Neural network and pattern recognition techniques for characterizing ultrasonic transducers

Abstract

A system in which transducers are characterized and classified using both pattern recognition and neural network algorithms is presented. Various techniques are investigated, compared, and analyzed. The hardware of the system used in collecting and measuring the characteristics of the transducers, and the parameters used in classifying the transducers into classes are described. The neural network and pattern recognition algorithms used to classify the transducers are introduced. The results of the classification are presented and the various algorithms are analyzed and compared. >