Combining classifiers for Offline Malayalam Character Recognition

Abstract

Offline Character Recognition is one of the most challenging areas in the domain of pattern recognition and computer vision. Here we propose a novel method for Offline Malayalam Character Recognition using multiple classifier combination technique. From the preprocessed character images, we have extracted two features: Chain Code Histogram and Fourier Descriptors. These features are fed as input to two feedforward neural networks. Finally, the results of both neural networks are combined using a weighted majority technique. The proposed system is tested using two schemes-Writer independant and Writer dependant schemes. It is observed that the system achieves an accuracy of 92.84% and 96.24% respectively for the writer independant and writer dependant scheme considering top 3 choices.