Handwritten Character Recognition Based On Nerual Network

Character are very important tools in human information communication. With the development of the computer and information technologies, the processing and recognition of character information has become very important. Off-line handwritten character recognition is one kind of pattern recognition technique. It involves the technologies of pattern recognition, digital image processing, digital signal processing, artificial intelligence and so on. It is widely used hi information processing, office automatic, machine translation. Research on Off-line handwritten character recognition has great worth.The information of Off-line handwritten character is very scarce (Only 2D pixel matrix), and the handwritten characters have so many classes, complex strokes and shape distortion. All of above have become the obstacle of character recognition. So, it is very important and necessary to study the technology of off-line handwritten character recognition. In this paper, we try to combine the neural network with pattern recognition technology and find a new solution of Off-line handwritten character recognition.The key of Off-line handwritten character recognition are feature extract and classifier design. In this paper, we extract the LLF (Local Line Fitting), CPT (Central Project Transform) and chain code features of characters, and then combine them with the statistical and structural information of characters for character recognition. In view of the feature of neural network and its advantages in pattern recognition, we have a great deal work in Off-line handwritten character recognition based on neural network. We present two recognition methods based on neural network. One of which is the hybrid neural network recognition system, a multi-level neural network classifier constructed by using the multi neural networks integration technology. The other one is the synthetical local nonlinear PCA neural network recognition model constructed by combining the nonlinear generalization of PCA and sub-space pattern recognition technology.We use the two recognition systems in handwritten digitals and characters recognition and obtain some satisfactory results. Compared with some traditional classifiers, our systems have better recognition performances.