Datasets
Standard Dataset
The Dominica dataset
- Citation Author(s):
- Submitted by:
- Guy Gubnitsky
- Last updated:
- Wed, 03/20/2024 - 17:22
- DOI:
- 10.21227/078x-0h20
- Data Format:
- License:
- Categories:
- Keywords:
Abstract
The Dominica database is used to evaluate and compare the performance of sperm whale click detectors. The database consists of 3 hours of recordings of sperm whale echolocation clicks and 4 hours of sound recordings containing delphinid clicks and transients from different sound sources, but no sperm whale clicks. Data were collected in September 2023 approximately 2 km west off the coast of Dominica Island in the eastern Caribbean using a home-built recorder equipped with an M36 Geospectrum hydrophone with near-constant sensitivity in the frequency range of 2 kHz to 20 kHz and a sampling rate of 192 kHz. The recordings lasted 5 days, with the recorder deployed at a depth of 20 m and anchored to a free-drifting buoy on the surface. The water depth was about 1,600 m.
This database is divided into two folders: Signal parts: consist of recordings containing sperm whales' clicks. Noise parts: consist of continuous recordings containing delphinid clicks and transients from various noise sources, but no sperm whale clicks. Additionally, a mat file named "Annotations_Dominica" is included, which contains a cell array with 28,300 annotations of sperm whale echolocation clicks. The array is structured in two columns: one describes the names of the recordings and the other contains a vector of the arrival time of the annotated clicks (in seconds).
Comments
Practical data for a fair comparison of click detectors and classifiers