Gujrati Character Recognition Using Adaptive Neuro Fuzzy Classifier

Abstract

This paper presents implementation of an Adaptive Neuro Fuzzy Classifier (ANFC) for recognition of isolated handwritten characters of Gujrati based on [2]. Authors aim to compare the performance of ANFC with weighted k - NN classifier proposed in [1] by them. Fuzzy classification is the task of partitioning a feature space into fuzzy classes. Authors exploit the method of employing adaptive networks based on [2] to solve a fuzzy classification problem. System parameters, such as the membership functions defined for each feature and the parameterized t-norms used to combine conjunctive conditions are calibrated with back propagation. Towards this aim, authors use a supervised learning procedure based on Scaled Conjugate Gradient (SCG) algorithm to update parameters in an adaptive network. Next, this architecture is deployed for the character recognition problem. From the experimental results, it is summarized that although adaptively adjusted classifier performs well as far as time complexity is concerned but fails to achieve better recognition rates than weighted k - NN. The results are discussed from the viewpoint of feature extraction methods discussed in [1] and their effectiveness on neuro fuzzy classifiers.