Handwritten Digit Recognition Of An Integrated BP Neural Network

In recent years, some facts indicate that as to a complex recognize problem, single method can not get a good performance, meantime there are some reciprocals during different recognize methods, they can improve the correctness via the method of get some recognize methods together, so these years getting some recognize methods together is becoming a hotspot in the field of pattern recognition .In this paper, a handwritten digit recognition system based on integrated neural network is set up. The system was consisted of two parts: Learning part, Recognition part. In the Learning part, seven BP neural networks are trained .There are two steps in the recognition part. They are feature extraction and combined recognition. They are feature extraction and combined recognition. During the realizing of the system, the following is down in this paper.1. Seven features for handwritten digits based on macroscopical , partial and microcosmic are extracted, which are applied in seven respective neural networks.2. In the signal classify method, make some improvements on BP neural network to quicken the network constringency speed and to avoid the fake saturation phenomenon. For example, change the learning-factor, change the S function , betterment the method of the farthest grads, fetch in the better ambulacrum and so forth.3. In the integrate arithmetic , We give the Optimal Linear Combination method base on the degree of judge. The idea of this method is ,first count the degree of every stylebook divide the training stylebook into some areas, in each area use the Optimal Linear Combination method to count some power value .In the test, We first count the degree of judge and then use the relevant power value to count the sort. Many combined...