IMPLEMENTATION AND EVALUATION OF DISPERSION-INVARIANT FEATURES FOR SIGNAL CLASSIFICATION by

IMPLEMENTATION AND EVALUATION OF DISPERSION-INVARIANT FEATURES FOR SIGNAL CLASSIFICATION Greg Okopal, M.S. University of Pittsburgh, 2007 When a sound wave interacts with an object, the acoustic energy may excite resonances in the object corresponding to its natural modes of vibration. The backscattered wave will then contain information which can be used to distinguish among different objects. As the wave propagates, it can be changed by the propagation channel, which complicates automatic classification of the echo. For example, in a dispersive channel, the duration of the wave increases with propagation distance. Our goal is to identify features of propagating waves that may be used for automatic classification. In this work, we implement and test a class of moment-like features that are invariant to specific propagation effects, in particular dispersion. Our tests of the classification utility of the “dispersion-invariant moments” (DIMS) are performed on numerical models of dispersive propagation and acoustic scattering from steel shells. We consider the case of real dispersion relations and in the conclusion discuss the implementation of complex dispersion and a future direction for research.