Acoustic Classification of Abyssopelagic Animals

The unique environment of the abyssal plains allows many simplifying assumptions, facilitating the acoustic classification of an animal into one of two groups. The most important assumptions are based on low population densities and available target strength histograms and swim rate histograms. The likelihood ratio is formed from this information and accepted signal processing theory. The likelihood function, a three-dimensional integral, is analytically simplified to one dimension and then solved numerically. A simulation based on this solution and measured data demonstrates that classification using the likelihood ratio approach is accurate, e.g., sensitivity 0.8. Although the measured data come from two abyssopelagic genera, the methods presented here are more generally applicable. Simulations based on hypothetical animal populations show that under certain conditions, a near perfect classification can be made, e.g., sensitivity and specificity greater than 0.969. © 1992 IEEE

Duke Scholars

Author Robert A. Malkin Biomedical Engineering