Real time automatic detection of Orcinus orca vocalizations in a controlled environment

The use of passive acoustic observation is a useful tool in bioacoustic studies of cetaceans, such as killer whales (Orcinus orca). Such studies require the recording, detection, and classification of vocalizations of animals during long periods of time. The manual processing of the recordings is an extremely time consuming task because of the large amount of data to process. Automatic detection and classification techniques are useful to improve the processing, increase the amount of information and, as a consequence, provide information for the conservation of these species. The Orca Ocean facilities of Loro Parque in the Canary Islands, Spain, were used as an experimental platform for developing devices to perform bioacoustic studies. Detection methods with low computational complexity were tested in order to capture vocalizations of four O. orca specimens in real time. The algorithms were also tested in other scenarios in order to determine their global performance. The sensitivity to noise was also tested. The most accurate method in this study was implemented and integrated with a continuous recording system generating an event database in real time, downsizing the storage demand to 7%. This allowed the storage of all the sound events produced in about one month on a standard computer harddisk and the generation of basic statistics on vocal activity of the animals.