EMPIRICAL RESULTS OF USING BACKPROPAGATION NEURAL NETWORKS TO SEPARATE SINGLE ECHOES FROM MULTIPLE ECHOES

Empirical results illustrate the pitfalls of applying an artificial ne ural network (ANN) tb classification of underwater active sonar return s. During training, a back-propagation ANN classifier ''learns'' to re cognize two classes of reflected active sonar waveforms. Waveforms in Class 1 have two major sonar echoes or peaks. Waveforms in Class 2 hav e one major echo or peak Our results show how the classifier ''learns' ' to distinguish between the two classes. Testing the ANN classifier w ith different waveforms having one major-peak and waveforms having two major peaks generated unexpected results: the number of echo peaks, w as not the feature used to separate classes.