Zone based Method to Classify Isolated Malayalam Handwritten Characters using Hu-Invariant Moments and Neural Networks
Paulose Raj, Amitabh Wahi · 2013
Handwritten Character Recognition of Indian languages have been a demanding task in image processing and pattern recognition. Structural complexity and likeness in the characters also increases the complexity in the classification of characters. In this study, Malayalam, a south-Indian language investigated for recognition of its characters using Hu-invariant moments.Moments applied to the preprocessed image after zoning the image. The image divided horizontally, vertically and diagonally to which moments are applied. Feed-forward backpropagation neural network used for classification of characters with two hidden layers. A better Recognition rate of 93.7 percentagenoted.