Target recognition in acoustics

Abstract

The classification/identification of underwater stationary objects can be acoustically achieved by means of passive or active sonar. Moving targets are the “easy case,” since they can be identified by their Doppler factor, and will not be discussed here. By means of passive sonar, one merely listens with the best possible receiver to the noises/sounds emitted by the noisy target. The targets are passively “identified” when the spectra of the radiated sound fields they emit are seen to match certain spectral lines known a priori for certain desired objects. Identification via active sonar is independent of how noisy the target may be. Here, the echoes returned by the targets when they are insonified by sonar pings are analyzed. These are classical inverse scattering situations in which a returned “signature” in the time or the frequency domain is used to extract physical characteristics of the targets in question. This paper will concentrate on this type of (active) classification technique [i.e., G. Gaunaurd, IEEE J. Ocean. Engr. OE-12, 419–422 (1987)], and discuss means to accurately determine size, shape, and composition information about submerged elastic targets from the RESONANCE features present in their monostatic and/or bistatic scattering cross sections. Emphasis will be placed on the resonance scattering techniques (RST) developed for poles in the complex-frequency domain, and their Fourier-related, time-dependent waveforms. The approach is ideally amenable to implementation by expert (AI) systems.