  These bounds account for prior information in the
estimation process and can handle general measurement models.
Conditions will be demonstrated when these bounds are tight to actual
estimator performance.

This talk will also discuss the use of these bounds for model-based
estimation approaches such as the use of Extended Kalman filters and
particle filters to estimate parameters of an evolving process from
measurements made at different times.  The extension of the standard
estimation lower bounds to these cases will be demonstrated. Some of
the more common cases where this extension can go wrong are also
discussed.

TIME/PLACE: Wednesday March 19 at 15:45 in EOW 430
SPEAKER: Wu-Sheng Lu
TITLE: Linear updates by SOCP for power symmetric perfect reconstruction
  filter banks
ABSTRACT:
Two-channel power-symmetric FIR filter banks are among the most popular
building blocks for multirate systems as they offer precise perfect
reconstruction (PR) property.  Many design methods for this class of
filter banks have been proposed since early 1980s.  This talk gives a
brief review of several important design methods with comments, and
describes a new method. The new method is based on a linear updating
strategy that can be readily realized with a second-order cone
programming (SOCP) solver.  We apply this technique to deduce two
design formulations, one with a least squares objective function and
the other with a minimax criterion. Examples are supplied to help
examine design performance and efficiency.

TIME/PLACE: Friday March 7 at 14:30 in EOW 430
SPEAKER: Diego Sorrentino
TITLE: Image Super-Resolution, an Overview
ABSTRACT:

As the number of digital imaging based applications grows, so does the
demand for higher quality images. Applications like surveillance,
remote sensing, and medical imaging greatly benefit from the
availability of high quality high resolution (HR) images.  The usual
approach to obtain higher resolution images implies building systems
with higher pixel count sensors. Unfortunately, this trend is reaching
a technological limit by which sensors pixel count cannot be increased
without severely degrading their overall quality.

One promising strategy to build systems that acquire HR images using
inexpensive low-resolution (LR) sensors is by resorting to signal
processing techniques known as 'Super-Resolution' (SR). SR algorithms
exploit the availability of several LR images from the same scene to
reconstruct a high quality HR version of it.

In this talk we provide an introduction to image super-resolution, as
well as an overview of some the most prominent techniques.

TIME/PLACE: Friday February 22 at 14:30 in EOW 430
SPEAKER:
TITLE:
ABSTRACT:
Rescheduled due to reading break.

TIME/PLACE: Friday February 8 at 14:30 in EOW 430
SPEAKER: Peter Hampton
TITLE: A new wave-front reconstruction method for adaptive optics systems
  using wavelets
ABSTRACT:

A new technique for wave-front reconstruction from gradient
measurements is presented.  This technique is based on obtaining the
Haar wavelet image decomposition of the original image by appropriate
filtering of the gradient measurement data.  The proposed technique is
illustrated with an example and its reconstruction accuracy and the
influence of the input SNR are discussed.  Results indicate that the
proposed technique is a computationally efficient and accurate
technique for image reconstruction from gradient measurements.

TIME/PLACE: Thursday January 24 at 15:45 in EOW 430
SPEAKER: Najith Liyanage
TITLE: Spectral Characteristics of 3D Windowed and Sampled Linear
  Trajectory(LT) Space-Time Signals
ABSTRACT:

Linear Trajectory Signals(LT) are an especially important class of MD
signals. They occur in many kinds of signal processing sytems, including
television, radar and seismic signal processing. An MD signal is a LT
signal if there exist a constant MD vector n such that the directional
derivative of the LT signal along n is zero everywhere in the domain of
the signal. In this talk we focus on a special LT signal with a square
object.

First, the Region of Support (ROS) of the spectrum of such a Linear
Trajectory (LT) signal and how it deviates from the ideal spectrum due to
practical processes 2D windowing and 3D spatio-temporal sampling are
discussed. Then, their ROS of the spectra are algebraically analyzed based
on the changes of parameters such as the speed of the object, the sample
rates and the window size. Finally, the algebraically explained
characteristics of the ROS of the spectra are simulated and verified using
Matlab.

