{"title":"A New High Speed Neural Model for Fast Character Recognition Using Cross Correlation and Matrix Decomposition","authors":"Hazem M. El-Bakry","country":null,"institution":"","volume":20,"journal":"International Journal of Computer and Information Engineering","pagesStart":2857,"pagesEnd":2877,"ISSN":"1307-6892","URL":"https:\/\/publications.waset.org\/pdf\/11267","abstract":"Neural processors have shown good results for\r\ndetecting a certain character in a given input matrix. In this paper, a\r\nnew idead to speed up the operation of neural processors for character\r\ndetection is presented. Such processors are designed based on cross\r\ncorrelation in the frequency domain between the input matrix and the\r\nweights of neural networks. This approach is developed to reduce the\r\ncomputation steps required by these faster neural networks for the\r\nsearching process. The principle of divide and conquer strategy is\r\napplied through image decomposition. Each image is divided into\r\nsmall in size sub-images and then each one is tested separately by\r\nusing a single faster neural processor. Furthermore, faster character\r\ndetection is obtained by using parallel processing techniques to test the\r\nresulting sub-images at the same time using the same number of faster\r\nneural networks. In contrast to using only faster neural processors, the\r\nspeed up ratio is increased with the size of the input image when using\r\nfaster neural processors and image decomposition. Moreover, the\r\nproblem of local subimage normalization in the frequency domain is\r\nsolved. The effect of image normalization on the speed up ratio of\r\ncharacter detection is discussed. Simulation results show that local\r\nsubimage normalization through weight normalization is faster than\r\nsubimage normalization in the spatial domain. The overall speed up\r\nratio of the detection process is increased as the normalization of\r\nweights is done off line.","references":null,"publisher":"World Academy of Science, Engineering and Technology","index":"Open Science Index 20, 2008"}