Abstract:

A broadband active sonar system using dolphinlike echolocation signals and biologically inspired signal processing algorithms was developed to classify proud and buried targets in real time. A dolphin simulator transducer was attached to a bottom-crawling remotely operated vehicle and used to transmit a dolphin click (120-kHz center frequency, 39-kHz bandwidth, 50-(mu)s duration) through seawater. Reflected target echoes were received and transmitted via cable to a shore-based computer system where the echoes were digitized at 1 MHz and subsequently processed. Two time-frequency representations of the echoes, one based on the short-time Fourier transform and the other on the Morlet wavelet, were processed in a hierarchical neural network system to derive target classifications. Echolocation returns were collected from six objects (cast iron pot, stainless steel sphere, glass jar, concrete tile, and coral rock) that were placed either on the ocean bottom or buried by bottom sediment. Echoes were separated into three target categories: (1) cast iron pot; (2) stainless steel sphere; and (3) a miscellaneous category consisting of the remaining four objects. Following supervised neural network training, the system was able to correctly identify 74%, 97%, and 88% of the category 1, 2, and 3 test echoes.