The paper develops an efficient but simple adaptive nonlinear classifier for recognition of handwritten Odiya numerals. The standard gradient and curvature features are extracted and nonlinearly mapped by sine/cosine expansions. These nonlinear inputs are fed to a low complexity classifier. The simulation results show excellent classification accuracy when test features are used. Index Terms--Character recognition, handwritten odiya numerals recognition, artificial neural network, gradient feature, curvature feature, principal component analysis and neural network
Conventional error based cost function provides unsatisfactory weight update of an adaptive system when outliers are present in the training signal. To alleviate this problem in this paper a hybrid approach using differential evolution (DE) and Wilcoxon norm is proposed to provide robust training in identification of complex nonlinear systems. Exhaustive simulation study shows superior performance of the new method compared to the conventional square error based minimization method.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.