Distributed acoustic sensing (DAS) is transforming acoustic monitoring, using existing fiber optic infrastructure to create vast listening arrays. DAS provides uniformly spaced virtual channels at intervals of a few meters over hundreds of km, with applications from geophysics and weather to vessel and marine mammal monitoring. A problem arises in that the large number of channels generates immense data volumes that overwhelm manual analysis. Hence, the need for automated approaches, especially for infrequent signals of interest. We develop and evaluate an automated algorithm for generic transient acoustic signal detection in DAS data. Our algorithm exploits the extensive available spatial aperture with multiple channels in the frequency–wavenumber domain to estimate signals against background noise over consecutive time windows. Signal energy estimation metrics (average, maximum, weighted average, and 95th percentile) are compared to a time-varying background noise estimate to derive a custom signal-to-noise ratio used for detection. We evaluate our algorithm performance on whale and ship signals using several months of data from a 130-km cable off Svalbard, taken in 2020 and 2023. The algorithm provides robust support for the creation of labeled DAS datasets that can be used in further processing for automated marine mammal and vessel monitoring at scale.
Schulze et al. (Tue,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: