Performance comparison of classifiers for bilingual gurmukhi-roman online handwriting recognition system

Abstract

Bilingual or multilingual script recognitionsystems for Online handwriting recognition (OHR) have been considered as a hot area of research. Here, more than one scripts have been used to generate handwriting samples.Unique writing style of each script increases the complexity of these systems. In this paper, performance comparison of classifiers is presented for bilingual Gurmukhi-Roman OHR system. This proposed systemprocessed intermixed bilingual handwritten text inputted through a digitizer tablet and pen. Various steps like input of handwriting samples, pre-processing, segmentation, feature extraction, classification and post-processing have been implemented to get digital data. The main emphasis has been given to the classification phase. Three different classifiers Multi-Layered Perceptron (MLP) neural network, Support Vector Machine (SVM) and Hidden Markov Model (HMM) have been implemented. The performance of these classifiers hasbeen compared and it is observed that MLPshows better results as compare to SVM and HMM classifiers.