In many control/signal processing applications, whether it be detecting the presence of a hostile submarine or a mine buried in the surf zone or communicating with a person lost in the hostile ocean environment, the processing of noisy, highly uncertain data to extract the desired information (e.g. target, mine or person location) is of extremely high importance. When signals are deeply buried in noise, then processing techniques that utilize representations of the underlying phenomenology to extract the desired information from the measured data must be employed. In this lecture, we discuss the model-based approach to signal and image processing capturing the underlying physics, instrumentation and noise in the form of mathematical models from which the measured data evolved. We argue that this approach is much more of a modeler’s rather than signal processor’s tool, since we are working directly in the physics of the problem. Phenomenologists operating in their own physical environment (comfort zone) helps remove much of the “magic” associated with the processing of their noisy signals and images.
Once the basic framework is established, we show how this approach can be developed to solve a wealth of applications ranging from the simple idea of estimating the direction-of-arrival of a far-field sonar source using a towed array, to creating an image of both near-field targets for localization, to estimating the potential failure of a vibrating structure. We briefly discuss ocean acoustic applications including sound-speed inversion and the development of an environmentally adaptive model-based processor for target localization. Finally, we discuss the applicability of such techniques to a diverse set of problems including nondestructive evaluation and position estimation for a high power laser beam. One of the major objectives of this lecture is to demonstrate the applicability, versatility and robustness of the model-based approach to a wide variety of physics-based applications advocating the attitude to “not give up” when the signals seem hopelessly buried in noise and uncertainty.
Back to Control Seminars Page