Esfahanian, Mahdi

This research presents a novel approach to categorize dolphin whistles into various types. Most accurate methods to identify dolphin whistles are tedious and not robust, especially in the presence of ocean noise. One of the biggest challenges of dolphin whistle extraction is the coexistence of short-time duration wide-band echo clicks with the whistles. In this research, a subspace of select orientation parameters of the 2D Gabor wavelet frames is utilized to enhance or suppress signals by their orientation. The result is a Gabor image that contains a noise free grayscale representation of the fundamental dolphin whistle which is resampled and fed into the Sparse Representation Classifier. The classifier uses the l1 norm to select a match. Experimental studies conducted demonstrate: a a robust technique based on the Gabor wavelet filters in extracting reliable call patterns, and b the superior performance of Sparse Representation Classifier for identifying dolphin whistles by their call type.

Ocean is home to a large population of marine mammals such as dolphins and whales and concerns over anthropogenic activities in the regions close to their habitants have been
increased. Therefore the ability to detect the presence of these species in the field, to
analyze and classify their vocalization patterns for signs of distress and distortion of their
communication calls will prove to be invaluable in protecting these species. The objective of this research is to investigate methods that automatically detect and classify vocalization patterns of marine mammals. The first work performed is the classification of bottlenose dolphin calls by type. The extraction of salient and distinguishing features from recordings is a major part of this endeavor. To this end, two strategies are evaluated with real datasets provided by Woods Hole Oceanographic Institution: The first strategy is to use contour-based features such as Time-Frequency Parameters and Fourier Descriptors and the second is to employ texture-based features such as Local Binary Patterns (LBP) and Gabor Wavelets. Once dolphin whistle features
are extracted for spectrograms, selection of classification procedures is crucial to the success of the process. For this purpose, the performances of classifiers such as K-Nearest Neighbor, Support Vector Machine, and Sparse Representation Classifier (SRC) are assessed thoroughly, together with those of the underlined feature extractors.