Implementation of algorithms described in
Shyam Madhusudhana, Anita Murray, and Christine Erbe. (2020). "Automatic detectors for low-frequency vocalizations of Omura’s whales, Balaenoptera omurai: A performance comparison." The Journal of the Acoustical Society of America 147(4). DOI: 10.1121/10.000108
for the detection of Omura's whale vocalizations.
Of the three methods described in the article, implementations of the following two are provided here -
- Blob Detector (implemented in Python)
- Entropy Detector (implemented in Matlab)
Instructions for using them are provided below. If you use these (in entirety or as parts), we request that you please cite the aforementioned article.
The Python3 implementation provided in blob_detector/omura_detector.py requires the Python packages DetectSound (which contains the general-purpose blob detection functionality), librosa (for loading audio files) and matplotlib (used for displaying detection results). These can be installed from command prompt as -
$> pip3 install librosa matplotlib
$> pip3 install git+https://github.com/shyamblast/DetectSound.gitAfter downloading omura_detector.py on your local machine, the detector can then be run as
$> python3 omura_detector.py /path/to/audio_file.wavto display the detection results on screen, or as
$> python3 omura_detector.py /path/to/audio_file.wav /save/to/output_file.txtto save the detection results to a file in Raven Selection Table format.
The Matlab implementation of the entropy detector is provided as a collection of 3 files in entropy_detector. After downloading the entire folder, set directory paths appropriately on lines 33 & 34 in OmuraEntropy2XSTD.m and then execute the script within Matlab.