Human performance in aural classification of sonar echoes

Abstract

Active sonar echoes from man-made objects and those from naturally occurring features in certain coastal environments can be difficult to distinguish when using traditional sonar processing techniques and visual displays. An approach being investigated at Defence R&D Canada-Atlantic for addressing this challenge is to exploit the capability of human hearing for discriminating between these two classes of echoes. Part of the work consisted in producing a human-performance baseline. Two human listening tests were designed and carried out. Both used stimuli that were generated from a sample of echoes recorded during actual sonar experiments at sea. In the second test, a 500-Hz high-pass filter was applied to the stimuli. Quantitative data from the rating-exercise portion of the tests were used to produce receiver-operating-characteristic (ROC) curves for modeling how well the participants could distinguish the two classes of echoes. For both tests, the results show that the listeners could hear differences, but performance was significantly better in the first test. Qualitative data collected from the questionnaire portion of the tests helped to interpret some of the performance results.