Automatic segmentation and identification of whistles produced by dolphins

Previously, abundant analysis and researches have been devoted to the marine acoustic signals of dolphins. In attempt to obtain the accurate signal segments and classify the categories of acoustic data, most researchers divide the signal and classify them manually. This paper aims to introduce a new method of automatically segmenting and identifying whistles and clicks of dolphins. By analyzing the spectrogram of acoustic signal with binarization processing by a self-adapting threshold, the time series about the sound contour could be obtained. Focusing on the time sequence, signals could be segmented automatically by heuristic segmentation algorithm, which could further automatically identify the clicks and whistle by Runs Test.