Classifying sonar signals with varying signal-to-noise ratio and bandwidth

Abstract

An automatic aural classifier developed at Defence Research and Development Canada has demonstrated the ability to distinguish target echoes from clutter using perceptual-based features inspired by sonar operators. Initially, the classifier was tested with echoes from explosive sources, but more recent research involved transmitting broadband waveforms from transducer sources. In sonar transducer operation, there is a trade off between source level and bandwidth, and the goal of this paper is to study how these factors affect echo classification. Source level relates to signal-to-noise ratio (SNR), which inherently affects classification since signals with low enough SNR cannot be distinguished from noise, let alone other signals. The dependence of classification performance on bandwidth is less obvious; however, the aural classification technique is based on a sub-band type of processing that mimics the basilar membrane in the human auditory system, and this model is not well adapted for narrow bands. Performance of the aural classifier is therefore expected to degrade as bandwidth is decreased. In this paper, the effect of SNR and signal bandwidth on echo classification is examined using echoes of varying SNR, and in various bands selected using band-pass filters. © 2013 Acoustical Society of America.