Relative impact of signal-to-noise ratio and propagation effects on the performance of an aural classifier

Abstract

Passive acoustic monitoring (PAM) is used to study marine mammals in their habitats, which cover diverse underwater environments. The distinct propagation characteristics of different ocean environments alters the time-frequency characteristics of a recorded signal. This may affect the accuracy of PAM systems. To develop a PAM system capable of operating under numerous environmental conditions, one must account for the impact of propagation. An aural classifier developed at Defence R&D Canada (DRDC) has successfully been used for inter-species discrimination of cetaceans. The aural classifier achieves accurate results by using perceptual signal features that model the features employed by the human auditory system. The current work examines the relative impacts of signal-to-noise ratio (SNR) and propagation effects on the performance of the aural classifier. DRDC’s pulse propagation model, Waveform Transmission Through a Channel (WATTCH), was used to simulate signals travelling through the ocean environment over ranges of 0–20 km. Noise was added to both these signals and the original signals, so that performance could be compared for three scenarios expected to decrease classifier performance: decreasing SNR, increasing propagation effects (frequency spreading, multipath, etc.), and combined SNR and propagation effects. In this presentation, the modeled results are compared to experimental data.