Time-frequency signatures based on fuzzy-clusters: Applications to echoes from absorbing spherical shells

Abstract

In order to develop an automatic target identification system for underwater objects, it is necessary to extract and classify the features contained in waves scattered from the targets. According to the acoustic scattering theory, the scattered signal contains the “signatures” corresponding to the structure and material content of the target. Pseudo-Wigner Distribution (PWD) function is applied on scattered waves, and the signatures are found and analyzed in the time-frequency plane. A method based on Fuzzy C-Means (FCM) algorithm is then introduced to eliminate the redundant features and represent the useful information (i.e. the time-frequency signatures) by FCM cluster centers. As a result, the amount of data that is required to identify the target is greatly reduced, and a target classification scheme can be simply developed based on the cluster representation.