Automated detection of Baleen Whales off Nova Scotia

PROJECT

Automated classification of baleen whales in the North Atlantic Ocean.

CLIENT

Department of Fisheries and Oceans Canada

PROJECT OVERVIEW

The Department of Fisheries and Oceans Canada (DFO) has been undertaking Passive Acoustic Monitoring in the North Atlantic Ocean. They wanted to detect and classify all baleen whales off Nova Scotia, but did not have access to multi-species detectors and classifiers. 

OUR SUpport

DFO engaged our bioacoustics expertise to train a specific classifier that could detect and classify calls from any species of baleen whales and allow the results to be easily verified themselves.

WE PROVIDED:

Data Analysis

DFO provided us with several months of raw acoustic data that was uploaded to our analysis systems where automated detectors processed the data. 

All baleen species were to be detected so species classification was achieved using four interconnected detection algorithms that used energy fluxes, spectral entropies, contour tracing and edge detection and harmonics. Cross-correlation between outputs and predetermined templates based on our bioacoustics library and literature lead to species classification. 

These detectors produced 2-min audio snippets of each detection as well as an image file of the spectrogram containing the detection. This was a standard data product provided to clients to give them confidence in the results because they can self-verify and check the outputs. 

Various calls from blue whales, fin whales, sei whales, humpbacks and North Atlantic right whales were detected, with bioacoustics properties and datetimes of calls provided to DFO.