Wavelet analysis of bioacoustic scattering and marine mammal vocalizations

Abstract

Wavelets have been used in numerous geophysical studies but few have examined their applicability to underwater acoustic signals. Wavelet transforms can remove noise from a given time series and allow data analysis at multiple levels of resolution. This unique ability is exercised as a feasible application to the signals in this thesis: a reflected scattered signal from a swimbladder-bearing fish, alewife (Alosa pseudoharengus), and several Odontocetes vocalizations. Both studies reveal that wavelet-based techniques show potential in providing viable information for these acoustic signals despite the lack of statistical analysis. The alewife portion shows a reasonable first order approximation to the absolute target strength and to the time delay correlation caused by the spatial separation of scattering features in the fish. The marine mammal application shows a possible real time method to estimate the mammal's range using the root mean square (RMS) energy of the decomposed signal. Because of wavelet function mismatch, both studies conclude that more extensive research is necessary to develop these techniques into systematic processes.