Many of the Physical phenomena encountered in practice are stochastic in nature, sharing two common characteristics:

• Nonstationarity
• Unknown statistics

To deal with these physical realities, we naturally turn to the use of adaptive signal processing.

In these few lines, I have mentioned some key words: physics, stochasticity, and adaptivity. To illustrate that these issues are intimately related, I have chosen "Adaptive Radar" as the topic for my lecture, with the ocean as the physical environment of interest.

After briefly reviewing some important aspects of ocean dynamics and how they influence sea clutter (i.e., radar backscatter form the ocean surface) and recognizing the importance of working with real-life data, I will describe an experimental research facility known as the "IPIX Radar" built at McMaster University. The IPIX radar is a sophisticated, instrument-quality, multi-function and transportable radar, which is fully coherent and polarimetric. It has been used to collect an extensive and highly invaluable radar database under differentenvironmental conditions at two sites: East coast of Canada, and Lake Ontario. Views of the many aspects of this database can be seen on the web site

http://soma.mcmaster.ca/ipix.

Sea clutter has been traditionally described as a stochastic process, with the compound K-distibution being the most principled distribution in both physical and statistical terms, which goes back to the 1970s. During the 1990s, influenced by chaos theory, experimental results were presented showing that sea clutter may be chaotic. In the second part of my lecture, I will show that presently available algorithms for estimating chaotic invariants, namely, correlation dimension and Lyapunov spectrum, are incapable of distinguishing between a chaotic process and a stochastic process in a reliable manner. I will also present results that show that dynamic reconstruction of sea clutter is very difficult. These results lead us to conclude that "Chaos can be a self-fulfilling prophesy" not only in the context of sea clutter but also other physical phenomena. Nevertheless, I will present experimental results that demonstrate that sea clutter is a nonlinear dynamical process with a discernible structure that manifests itself in the hybrid form of amplitude modulation and frequency modulation.

In the third part of my lecture, I will address the task of how to reliably detect a small target (e.g., small piece of ice or fishing boat) in the presence of sea clutter. Here I will describe two approaches for adaptive target detection:

• Adaptive pattern classification, using time-frquency images based on the Wigner-Ville' distribution
• Adaptive state estimation, involving the joint use of two filters: A Kalman filter for tracking the evolution of the sea clutter, measured by its short-time Fourier transform; the second filter for tracking the dynamics of the target, if present.

A distinctive feature of radar is the co-location of the transmitter and the receiver, which means that (unlike a communication system) the transmitter has ready access to the received signal. This observation sets the stage for the fourth and final part of my lecture:

"How to design a procedure for adaptation of the transmitted radar waveform for the optimum detection of a dynamic target in a sea clutter dominated environment."

In the context of a multifunction radar, the challenge is to identify the radar discriminants that are most sensitive to variations in target dynamics and the development of a criterion that facilitates the adaptation of the transmitted radar waveform for optimum target detection in a real-time fashion. With adaptive radar in mind, I will finish the lecture with a description of the neurobiological echolocation system of the bat, which carries out complex neural computations within a brain the size of a plum. Yet it can perform and capture its target (i.e., a flying insect) with a facility and success rate that would be the envy of a radar and sonar engineer.

Simon Haykin - received the degrees of B.Sc. (First Class Honours), Ph.D., and D.Sc., all in electrical engineering from the University of Birmingham, England. On the completion of his Ph.D. studies, he spent several years from 1956 to 1965 in industry and academe in England. In January 1966, he joined McMaster University, Hamilton, Ontario, Canada, as Full Professor of Electrical Engineering; he has stayed there every since.

In 1972, in collaboration with several faculty members, he established the Communications Research Laboratory (CRL), specializing in signal processing applied to radar and communications. He stayed on as the CRL Director until 1993. In 1996, the Senate of McMaster University established the new title of University Professor; in April of that year, he was appointed the first University Professor from the Faculty of Engineering. Professor Haykin is a Fellow of the IEEE and a Fellow of the Royal Society of Canada. In 1999 he was awarded the honorary degree of Doctor of Technical Sciences by ETH, Zurich, Switzerland.

Professor Haykin’s primary research interests have focused on adaptive signal processing applied to radar and communications, on which he has published many papers. He is the author, co-author, editor of over 40 books, which include the widely used text books: Communications Systems (4th edition, Wiley), Adaptive Filter Theory (4th edition, Prentice-Hall), and Neural Networks: A Comprehensive Foundation (2nd edition, Prentice-Hall); these three books have been translated into many different languages all over the world. He is the Founding Technical Editor of the Wiley Series on "Adaptive and Learning Systems for Signal Processing, Communications, and Control".