Using a binaural biomimetic array to identify bottom objects ensonified by echolocating dolphins.

The development of a unique dolphin biomimetic sonar produced data that were used to study signal processing methods for object identification. Echoes from four metallic objects proud on the bottom, and a substrate-only condition, were generated by bottlenose dolphins trained to ensonify the targets in very shallow water. Using the two-element ('binaural') receive array, object echo spectra were collected and submitted for identification to four neural network architectures. Identification accuracy was evaluated over two receive array configurations, and five signal processing schemes. The four neural networks included backpropagation, learning vector quantization, genetic learning and probabilistic network architectures. The processing schemes included four methods that capitalized on the binaural data, plus a monaural benchmark process. All the schemes resulted in above-chance identification accuracy when applied to learning vector quantization and backpropagation. Beam-forming or concatenation of spectra from both receive elements outperformed the monaural benchmark, with higher sensitivity and lower bias. Ultimately, best object identification performance was achieved by the learning vector quantization network supplied with beam-formed data. The advantages of multi-element signal processing for object identification are clearly demonstrated in this development of a first-ever dolphin biomimetic sonar.

Multiecho processing by an echolocating dolphin.

Bottlenose dolphins (Tursiops truncatus) use short, wideband pulses for echolocation. Individual waveforms have high-range resolution capability but are relatively insensitive to range rate. Signal-to-noise ratio (SNR) is not greatly improved by pulse compression because each waveform has small time-bandwidth product. The dolphin, however, often uses many pulses to interrogate a target, and could use multipulse processing to combine the resulting echoes. Multipulse processing could mitigate the small SNR improvement from pulse compression, and could greatly improve range-rate estimation, moving target indication, range tracking, and acoustic imaging. All these hypothetical capabilities depend upon the animal's ability to combine multiple echoes for detection and/or estimation. An experiment to test multiecho processing in a dolphin measured detection of a stationary target when the number N of available target echoes was increased, using synthetic echoes. The SNR required for detection decreased as the number of available echoes increased, as expected for multiecho processing. A receiver that sums binary-quantized data samples from multiple echoes closely models the N dependence of the SNR required by the dolphin. Such a receiver has distribution-tolerant (nonparametric) properties that make it robust in environments with nonstationary and/or non-Gaussian noise, such as the pulses created by snapping shrimp.

Discrimination of complex synthetic echoes by an echolocating bottlenose dolphin.

Bottlenose dolphins (Tursiops truncatus) detect and discriminate underwater objects by interrogating the environment with their native echolocation capabilities. Study of dolphins' ability to detect complex (multihighlight) signals in noise suggest echolocation object detection using an approximate 265-micros energy integration time window sensitive to the echo region of highest energy or containing the highlight with highest energy. Backscatter from many real objects contains multiple highlights, distributed over multiple integration windows and with varying amplitude relationships. This study used synthetic echoes with complex highlight structures to test whether high-amplitude initial highlights would interfere with discrimination of low-amplitude trailing highlights. A dolphin was trained to discriminate two-highlight synthetic echoes using differences in the center frequencies of the second highlights. The energy ratio (delta dB) and the timing relationship (delta T) between the first and second highlights were manipulated. An iso-sensitivity function was derived using a factorial design testing delta dB at -10, -15, -20, and -25 dB and delta T at 10, 20, 40, and 80 micros. The results suggest that the animal processed multiple echo highlights as separable analyzable features in the discrimination task, perhaps perceived through differences in spectral rippling across the duration of the echoes.

Discrimination of amplitude-modulated synthetic echo trains by an echolocating bottlenose dolphin.

Bottlenose dolphins (Tursiops truncatus) have an acute ability to use target echoes to judge attributes such as size, shape, and material composition. Most target recognition studies have focused on features associated with individual echoes as opposed to information conveyed across echo sequences (feature envelope of the multi-echo train). One feature of aspect-dependent targets is an amplitude modulation (AM) across the return echoes in the echo train created by relative movement of the target and dolphin. The current study examined whether dolphins could discriminate targets with different AM envelopes. "Electronic echoes" triggered by a dolphin's outgoing echolocation clicks were manipulated to create sinusoidal envelopes with varying AM rate and depth. Echo trains were equated for energy, requiring the dolphin to extract and retain information from multiple echoes in order to detect and report the presence of AM. The dolphin discriminated amplitude-modulated echo trains from those that were not modulated. AM depth thresholds were approximately 0.8 dB, similar to other published amplitude limens. Decreasing the rate of modulation from approximately 16 to 2 cycles per second did not affect the dolphin's AM depth sensitivity. The results support multiple-echo processing in bottlenose dolphin echolocation. This capability provides additional theoretical justification for exploring synthetic aperture sonar concepts in models of animal echolocation that potentially support theories postulating formation of images as an ultimate means for target identification.

Acoustic identification of female Steller sea lions (Eumetopias jubatus).

Steller sea lion (Eumetopias jubatus) mothers and pups establish and maintain contact with individually distinctive vocalizations. Our objective was to develop a robust neural network to classify females based on their mother-pup contact calls. We catalogued 573 contact calls from 25 females in 1998 and 1323 calls from 46 females in 1999. From this database, a subset of 26 females with sufficient samples of calls was selected for further study. Each female was identified visually by marking patterns, which provided the verification for acoustic identification. Average logarithmic spectra were extracted for each call, and standardized training and generalization datasets created for the neural network classifier. A family of backpropagation networks was generated to assess relative contribution of spectral input bandwidth, frequency resolution, and network architectural variables to classification accuracy. The network with best overall generalization accuracy (71%) used an input representation of 0-3 kHz of bandwidth at 10.77 Hz/bin frequency resolution, and a 2:1 hidden:output layer neural ratio. The network was analyzed to reveal which portions of the call spectra were most influential for identification of each female. Acoustical identification of distinctive female acoustic signatures has several potentially important conservation applications for this endangered species, such as rapid survey of females present on a rookery.