Research of numeral character recognition technology based on wavelet analysis and RBF neural networks

Taking advantage of the ability of revealing the images details and the local characteristics in the time-frequency domain, this paper presents a new method of character images recognition, which is based on wavelet packets analysis and RBF neural networks. Firstly this paper decomposes character images with wavelet analysis, then reconstructs the discrete wavelet coefficients and calculates energy values, then extracts energy values from various character images to construct energy eigenvectors as the input of the RBF neural networks. By choosing the number of hide nodes and the learning algorithm of weight, a perfect RBF neural network can be created. At last the RBF neural network carries on identifying the numeral character images. The experiment results show that a high rate of recognition can be obtained by this method.