A wavelet based time-frequency descriptor for automatic classification of acoustic signals of fishes

Fishes are capable of emitting sounds for diverse purposes, which can be recorded through microphone sensors. This work aims to develop a methodology for automatic classification of fishes through their acoustic signals by applying pattern recognition techniques. Fish sounds are preprocessed and segmented automatically to extract each call. These calls are parametrized using local wavelet acoustic pattern (LWAP) and approximation coefficients obtained using wavelet decomposition, to yield useful information for classification. Three machine learning algorithms: discriminant analysis (DA), k- nearest neighbors (k-NN) and support vector machines (SVM) are used for the fish call identification. Syllables are identified with accuracies of 84, 89 and 56 percentages for DA, k-NN and SVM classifiers.