SSC San Diego TD 627 Revision D,
Annotated Bibliography of Publications from the U.S. Navy's Marine Mammal Program, May 1998


2. NEURAL NETWORKS

Anderson, L.N., A.R. Rasmussen, W.W.L. Au., P.E. Nachtigall, and H.L. Roitblat. 1994.Neural Network Modeling of a Dolphin’s Sonar Discrimination Capabilities. (Abs.). Jour. Acoust. Soc. Am. 96 (5, Pt. 2): 3316.

A previous modeling experiment used only spectral information of a dolphin’s echo. In this study, both time and frequency information were used to model the dolphin discrimination capabilities.

Au, W.W.L. 1994. Comparison of Sonar Discrimination: Dolphin and an Artificial Neural Network. Jour. Acoust. Soc. Am. 95 (5, Pt. 1): 2728-2735.

Discusses comparison of capabilities of a dolphin and an artificial neural network to determine wall thicknesses of cylinders. The network was used to examine broadband echo features from the same cylinders employed in a earlier study of dolphin capabilities. Results indicate use of neural network technology may assist in understanding dolphin sonar and provide insights on different cues available for target discrimination.

Au, W.W.L. and P.E. Nachtigall. 1994. Artificial Neural Network Modeling of Dolphin Echolocation. (Abs.) Harderwijk Marine Mammal Sensory Symposium, Harderwijk, The Netherlands, April 28-May 3.

Discusses general concepts of artificial neural network modeling, with its application to dolphin echolocation in target discrimination tasks. Emphasizes a study comparing performance of a counterpropration neural network and a dolphin in discriminating wall thickness differences of aluminum cylinders.

Au, W.W.L. and P.E. Nachtigall. 1995. Artificial Neural Network Modeling of Dolphin Echolocalocation. In: Sensory Systems of Aquatic Mammals, R. A. Kastelein, J.A. Thomas, and P.E. Nachtigall (eds.). DeSpil Publishers, Woerden, The Netherlands, pp. 183-199.

See Au and Nachtigall, 1994, above.

Au, W.W.L., L. Anderson, A. Rasmussen, H.L. Roitblat, and P.E. Nachtigall. 1995. Neural Network Modeling of a Dolphin's Sonar Discrimination Capabilities. Jour. Acoust. Soc. Amer. 98: 43-50.

A counterpropagation neural network, a backpropagation neural network and a model using Euclidean distance measures were used to analyze digitized echoes from cylinders in an effort to model dolphin sonar discrimination capabilities. Both time and frequency information were used in the study. The backpropa-gation network outperformed the others, using spectral-only features of the echoes and also combined time and frequency features. In some specific instances, the backpropagation network also performed better than the dolphin.

Au, W.W.L., R.H. Shizumura, P.E. Nachtigall, R.J. Hicks, H.L. Roitblat, and G. Moons. 1996. Aspect-Independent Sonar Recognition of Cylinders Using Dolphin-Like Signals. (Abs.). Jour. Acoust. Soc. Am. 100 (4, Pt. 2): 2643.

Echoes of a dolphin-like sonar signal from four different cylinders were collected and used in a neural network recognition study. The network achieved better than 90 percent recognition of the echoes.

Helweg, D.A., H.L. Roitblat, and P.E. Nachtigall. 1993. Using a Neural Network to Model Dolphin Echolocation. In: Artificial Neural Networks and Expert Systems, N. Kasabov (ed.). IEEE Computer Society Press, Los Alamitos, CA, pp. 247-251.

A biomimetic neural network was used to model a bottlenose dolphin's ability to recognize aspect-dependent targets. Researchers used echo trains recorded during the dolphin trials to train an Integrator Gateway Network (IGN) to discriminate among the targets using echo spectra. The dolphin and the IGN learned to recognize geometric targets, even though orientation could vary. Results support the notion that ensonified underwater objects with complex shapes and echoes may be reliably classified using neural network architectures that are motivated through understanding of dolphin echolocation signals and performance.

Helweg, D.A. and P.W.B. Moore. 1997. Classification of Aspect-Dependent Targets by a Biomimetic Neural Network. NRaD TR 1747. 6 pp.

A biomimetic neural network was used to model a bottlenose dolphin's ability to recognize aspect-dependent targets. The results support the notion that ensonified mines with complex shapes and echoes may be reliably classified using neural network architectures motivated through understanding of marine mammal system echnolocation signals and performance.

Moore, P. W. B., H. L. Roitblat, P. E. Nachtigall, and R. H. Penner. 1990. Classifying Dolphin Echoes Using an Integrator Gateway Artificial Neural Network. Jour. Acoust. Soc. Am. 90 (2): 2334.

