Analysis of the neuro-fuzzy classifiers for fish species identification from acoustic echoes

Abstract

Some of the variety of classification methods have been recently implemented also for fish species identification from their acoustic echoes. The classification systems employing artificial neural networks and fuzzy logic separately are not satisfactory, so the combined neuro-fuzzy approach were carried out. A couple of neuro-fuzzy classifiers were built and tested using the NEFCLASS program, which discovered the rules and found the shape of membership functions for determining correct class categories for a given input data set. The systems were tested on fish echoes acquired by DT4000 digital echo sounder. The results are promising with reference to the classifiers performance and their generalization abilities.