Recognizing transient low-frequency whale sounds by spectrogram correlation

A method is described for the automatic recognition of transient animal sou nds. Automatic recognition can be used in wild animal research, including s tudies of behavior, population, and impact of anthropogenic noise. The meth od described here, spectrogram correlation, is well-suited to recognition o f animal sounds consisting of tones and Frequency sweeps. For a sound type of interest, a two-dimensional synthetic kernel is constructed and cross-co rrelated with a spectrogram of a recording, producing a recognition functio n-the likelihood at each point in time that the sound type was present. A t hreshold is applied to this function to obtain discrete detection events, i nstants at which the sound type of interest was likely to be present. An ex tension of this method handles the temporal variation commonly present in a nimal sounds. Spectrogram correlation was compared to three other methods t hat have been used for automatic call recognition: matched filters, neural networks, and hidden Markov models. The test data set consisted of bowhead whale (Balaena mysticetus) end notes from songs recorded in Alaska in 1986 and 1988. The method had a success rate of about 97.5% on this problem, and the comparison indicated that it could be especially useful for detecting a call type when relatively few (5-200) instances of the call type are know n. (C) 2000 Acoustical Society of America. [S0001-4966(00)01706-9].