Modified quadratic classifier for Handwritten Malayalam Character recognition using Run Length Count

The strength of the selected feature and the effectiveness of the classifier are the two key factors determining the performance of a handwritten Character Recognition System. In this work, we implemented a feature extraction method based on Run Length Count (RLC) for the offline recognition of Handwritten Malayalam Characters. RLC is the count of contiguous group of 1's encountered in a left to right / top to bottom scan of a character image or block of an image. Fixed Meshing strategy is followed for blocking the character images and RLC of the different blocks forms the feature vector for classification. For classification, we implemented Modified Quadratic Discriminant function (MQDF), which is a successful statistical approach for Handwritten Character Recognition. The classifier gives 94.18% accuracy for a feature vector of size 51, which is a significant achievement in isolated Malayalam HCR systems. The study was carried out with a database containing 15,000 handwritten Malayalam character samples. Feature extraction with RLC contributes to the work by its simplicity and lesser execution time. Compared to neural network, MQDF require much lesser training time.