Examining the effects of propagation on perceptual features used for automatic aural classification

A prototype aural classifier has been developed at Defence R&D Canada that uses perceptual signal features which model the features employed by the human auditory system. Previous effort has shown the classifier reduced false alarm rates and successfully discriminated cetacean vocalizations from several species. The current paper investigates the robustness of the aural features against propagation effects for two of those species - the bowhead and humpback whales. This is achieved by comparing the classification results of the original vocalizations to classification results obtained after the vocalizations were re-transmitted underwater over ranges of 2 to 10 km. To gain additional insight into the propagation effects, synthetic bowhead and humpback vocalizations, with features similar to the most important aural features for classification of bowhead and humpback vocalizations, were also transmitted. In this paper, the classifier performance on both real and synthesized vocalizations are compared before and after propagation, to quantify the effect of propagation on the features used in the aural classifier.