Broadband active sonar classification

Active sonar performance is limited by noise, reverberation or false alarm rate (FAR). High FAR is overcome through the use of signal classification, which depends on the information content of the signals. Information content and classification are addressed using a set of broadband sonar echoes. The set of broadband echoes generated by an explosive source scattered from 30 different geographic locations is described. A continuous wavelet transform is shown that simulates the audible content of the echoes. The root-mean-square log scalogram distance, an audio distance defined with wavelet scalograms, between echoes is shown to decrease with decreasing sonar-source bandwidth. Results from multidimensional scaling (MDS), additive tree analysis, and exit interviews from a previous human subject listening study are shown. The results suggest that subjects use a process other than one defined by a single feature space. A human-subjects-research experimental protocol is described that will elucidate an alternative approach to signal classification. Funding was provided by the Office of Naval Research of the United States Navy.