The Research Of Character Recognition Technology And Application Based On Neural Network

Character recognition mainly includes the analysis and processing of the text information, the text feature extraction and text recognition judgment. It is mainly applied in reading, information entry and office automation and other fields. Feature extraction is the core of Character recognition technology, The outstanding feature extraction algorithm determines the character recognition speed and recognition accuracy. Character feature is divided into structural features and statistical features, structural characteristics including skeleton features, contour features and directional line element features; statistical characteristics of grid characteristics, peripheral features and complexity index characteristics. However, due to the presence of pollution and similar shaped character, these factors brought a certain degree of difficulty to character recognition.Since the birth of the neural network, it constantly moving forward and continue to improve its ability to deal with the problem, it is involved in more and more science areas, and has been widely applied to many fields. RBF (Radial Basis Function) neural network is a forward feedback neural network consist of two levels. During to the advantages of learning fast, it get more and more attention. It equips ability of self-learning, self-organization, adaptive function, high-speed in processing of data in parallel. Experiments show that the network can approximate any nonlinear function, and there are bright application prospects in the character recognition research, we can take advantage of its extremely fast learning ability, the convergence rate characteristics and the best approximation in improving recognition speed and accuracy.The subject research and analysis the character recognition technology and its application, compared and analyzed the character feature extraction algorithm are used widely, and on this basis, a character recognition system based on RBF neural network was designed. As the adoption of improved complexity index feature e ktraction algorithm to extract text features, special treatment was done to special characters and similar characters. After a large number of sample s was trained, the feature vectors extracted from the characters object information as the input of the system in order to achieve the identification of the characters.