See other articles by Moore, et al., 1990, below.

Moore, P. W. B., H. L. Roitblat, R. H. Penner, and P. E. Nachtigall. 1990. An Integrator Gateway Network for Recognizing Dolphin Echoes. Government Neural Network Applications Workshop, August 29-31, 1990, San Diego CA.

The application of the gateway integrator neural network for classifying various signals was presented in this classified workshop.

Moore, P. W. B., H. L. Roitblat, R. H. Penner, and P. E. Nachtigall. 1990. An Integrator Gateway Network for Recognizing Dolphin Echoes. Conf. on Neural Networks for Decision, Estimation, and Control, West Greenwich, RI.

A new neural network design based on the properties of the echolocating dolphin was presented and discussed in a classified conference on government signal-processing approaches.

Moore, P. W. B., H. L. Roitblat, P. E. Nachtigall, R. H. Penner, and W. W. L. Au. 1990. Sonar Target Recognition by an Artificial Neural Network. Naval Research and Development Information Exchange Conf., Naval Air Development Center, Warminster, PA, p. 48.

Detailed presentation of the integrator gateway network. This network (patent applied for) combines information from multiple signals and resets between trains of signals. This artificial neural network model was compared against a standard neural network model that did not include the integrating components and was found to improve object recognition substantially.

Roitblat, H. L., P. W. B. Moore, R. H. Penner, and P. E. Nachtigall. 1989. Clicks, Echoes, and Decisions: The Use of Information by a Bottlenosed Dolphin (Tursiops truncatus). (Abs.) Eighth Biennial Conf. on the Biology of Marine Mammals, Society of Marine Mammalogy, Pacific Grove, CA., p. 56.

Describes the pattern by which the dolphin searched alternative comparison stimuli in a delayed matching-to-sample task and some preliminary neural network models for dolphin echolocation.

Roitblat, H.L., P. W. B. Moore, P. E. Nachtigall, R. H. Penner, and W. W. L. Au. 1989. Natural Echolocation With an Artificial Neural Network. International Journal of Neural Networks: Research and Applications 1 (4): 239-248.

The performance of a dolphin performing in a matching-to-sample echolocation task was simulated with a counterpropagation artificial neural network. The neural network performance compared well with that of the dolphin when echoes collected while the dolphin echolocated were used.

Roitblat, H. L., P. W. B. Moore, P. E. Nachtigall, R. H. Penner, and W. W. L. Au. 1989. Dolphin Echolocation: Identification of Returning Echoes Using a Counterpropagation Network. International Joint Conf. on Neural Networks, Vol. 1, IEEE and International Neural Network Society, Piscataway, NJ, pp. 295-301.

Describes preliminary work on using a counterpropagation artificial neural network to recognize echoes from objects ensonified in a test pool by an artificial dolphin click and in Kaneohe Bay by a dolphin during performance of a delayed matching- to-sample task. Selected echoes were analyzed and successfully recognized by the network. Target recognition abilities of an echolocating dolphin and the neural network were also compared. In a noisy natural environment, the dolphin was 94.5 percent correct and the network was 96.7 percent correct. Possible applications of neural networks to echolocation studies are discussed.

Roitblat, H. L., P. W. B. Moore, P. E. Nachtigall, and R. H. Penner. 1991. Natural Dolphin Echo Recognition Using an Integrator Gateway Network. In: Advances in Neural Information Processing Systems, Vol. 3. D. S. Touretsky, J. E. Moody and R. Lippman (eds.). Morgan Kaufmann, San Mateo, CA, pp. 273-281.

Discusses the integrator gateway network for recognizing objects ensonified by dolphin echolocation signals.

Roitblat, H. L., P. W. B. Moore, P. E. Nachtigall, and R. H. Penner. 1991. Biomimetic Sonar Processing: From Dolphin Echolocation to Artificial Neural Networks. In: From Animals to Animats, J. A. Meyer and S. Wilson (eds.). MIT Press, Cambridge, MA, pp. 66-76.

Describes a dolphin’s recognition performance and some aspects of a neural network model of echo recognition that incorporated properties of the sequential sampling model to combine information from successive dolphin echoes.

Root, W. A., and S. H. Ridgway. 1991. Neural Network Applications in Dolphin Response-time Studies. Jour. Acoust. Soc. Am. 90: 2334.

Dolphins (Tursiops truncatus) were trained to make different sounds in response to two different acoustic stimuli produced by a computer system. A neural network was shown to be better at identifying response type and setting response latency than was a previously employed discriminant analysis routine.

Top

Contents