Performance comparison of neural networks and pattern recognition techniques for classifying ultrasonic transducers

Neural networksandpatternrecognition techniques have been effectively used in characterizingand classifying patterns.In this paper, we focus on the performanceof both techniques in classifying ultrasonic transducers.The work was aimed at comparingthe performanceof both techniques for : (1) preclustering of transducers into classes; (2)~ognizing transducers.‘l’hecharacterizationalgorithms used are based on neural network and pattern recognition techniques. The competitive learning technique provides poor results as compared to the K-Means for prtxlustering. While for recognition, it is found that artificial neural network techniques provide tkr better classifkation nsults as compared to the pattern remgnition techniques. A multilayerbackpmpagation neural network is developed for characterizing the transducers which provides a misclassifwation rate of 6%. Two other multilayer neural networks, sum-of-products and a newly devised neural network called hybrid sum-of-products have a misclassification rateof 10%and 7%, respectively. The best patternrecognition technique for this application is found to be the perception which provides a misclassification rate of 23%. The worst patternrecognition technique is found to be the Bayes’ theorem method which provides a misclassification rate of 54%. Index Terms Performance Evaluation, Neural Networks, Pattern Recognition, Ultrasonic Transducers, Classification, CharacterizationTechniques. Permission to copy without fee all or part of this material is granted provided that tha oopies are not made or distributed for direct commercial advantage, the ACM copyright notica and the title of the publication and ite date appear, and notice is givan that cop~ngiaby permissionof theAssociationfor Computing Machinery.Tc copyotherwise,cr to republish,requiresa fee and/orspecificpermieeion. 01992 AcM o.aglgl.sOz-x/gz~ z/lzsd...$ 1.So