Sonar signal classification using the BCM learning algorithm

Previous work by the author has demonstrated the capability of the Bienenstock Cooper Munro (BCM) model (proposed in 1982) of neural synaptic modification to perform feature extraction, thus enhancing the performance of automated classifiers. Recent work has applied the BCM algorithm to sonar images of minelike objects, with the output of the BCM networks fed into a neural network classifier. This paper demonstrates the capability of this approach to classify, objects as minelike or non-minelike, and to further classify the minelike objects by type.