Neural networks and the classification of complex sonar signals

The author examines the ability of neural networks to extract high-order information from a set of input patterns, and relates this ability to advantages over previous approaches to both active and passive sonar signal classification. The basic computational structure of feedforward neural networks is reviewed and the ability of these networks to extract high-order information from various signals is examined. The hierarchical neural network, is examined as an alternate means of extracting information from highly structured non-Gaussian sonar signals. The promise of dynamic neural networks as an approach to the classification of complex sonar signals is discussed