Phase 1: DCL System Research Using Advanced Approaches for Land-based or Ship-based Real-Time Recognition and Localization of Marine Mammals - HPC System Implementation

TL;DR

This research develops advanced algorithms and high-performance systems for real-time detection, classification, and localization of marine mammals using long-term acoustic datasets, aiming to automate recognition and reduce manual annotation.

Contribution

The paper introduces novel deep learning and HPC-based methods for automatic marine mammal detection and classification in real-time from passive acoustic data.

Findings

Effective detection and classification of marine mammals achieved

Real-time processing capabilities demonstrated

Automated measurement of species-specific sounds implemented

Abstract

We aim to investigate advancing the state of the art of detection, classification and localization (DCL) in the field of bioacoustics. The two primary goals are to develop transferable technologies for detection and classification in: (1) the area of advanced algorithms, such as deep learning and other methods; and (2) advanced systems, capable of real-time and archival and processing. This project will focus on long-term, continuous datasets to provide automatic recognition, minimizing human time to annotate the signals. Effort will begin by focusing on several years of multi-channel acoustic data collected in the Stellwagen Bank National Marine Sanctuary (SBNMS) between 2006 and 2010. Our efforts will incorporate existing technologies in the bioacoustics signal processing community, advanced high performance computing (HPC) systems, and new approaches aimed at automatically detecting-classifying and measuring features for species-specific marine mammal sounds within passive acoustic data.