ShipsEar: An underwater vessel noise database

There is a manifest shortage of audio databases available to underwater acoustics researchers. With the aim of palliating this situation, ShipsEar, a database of underwater recordings of ship and boat sounds, has been made available to the research community at http://atlanttic.uvigo.es/underwaternoise/. The database is currently composed of 90 records representing sounds from 11 vessel types. It includes detailed information on technical aspects of the recordings and environmental and other conditions during acquisition. To demonstrate the usefulness of ShipsEar, a vessel classifier was developed, based on cepstral coefficients and Gaussian mixture models. It was tested on a subset of ShipsEar database in which the original 11 vessel types were merged into 4 vessel size classes. The system yielded an overall classification rate of 75.4%, and 100% accuracy in detecting vessel presence. ShipsEar is potentially useful for the development and testing of applications based on processing underwater vessel sound.