Isolated handwritten digit recognition is still a difficult task for a computer although a lot of research has been done on this topic for over two decades. The main difficulties arise due to a large number of variations and styles of digit patterns. Literature reveals that a great amount of research has been made to solve the problem. A variety of feature selection methods and classification methods have been proposed to improve the performance of the recognition system. In this paper, we have studied the functional link artificial neural networks (FLANN) and single-layer perceptron network for the task of classification, where the aim is to classify the input numeral as one of two classes. In contrast to multilayer perceptron network which increases the complexity of the network, we have compared the performance of two single-layer feedforward structures for the task of digit recognition. Using the functionality-expanded features, FLANN overcomes the nonlinearity nature of the problem which is commonly encountered in single-layer perceptron. Experimental results demonstrate that on a database of 30 digit patterns written by 30 people, FLANN-based classifier exhibits a recognition rate of 90 % when compared with perceptron-based classifier which exhibits only 30 % accuracy. © Springer India 2015.
CITATION STYLE
Samal, A., Panda, J., & Das, N. (2015). Performance Comparison of Single-Layer Perceptron and FLANN-Based Structure for Isolated Digit Recognition. In Advances in Intelligent Systems and Computing (Vol. 308 AISC, pp. 237–246). Springer Verlag. https://doi.org/10.1007/978-81-322-2012-1_25
Mendeley helps you to discover research relevant for your work.