Handwritten digit recognition based on improved BP neural network

Abstract

Due to the different writing habits, it is difficult to achieve the recognition of handwritten numbers. The artificial neural network has been widely used in character recognition because of its strong self-learning ability, adaptive ability, classification ability, fault tolerance and fast recognition. BP neural network is used to identify handwritten numerals in this paper. In order to obtain a higher correct rate, this paper improves the traditional BP neural network and experiments with the MNIST data set on the MATLAB simulation platform. The experimental results show that the improved network converge faster and the classification is more accurate.