Research Areas of the ASPC Group

Our goal is to develop high performance, robust and efficient signal processing algorithms for the advancement of modern communications. Our major research areas of interest are: Short descriptions of our research on these topics are provided below. (You can follow the appropriate link above.) Links to more detailed descriptions are (or will be) provided there.

The research support of a number of companies and organizations, including has been and continues to be mutually beneficial. (More detailed acknowledgments of support will be provided on the research area pages, when they are finished.)

Short descriptions of research areas

with links to longer descriptions where available.
High Resolution Array Signal Processing:
Current direction finding algorithms usually perform well when the source is well localized. However, in some applications (e.g., scatter propagation communications and low-angle tracking) the source appears to be dispersed and the performance of the current algorithms degrades. We are developing algorithms which do not require a localized source assumption and consequently provide improved performance when the source is, or appears to be, dispersed. In other applications (e.g., cellular communications) transmitted signals undergo multi-path propagation and consequently appear to be correlated sources, degrading the performance of current algorithms. We have designed an efficient spatial smoothing method to approximate the data covariance matrix by a positive semi-definite Toeplitz matrix to counter this degradation. Whilst the initial simulation results are very encouraging, further testing is underway.

Beamforming is emerging as an efficient technique for enhancing the capacity of cellular communication systems. However, the computational load of the chosen beamforming algorithm must be light enough for the rapid movement of a mobile unit. We have proposed a class of such `fast beamforming' algorithms based on the cyclostationary properties of most communication signals (the `CAB' algorithms), and are currently modifying their structures in order to obtain greater robustness.

List of major research areas

Multiple Signal Transmission:
The increasing demand for high-speed flexible digital communication services has focussed our attention of the development of high-capacity robust multiplexing schemes. We have proposed two schemes: Wavelet Packet Division Multiplexing (WPDM) and Cyclic Frequency Division Multiplexing (CFDM).

WPDM:
A wavelet packet basis provides a set of orthonormal waveforms which overlap in both time and frequency. We are investigating the application of such waveforms in modulation and multiplexing, for improved bandwidth efficiency over conventional frequency division and time division multiplexing schemes. Such a 'wavelet packet division multiplexing' scheme provides a highly flexible structure and provides substantial robustness to common adverse channel environments. Furthermore, by exploiting the close relationships between wavelets and multi-rate filter banks we can obtain simplified transmitter and receiver structures. (Jiangfeng Wu's thesis on WPDM is avaliable online.)
CFDM:
The ability of a frequency-shift (FRESH) filter to separate spectrally overlapping signals with distinct cyclic frequencies exposes the potential for transmission of overlapped signals for spectrally efficient communication. We are currently identifying trade-offs in the system structure to balance capacity increases against receiver complexity and the fidelity of the demultiplexed signals.

List of major research areas

Applications of Column Generation Algorithms:
Interior point column generation algorithms are a class of efficient optimization algorithms for finding a point in a convex set. These algorithms examine one constraint at a time and update the iterate accordingly. We are developing stochastic versions of column generation algorithms for use in adaptive filtering applications (system identification, channel equalization, speech coding, model predictive control, etc).

List of major research areas

Applications of Multi-Rate Filtering and Wavelets:
In addition to the application of multi-rate filter banks and wavelets to multiplexing (above) we are also investigating the application of wavelet packet and other subband decompositions to the reduction of the computational complexity of adaptive filtering algorithms (in particular for echo cancellation), as part of a data compression tool for target tracking via multi-sensor Kalman filtering, and as part of speech and audio compression schemes.

List of major research areas

Data Association and Multi-Target Tracking:
We are developing a software system for real-time multi-target tracking. A major problem that has to be resolved is how to associate data to the tracks. We have used the state-of-the-art homogeneous self-dual linear programming algorithm, as well as the $\epsilon$-relaxation algorithm to solve the underlying assignment problem. The intermediate results are combined to update the tracks. These efficient new data association methods not only deliver real-time performance, but also offer improved tracking accuracies.

List of major research areas

Distributed Detection in Multi-Sensor Networks:
We are developing methods to optimize the system performance of a multi-sensor network for detection of a target under correlated noise. The issues being studied include the selection of optimal local decision rules and optimal fusion rule at the fusion center. These results form a natural extension of the classical Bayesian detection theory which deals with one sensor case.

List of major research areas

Radar Pulse Classification:
We are also working on ways to (dynamically) classify the incoming radar pulses according to the (unknown) emitters. Due to the nature of the application, the carrier frequency and the inter-pulse information are not usable. We use clustering analysis and the MDL principle to solve this problem. A wavelet compression technique is also being considered in order to reduce the computational complexity of the clustering step.

List of major research areas

Back to the Advanced Signal Processing for Communications Group home page.
Tim Davidson (davidson@mail.ece.mcmaster.ca), with additional material from Tom Luo.
Last change: 24 June 1997.