Handwritten Character Input And Recognition Based On Computer Vision

With the popularity of consumer electronics end products, a relatively high demand is proposed for the input of the information. The way of original keyboard input can not meet the requirements in both of the input efficiency and appearance, while, the way of the handwritten character input as an efficient input method has much of the people of all ages. So, the way of key input will be replaced by handwritten character input because of its excellent and convenient performance, which has become a direction of development. The recognition of handwritten character has been a focus for researchers, which is one of important methods to solve the problem about the input of handwritten character. Therefore, the research about the technology of handwritten character input and recognition has a profound theoretical significance and broad application prospects.Firstly, it was introduced that the basic principles about the image processing and the feature extraction. The design of the non-contact handwritten character input system has been done by using an ordinary camera and the laser pointer; Secondly, the experiment of feature extraction and classification has been done about number, letter and character by using the knowledge of the pattern recognition.Considering the gray distribution features of the spot under the different circumstances, the center of the laser spot was extracted to resumpte the spot image with the method of gray-scale gravity, and the recovering and denoising about the spot track were completed combinating the knowledge of the image morphology. The method of the spot location and the track recovery proposed in this paper is simple and practical, which does not require complicated mathematical model.The identification of three types of characters were described in this paper. It is identified using the two features of topology characteristics and moment invariants for digital letter and the alphabet; while the handwritten character make full use of the elastic deformation of the ability of local grid and Gabor filter extraction in the direction of the Chinese characters and combining with the features of stroke K-neighbor classifier for classification and recognition, the highest recognition rate can achieve92%. The paper in non-contact handwritten character input and recognition to do some research, the experimental results show that the proposed method is reliable and practical, also, obtained high recognition rate.