Abstract:
Social animals such as Steller sea lions Eumetopias jubatus often produce vocalisations that are very consistent and easy to discriminate, thus reducing potentially hazardous social errors in reunions between mothers and pups, between neighbouring territorial males, or other social relationships where individual recognition is beneficial. We have developed a system for accurately categorising calls to the appropriate individual, using parallel processing software algorithms (neural networks). An ability to quickly and easily identify individuals by voice will greatly aid researchers monitoring social behaviour on crowded rookeries where visual identification can often be difficult and uncertain.