Acoustic modeling, phantom echolocation experiments, and time-frequency analysis of dolphin sonar for improved technological sonar systems

Dolphin echolocation has been demonstrated to be a remarkable natural sonar system, one which may greatly exceed the performance of technological sonar systems in terms of detection and discrimination of targets. A dolphin's ability to discriminate targets may depend on the characteristics of echo highlights within a received signal. Previous experiments with dolphins varied the physical parameters of targets, but did not fully investigate how changes in individual echo components within the scattered waveforms affected the dolphin's subsequent response. The experiments utilize a phantom echo system to test a dolphin's detection response to relative amplitude differences of echo highlights and time separation differences between echo highlights. By electronically manipulating the amplitude and temporal separation of the highlights, the underlying acoustic cues are more efficiently investigated. Based on the dolphin's performance, the animal is sensitive to changes in the relative amplitudes and temporal separations of the highlights. Center frequency and rms bandwidth are calculated for the clicks collected during the experiments and the clicks are clustered into four classes using model based clustering. Echo signatures are obtained from elastic, isotropic spheres for each class of clicks with an acoustic scattering model. The results from the model are compared with experimentally obtained results. The joint time-frequency content of the resulting echo signals are obtained by the reduced interference distribution (RID). The RIDs are examined for each signal class for four different spherical targets. RID correlation values are obtained for a standard target versus comparison targets using a time-frequency correlator. Matching pursuit decomposition is applied to the signals to study frequency changes within the dolphin's functional bandwidth during discrimination tasks. The matching pursuit algorithm extracts subtle frequency differences that traditional time-frequency analysis techniques do not reveal. Furthermore, ROC analysis is applied to the relative energies of the matched waveforms to determine probability of discrimination. The results support the hypothesis that dolphins may discriminate by altering outgoing clicks and utilizing the time-frequency information returned by the targets. In a similar manner, by altering transmission signals and incorporating time-frequency signal processing algorithms, it may be possible to improve technological sonar models.