Model-based signal processing approach to animal bioacoustics: A brief overview

The model-based approach to signal processing is generally founded on the fundamental concept of incorporating any a-priori knowledge of the underlying phenomenology from which the signal evolved along with measurement instrumentation and uncertainty (noise, parameters, etc.) in the form of mathematical models that are embedded in the processor. In this way, the phenomenologist, experimenter, and signal processor combine all of their possible knowledge into a scheme enabling each to think within their own comfort zones while developing a powerful approach to extract the illusive information they desire. In this overview, we present the concepts required to develop mode-based processing schemes that can be used in a wide variety of animal bioacoustics applications ranging from signal estimation, tracking, identification, detection, and classification. We discuss the development of this approach incorporating acoustic applications that can be extrapolated to animal bioacoustics problems. We can express all of these techniques in terms of a model-based framework enabling the use of such powerful estimation techniques such as Kalman filters (Gaussian case) to Bayesian particle filters (non-Gaussian case) for solution to a wide variety of problems.