|International Journal of Applied Information Systems|
|Foundation of Computer Science (FCS), NY, USA|
|Volume 9 - Number 3|
|Year of Publication: 2015|
|Authors: J. O. Adigun, E. O. Omidiora, S. O. Olabiyisi, O. D. Fenwa, O. Oladipo, M. M. Rufai|
J. O. Adigun, E. O. Omidiora, S. O. Olabiyisi, O. D. Fenwa, O. Oladipo, M. M. Rufai . Development of a Genetic based Neural Network System for Online Character Recognition. International Journal of Applied Information Systems. 9, 3 ( June 2015), 1-8. DOI=10.5120/ijais15-451376
Character Recognition has been one of the most intensive research during the last few decades because of its potential applications. However, most existing classifiers used in recognizing online handwritten characters suffer from poor feature selection and slow convergence which affect training time and recognition accuracy. Hence, this paper focused on integrating an optimization (genetic algorithm) into modified backpropagation neural network to enhance the performance of character recognition. This paper proposed a methodology that is based on extraction of features using stroke number, invariant moments, projection and zoning. Genetic algorithm was use as feature selection to optimize the subset of the character for classification. A Modified Genetic Algorithm (MGA) was modified to reduce character recognition errors using fitness function and genetic operators. However, an integration of optimization algorithm (modified genetic algorithm) into an existing modified backpropagation (MOBP) learning algorithm was employed as classifier. For further enhancement of classifier, three classifiers (C1, C2 and C3) were formulated from MGA-MOBP model and evaluated using training time and correct recognition accuracy. C3 performed better than C1 and C2 in terms of convergence rate, correct recognition accuracy and feature selection (its ability to remove irrelevant features of character images). The results of the developed system achieved a false recognition of 0.56% and 99.44% overall recognition accuracy compared with existing models.