TIME/PLACE: Friday January 11 at 14:30 in EOW 430
SPEAKER:
TITLE:
ABSTRACT:
*** CANCELLED ****

TIME/PLACE: Friday December 7 at 14:30 in EOW 430
SPEAKER: Ping Wan
TITLE: Channel Estimation for OFDM in Fast Fading and Frequency Selective
  Channels
ABSTRACT:
Orthogonal frequency-division multiplexing (OFDM) is an efficient
technique for high data rate transmission in wireless communications. A
cyclic prefix (CP) is inserted at the transmitter and removed at the
receiver. If the length of CP is equal to or long than the delay spread
of the channel, intersymbol interference (ISI) due to multipath
propagation is eliminated. In slow fading channels, the channel is
stationary over one OFDM symbol, a simple one-tap equalizer can be
employed. However, in fast fading channels, the channel is variant over
the period of an OFDM symbol, the orthogonality between subcarriers is
destroyed, resulting in intercarrier interference (ICI) which impacts
the bit-error rate (BER) performance. To mitigate ICI, channel
equalization schemes are employed which rely on accurate channel
estimation. The channel is estimated using pilot symbol assisted
modulation (PSAM). In fast fading channels, due to the interference
from the neighboring symbols in the frequency domain, pilot symbols are
inserted in the time domain. By taking the traditional Rayleigh fading
channel into account, several channel estimators have been derived such
as the linear minimum mean square error (LMMSE) estimator and the least
squares (LS) estimator. In order to get low-complexity equalizers in
the fast fading channel, the basis expansion model (BEM) with a low
number of parameters is employed to approximate the channel such as the
oversampled BEM, the Slepian BEM and the optimal BEM, where only the
BEM coefficients are estimated. For the oversampled and Slepian BEM,
the coefficients are estimated by the LS estimator. We propose a now
optimal BEM where the coefficients are estimated by the LMMSE approach.
The proposed model minimizes the mean squared LS estimator error for a
fixed number of parameters. This model is also robust when the Doppler
frequency is varied from the design value.  Simulation results show the
channel estimation based on the optimal BEM improves the BER
performance with a low computational cost. Use of a parameter model
allows the use of a Kalman filter to estimate the channel in the fast
fading. A dynamic model based on the evolution of the BEM parameters in
sequential time segments of the channel is presented for OFDM systems
in fast fading channels.

TIME/PLACE: Thursday November 8 at 15:45 in EOW 430
SPEAKER: Mohamed Yassein
TITLE: An Image Normalization Technique based on Geometric Properties of
  Image Feature Points
ABSTRACT:

Image normalization techniques have been used as a preprocessing step
in many applications such as pattern recognition and  classification,
image retrieval, and image watermarking. Image normalization is used in
such applications with the purpose of representing objects, patterns
(or the entire image) regardless of changes in their orientation, size,
or position.

This talk will introduce a technique for image normalization based on
the geometric properties of image feature points. The proposed
technique consists of two main steps, feature point extraction and
estimating the normalization parameters. The feature point extraction
is carried out using an extractor that based on scale-interaction of
Mexican-hat wavelets. The locations of the extracted feature points are
used to estimate the normalization parameters needed for normalizing
the image.  The proposed technique can be used to normalize input
images in a rotation, translation and scale invariant manner.
Experimental results illustrate the accuracy of normalization and the
robustness against several geometric distortions, image degradations
and common image-processing operations.

TIME/PLACE: Friday November 23 at 14:30 in EOW 430
SPEAKER: Eugene Hyun
TITLE: Location Aware Computing
ABSTRACT:

Location aware computing consists of the use of location information to
improve the value of a wireless network for the users.  Current
technologies will be discussed as well as our novel approach; To use
the power received signal strength (RSS) from Wireless Local Area
Networks (WLAN) base stations. We implement a non-parametric kernel
function to estimate the terminal's location. A presentation of our
experimental setup and results will be presented.

TIME/PLACE: Friday September 14 at 14:30 in EOW 430
SPEAKER: Parameswaran Ramachandran
TITLE: Recent Research Progress Related to the Location of Hot Spots in
  Proteins Using Digital Filters
ABSTRACT:
Accurate and efficient location of hot spots in proteins is of prime
importance in order to understand protein function. In an earlier
work, we proposed and demonstrated a hot-spot location technique using
a narrowband bandpass inverse-Chebyshev digital filter. In this talk,
we present a preview of a new optimization-based filter design
technique that can be used to design better digital filters capable of
yielding more accurate hot-spot locations. We demonstrate the
technique using several example protein sequences.

TIME/PLACE: Friday March 16 at 14:30 in EOW 430
SPEAKER: Deepali Arora
TITLE: Overview of Wireless Communications and Standards
ABSTRACT:

Wireless communications is a rapidly growing segment of the communications
industry, with the potential to provide high-speed high-quality information
exchange between portable devices located anywhere in the world. Potential
applications enabled by this technology include multimedia Internet-enabled
cell phones, video teleconferencing and distance learning, to name just a
few. However, supporting these applications using wireless techniques poses
significant technical challenges.

