Handwritten Character Recognition System using Deep Learning Models for Tamil Language

Abstract

Just like its countless achievement in resolving various computer image processing-based problems, the deep learning presented novel end-to-end tactic to many Handwritten Character Recognition (HCR) with very optimistic outcomes for many languages in recent years. Convolutional Neural Networks (CNN) are the most efficient technique adopted in computer vision and character identification with automatic feature extraction technique. This proposed research work recommends an innovative handwritten Tamil character identification system using AlexNet model. In Tamil linguistic, the multiple letters have alike features, therefore identifying exact letter is an exciting mission. This research work offered that, this deeper architecture can assistance for Tamil Handwritten Character Recognition (THCR) a lot to attain greater performance, in the meantime can be designed with less parameter. 12 primary, 18 consonants and other compound characters, totally 156 classes of Tamil character of Tamil language are recognized with better classification rate and attained nominal processing time using this model.