Effective Strategies for Automatic Analysis of Acoustic Signals in Long-Term Monitoring

Abstract

Hydrophones used in Passive Acoustic Monitoring generate vast amounts of data, with the storage requirements for raw signals dependent on the sampling frequency, which limits the range of frequencies that can be recorded. Since the installation of these observatories is costly, it is crucial to maximize the utility of high-sampling-rate recordings to expand the range of survey types. However, storing these large datasets for long-term trend analysis presents significant challenges. This paper proposes an approach that reduces the data storage requirements by up to 85% while preserving critical information about Power Spectral Density and Sound Pressure Level. The strategy involves generating these key metrics from spectrograms, enabling both short-term (micro) and long-term (macro) studies. A proposal for efficient data processing is presented, structured in three steps: the first focuses on generating key metrics to replace space-consuming raw signals, the second addresses the treatment of these metrics for long-term studies, and the third outlines the creation of event detectors from the processed metrics. A comprehensive overview of the essential features for analyzing acoustic signals is provided, along with considerations for the future design of marine observatories. The necessary calculations and processes are detailed, demonstrating the potential of these methods to address the current data storage and processing limitations in long-term acoustic monitoring.