Research And Implement Of Cascade Neural Networks Handwritten Numeral Recognition

Thinking Science generally thought that the human mind is divided into three basic ways of thinking that is abstract thinking (logical) thinking, the image (visual) thinking and inspiration (Epiphany). Simulation of artificial neural network like the second way people think. This is a non-linear dynamic system, its characteristic is the information distributed storage and parallel co-processing. Although the structure of a single neuron is extremely simple, limited functionality, a large number of neurons in a network system can bring about plentiful function realization.Handwritten digit recognition is a long standing research focus, but also character recognition in a particular problem. Handwritten digit recognition in a specific environment, particularly wide range of applications. When it comes to digital identification, people often require a high level of recognition reader reliability, particularly in relation to the amount of the number of recognition, therefore, for the treatment of such issues a key element of system design is to design a highly reliable and high recognition rate of the method of handwritten numeral recognition. It can be said, however, has no handwritten numeral recognition devices to achieve a perfect recognition effect.Neural Network Applications and handwritten numeral recognition study of the neural network from a single direction to the integration, due to neural network ensemble can significantly improve the system capacity of the prosperity, it will be integrated neural networks applied to handwritten digits has become a research hotspot.This paper presents a BP neural network based on cascaded network, based on BP neural network to further improve the structure of the network as a whole identify with the structure of rational design of recognition systems to improve the recognition rate, thereby enhancing the recognition rate and the entire system's ability.