This talk presents a brief overview of existing and emerging wireless
technologies and standards. Some of the topics that will be discussed
include different multiple access techniques, channel fading and techniques
to combat the effects of fading and interference on system performance. A
brief overview of cross-layer systems will also be discussed. Finally, some
of the applications of TDMA and CDMA systems in current wireless systems
will also be discussed in brief.

TIME/PLACE: Thursday March 29 at 15:45 in EOW 430
SPEAKER: Mathieu Lagrange
TITLE: Sound Source Tracking and Formation using the Normalized Cuts
ABSTRACT:

The goal of computational auditory scene analysis (CASA) is to create
computer systems that can take as input a mixture of sounds and form
packages of acoustic evidence such that each package most likely has
arisen from a single sound source. We formulate sound source tracking
and formation as a graph partitioning problem and solve it using the
normalized cut which is a global criterion for segmenting graphs that
has been used in Computer Vision.  We describe how this formulation can
be used with sinusoidal modeling, a common technique for sound
analysis, manipulation and synthesis.  Several examples showing the
potential of this approach and its scalability to the analysis of
multimodal (Audio+Video) systems will be provided.

TIME/PLACE: Friday April 13 at 14:30 in EOW 430
SPEAKER: Di Xu
TITLE: An Improved Multiscale Normal-Mesh-Based Image Coder
ABSTRACT:

Three modifications to the normal-mesh-based image coder of Jansen,
Baraniuk, and Lavu are proposed, namely, the use of a data-dependent
base mesh, an alternative representation for normal/vertical offsets,
and a different scan-conversion scheme based on bicubic
interpolation. Experimental results show that our proposed changes
lead to improved coding performance in terms of both objective and
subjective image quality measures.

TIME/PLACE: Thursday April 26 at 15:45 in EOW 430
SPEAKER: Michael Adams
TITLE: Triangulations and Signal Approximation
ABSTRACT:

Although regular (i.e., lattice-based) sampling is ubiquitous in signal
processing, such sampling is usually highly suboptimal.  Due to the
nonstationary behavior of signals, irregular sampling often facilitates
better signal approximations.  In this talk, the speaker will introduce
triangulations and explain how they can be employed in this context
for representing signals defined on planar domains (such as images).
The application of image approximation will be considered.

TIME/PLACE: Thursday May 10 at 15:45 in EOW 430
SPEAKER: Pan Agathoklis
TITLE: A Feature-based Algorithm for Image Registration
ABSTRACT:

In this talk an algorithm for aerial image registration will be
presented. The objective of this algorithm is to register aerial images
having only partial overlap, which are also geometrically distorted due
to the different sensing conditions and in addition they may be
contaminated with noise, may be blurred, etc. The geometric distortions
considered in the registration process are rotation, translation and
scaling. The proposed algorithm consists of three main steps: feature
point extraction using a feature point extractor based on
scale-interaction of Mexican-hat wavelets, obtaining the correspondence
between the feature points of the firstt (reference) and the second
image based on Zernike moments of neighborhoods centered on the feature
points, and estimating the transformation parameters between the first
and the second images using an iterative weighted least squares
algorithm. Experimental results illustrate the accuracy of image
registration for images with partial overlap in the presence of
additional image distortions, such as noise contamination and image
blurring.

TIME/PLACE: Friday June 8 at 14:30 in EOW 430
SPEAKER: Ze Dong
TITLE: Photography as a Hobby
ABSTRACT:
This presentation is not about research, but introduces photography as
a hobby. The talk gives a brief overview of development of photography,
and then focuses on the digital photography. Basic terminology and
techniques in photography will be explained. The criterions judging the
quality of pictures are discussed by way of sample images.

TIME/PLACE: Friday June 22 at 14:30 in EOW 430
SPEAKER: Wu-Sheng Lu
TITLE: Variational Methods for Image Processing: Part I
ABSTRACT:
In the conventional numerical optimization setting for an engineering
analysis or design problem, the number of parameters to be optimized
must be finite.  On the other hand, images are two-variable functions
and important image features such as edges and object contours are
implicitly defined one-variable functions, and processing of images and
image sequences such as de-noising, de-blurring, segmentation, etc.
involve selection of best candidate images with noise and blurring
effects removed and best contours highlighting objects of interest.
Naturally, the "design parameters" in an effective optimization
formulation for such image processing tasks are functions rather than
finite-dimensional vectors. The mathematical foundation of such
optimization formulation and solution methods is the calculus of
variations (CoV), and the optimization techniques based on CoV are
generally referred to as variational methods.  In this talk, we give a
brief overview of several successful variational methods developed in
the 1990s and some recent research activities in the field.

