Passive acoustic monitoring is an essential technique for studying marine mammal populations, especially in remote or visually obstructed marine environments. Wide-baseline hydrophone arrays can cover large areas of the ocean and enable spatial localization. While these configurations offer much insight, they present unique signal processing challenges due to complex propagation effects and multipath arrivals. In this work, we introduce a multi-sensor tracking algorithm that employs a probabilistic framework based on factor graphs to leverage multipath propagation for localizing and tracking vocalizing sperm whales. Our approach incorporates the time-differences-of-arrival of direct and surface-reflected paths received by a single receiver and pairs of receivers. The likelihood function within the factor graph integrates the acoustic ray-tracing engine Bellhop that accounts for variable sound speed profiles to model multipath arrival times. The proposed method inherently addresses data association ambiguities and measurement uncertainties, efficiently combining information from multiple hydrophones and hydrophone pairs to probabilistically estimate whale trajectories. We demonstrate the performance of the proposed method using synthetic data simulating whale vocalizations and field-measured acoustic data of single or multiple sperm whales. The results show that localization accuracy and tracking continuity improve when multipath propagation is used.
Watkins et al. (Wed,) studied this question.