A Novel Approach for Handwriting Recognition in Malayalam Manuscripts using Contour Detection and Convolutional Neural Nets

Neural Networks is a hot area of research for various kind of pattern recognition. Hand writing recognition is a domain coming under the field of pattern recognition which had captured a high research interest during the last 10 years. It is a complex process mainly due to the huge character set, complexity and similarity of Malayalam letters. Our project aims to digitalize Malayalam handwritten text from Malayalam manuscripts like palm leafs, official documents in Government offices etc. An intensive literature survey was conducted on the various methods of handwriting recognition in various languages to find a best suitable approach for digitalizing Malayalam language. A novel approach to digitalize Malayalam manuscripts using contour detection for segmentation and Convolutional Neural Network for classification was proposed. The system was trained using the samples collected from museums and a neural network is constructed for character identification. According to the results obtained the system is proved to have less overhead without compromising the accuracy.