The Research And Application On Neural Network Pattern Recognition In The Dispenser

To recognize a pattern is a human basic activity , people hope that the machine can replace the mankind and do the work , so the theory and method of the pattern-recognition .which has already developed into multi- disciplinary disciplines that cross of one door now, have caused great interests of people and carried on long-term research. For different purpose , different method can be used. Because artificial neural network can simulate better image thinking of people, and can run side by side with stronger fault-tolerant ability and association ability and also stronger learning ability in coordination on a large scale, it has become a domestic and international focus that it is applied to pattern- recognition for scientific worker.The development overview for the pattern-recognition neural network and the use of neural network to recognize the pattern is described at first in this text, then some basic conception , method of the pattern-recognition is analysed, and also point out its difficulty and request, at the same time , after providing the neural method of carrying on the pattern-recognition of network and compare the similarities and differences between it and traditional method, the characteristic and superiority of the pattern-recognition using the neural network method is carried on; The first key problem of the neural network mode recognition system emphatically have studied, namely mode pretreatment and characteristic drawn, then it shows the basic K-L varies to realizy PCA and how to realize the method step of PCA especially; On this basis, this text continues the detailed research of the theoretical foundation of neural network for pattern-recognition, thus stated the feasibility of its realization; Then it provide a realization thought of BP neural network mode classifying device to the core of this text. With above-mentioned research results, a concrete neural network mode recognition system, namely dispenser system is succeeded in developing and designing. With the aid of MATLAB , it have carried on samples and trained and finished the neural network structural design to the prototype of a machine that is developed in improved BP algorithm. Laboratory test to prototype of a machine carried on indicates their operation stability , a high discerning rate relatively for the paper currency, and have a greater practical value , thus indicated the validity of the achievement from this text research and development.