A neural network method for recognition of individual Orcinus orca based on their acoustic behaviour: phase 1

Using a backpropagation neural network algorithm three experiments were designed to discriminate between the acoustic behaviour of Orcinus orca. The first experiment established the procedures by discriminating human vocalisations. The second and third experiments applied the procedures established in the first to the discrimination of Orca acoustic behaviour. Experiment two and three were designed to answer the following questions: (1) is it possible to discriminate between the calls of individual whales within the same pod, i.e., identify the whales by their calls, and (2) when the same call is made by different whales is it possible to identify the whales? The results indicate that the neural networks are able to accurately discriminate between individual calls and also between different whales making the same call. Implications of these results for the on-line classification of whales and how this will affect the study of Orca are discussed.<
>