TIME/PLACE: Friday July 20 at 14:30 in EOW 430
SPEAKER: Wu-Sheng Lu
TITLE: Variational Methods for Image Processing: Part II
ABSTRACT: [This is a continuation of the speaker's earlier recent talk.]
In Part II of the talk we shall cover two topics. We start with a quick
review of the anisotropic diffusion method by Perona and Malik (PM) and
present one of the many developments following the PM model, which was
proposed by Catte, Lions, Morel and Coll in 1992 and has since
attracted considerable attention.  Next, we describe an active contour
model proposed by Chan and Vese in 2001 for segmentation. The Chan-Vese
model is based on the level set method originated from the work of
Osher and Sethian on analyzing and computing moving fronts and
minimization of the Mumford-Shah functional. We use examples to
illustrate several key theoretical elements of the model and
demonstrate the effectiveness of the model for image segmentation.

TIME/PLACE: Friday August 3 at 14:30 in EOW 430
SPEAKER: Peter Hampton
TITLE: Phase Correction Distribution Methods for a Woofer-Tweeter System of
Deformable Mirrors
ABSTRACT:
In this talk I will introduce the Woofer-Tweeter test bench. A
deformable mirror is an array of actuators that can bend a reflective
surface. Similarly to audio aplications, the name Woofer refers to a
low spatial and temporal frequency deformable mirror and the tweeter is
a high spatial and temporal frequency deformable mirror. This system
was used to prove a concept of using these two  mirrors co-operatively
to correct for distortions of star light. Fundamental concepts to
classic adaptive optics will be presented.

TIME/PLACE: Friday August 17 at 14:30 in EOW 430
[There are two presentations, each of which will take about 30 minutes.]
*** PRESENTATION 1 ***
SPEAKER: Arjuna Madanayake
TITLE: FPGA Processors for Real-time 2D/3D IIR Frequency-planar/Beam Filters
ABSTRACT:
Broadband antenna arrays are finding increasing beam forming
applications for ultra wideband (UWB) communications, radar imaging,
radio astronomy, navigation and space research.  Beam forming for
antenna arrays is possible using various DSP algorithms based on finite
impulse response (FIR) and infinite impulse response (IIR) filters.  In
this talk, we briefly review MDSP approaches to beam forming using
first-order spatio-temporal 2D/3D IIR frequency-planar/beam digital
plane-wave filters, and discuss our measurable progress towards
single-chip real-time MDSP hardware implementations for eventual use
with broadband smart antenna arrays. We discuss the design of novel
distributed-parallel-processor (DPP) massively-parallel systolic-array
hardware architectures that can potentially lead to 2D/3D real-time
VLSI circuits for RF filtering at frame-rates of several GHz. We also
report on intermediate frequency (<100MHz frame rate) field
programmable gate array (FPGA) based prototypes of both 2D and 3D IIR
DPP plane-wave filter circuits that we built and tested at the
University of Calgary.
*** PRESENTATION 2 ***
SPEAKER: Thushara Gunaratne
TITLE: Beamforming of Temporally-Broadband-Bandpass Plane Waves using 2D
  FIR Trapezoidal Filters
ABSTRACT:
Beamforming is an important array processing application that has been
successfully used to selectively enhance desired spatio-temporal (ST)
plane waves (PWs) according to their directions of arrival (DOAs) to
recover the associated 1D temporal intensity functions, alleviating
both co-channel interference, caused by other propagating ST signals,
and non-propagating broadband noise. In this lecture, we present two
new discrete-domain methods proposed for the beamforming of temporally
broadband bandpass ST PWs using 2D FIR filters with novel
trapezoidal-shaped passbands. The first method employs a
complex-coefficient 2D FIR filter having an asymmetric
trapezoidal-shaped passband at baseband and achieves the minimum
possible temporal sampling rate for a given temporal bandwidth. The
second method employs a real-coefficient polyphase 2D FIR filter having
a rectangularly-symmetric double-trapezoidal-shaped passband and
requires 50% less computational complexity for the same temporal
bandwidth. Experimental results have confirmed that both methods are
equally capable of enhancing the desired temporally broadband bandpass
ST PWs according to their DOAs under severe co-channel interference and
multipath fading.

