Bottom type identification using combined neuro-fuzzy classifier operating on multi-frequency data

Abstract

The paper introduces a novel approach to acoustic methods of characterising the bottom type by using a neuro-fuzzy classifier which processes the bottom backscatter data collected with an echosounder on three different operating frequencies. The classifier combining fuzzy logic and artificial neural networks was created using NEFClass system. It constitutes a fuzzy system, which can be viewed as a special 3-layer feed-forward neural network architecture, where the nodes of the second layer represent fuzzy rules. These rules are derived from a set of training data separated into crisp classes. In training and testing stages, apart from using single-frequency data, sets of dual-frequency and triple-frequency data combined together were used in order to enhance the classifier's performance. The results show that combining dual-frequency, or moreover triple-frequency data, clearly improves the generalisation ability of the classifier. The bottom backscattered echoes were acquired from acoustic surveys carried out on Lake Washington using the single-beam digital echosounder working on three frequencies: 38 kHz, 120 kHz and 420 kHz.