Assessing the Performance of Omni-Directional Receivers for Passive Acoustic Detection of Vocalizing Odontocetes: Initial Analysis

摘要

The purpose of this study was to evaluate the performance of inexpensive, passive, omni-directional receivers as a means to detect vocalizing Odontocetes using conditional statistics. To evaluate and predict performance, it was necessary to establish probability of detection as a function of (a) signal to noise ratio or range at a given source level and (b) probability of false alarm. For this purpose, a model of the probability distribution function of the detector output was derived from experimental data, For the experiment a series of short duration digital recordings of selected odontocete vocalizations were broadcast underwater from a moving platform. The vocalizations were monitored and digitally recorded at a stationary underwater array consisting of three vertically distributed hydrophones. Over a period of three days, several hundred iterations of each signal with the transmitter at ranges varying from 300 meters to 12000 meters were recorded. A monitoring hydrophone (co-located with the transmitter) was used to monitor the signal source level. The raw data was fed to two 'automatic detectors' consisting of different data processing routines developed in MATLAB. The output of each detector was subjected to statistical analysis. Other factors also considered in the analysis were: signal used, range, and wind (as a proxy indicator of noise generated by surface wave action). A statistical test was employed to systematically find a best fit probability distribution function model of detector output. From this empirical model, detector performance was estimated.