TIME/PLACE: Thursday August 30 at 15:45 in EOW 430
SPEAKER: Michael McGuire
TITLE: An Analysis of Faster-Than-Nyquist Signalling
ABSTRACT:
In 1975, J.E. Mazo demonstrated that it is possible to increase the
data rate of a digital communications system operating using 4-QPSK
modulation over an additive white Gaussian noise (AWGN) channel by
increasing the rate of the data pulses above the standard Nyquist
criteria limit without  increasing the bit error rate or required
transmission bandwidth.  This so-called Faster-than-Nyquist (FTN)
signalling has not been used in real communications systems since the
previously proposed implementation schemes require large amounts of
computation in the receivers. This paper introduces a reformulation of
FTN signalling in terms of a non-rectangular matrix multiplied by a
sample vector of modulated 4-QPSK symbols.  With this formulation, the
receiver computing cost to detect the transmitted data for an AWGN
channel is well within the complexity bounds for standard digital
communication systems.  This formulation enables an analysis of FTN
signalling so that it may be directly compared to standard higher order
modulation and data coding techniques.  This presentation compares the
attributes of FTN signalling in terms of required bandwidth, mean
power, peak to average power ratio, and implementation complexity to
other signalling schemes. Bounds on the performance gains for more
general and higher dimensional FTN signalling schemes are also
presented.

TIME/PLACE: Thursday September 27 at 15:45 in EOW 430
SPEAKER: Ana Maria Sevcenco
TITLE: Edge detection techniques in image processing
ABSTRACT:
Edge detection is the process which identifies the sharp changes of
magnitude intensity in an image. A less strict definition is that edge
detection identifies the "meaningful" discontinuities in intensity
values of an image. Either way, the edge detection process can be
regarded as one of the basic operations in image segmentation, image
analysis, image recognition, image enhancement, image compression, or
other image processing fields.  There are many ways to perform edge
detection such as first or second order derivative methods, wavelet
techniques, or more recent algorithms as active contours. In this talk
we briefly present the Canny edge detector, as a point-based oriented
method using the gradient information of the image, the multi-scale
edge detection using the wavelet transform, and the active contours
algorithm, as a function-based oriented method in which an evolving
curve, subject to constraints from a given image, detects the objects
in that image.

TIME/PLACE: Friday October 12 at 14:30 in EOW 430
SPEAKER: Yihai Zhang
TITLE: Low Complexity Joint Semiblind Detection for OFDM Systems over
Time-Varying Channels
ABSTRACT:
Orthogonal frequency division multiplexing (OFDM) modulation is widely
used in communication systems to meet the demand for ever increasing
data rates. However, it is sensitive to Doppler spread caused by user
mobility, which results in intercarrier interference (ICI).  In the
design of most ICI reduction algorithms for OFDM systems, perfect
channel information is assumed at the receiver.  Many channel
estimation methods have been proposed for OFDM systems based on pilot
symbols.  In most cases, channel estimation with pilot symbols is a
robust method.  However, there are disadvantages such as bandwidth loss
and overhead, which can be excessive in fast fading channels.  This
motivates the use of joint blind or semiblind detection methods for
OFDM systems.  In this talk, a low complexity joint semiblind detection
algorithm for OFDM systems over time-varying channels is presented
based on the channel correlation and noise variance.  Results are given
which demonstrate that the proposed algorithm provides comparable
performance with lower computational complexity than that of a sphere
decoder.

TIME/PLACE: Thursday October 25 at 15:45 in EOW 430
SPEAKER: Deepali Arora
TITLE: The hidden beam problem in wireless communication systems 
ABSTRACT:

The use of IEEE 802.11 wireless LAN systems (WLANs) over the last few
years has increased tremendously and WLANs are no longer limited to
homes and offices. The advancements in antenna technology and there use
in WLANs has shown that the directional antennas are preferable over
omni-directional antennas not only in terms of increased range, better
directivity and substantial reduction in interference but also in large
power savings which is a critical issue in wireless systems. To exploit
the benefits offered by the directional antennas, several researchers
have re-designed multiple access control (MAC) protocols by replacing
omni-directional antennas with directional antennas. However, when used
in IEEE 802.11 systems, directional antennas present their own share of
issues such as the hidden node problem, the hidden beam problem and
deafness. While some of these issues such as the hidden node problem
and deafness have been widely investigated, the hidden beam problem has
remained largely unaddressed.

In this presentation an overview of the hidden beam problem and its
effect on systems relying on carrier sensing is presented. Some of the
earlier work done to solve the hidden beam problem will also be
discussed in brief and a new technique and some initial results will be
presented.

Other Related Links