A Kannada Handwritten Character Recognition System Exploiting Machine Learning Approach

Handwritten character recognition plays an important role when the handwritten text on paper, postcards, etc. requires conversion of the handwritten text into digitized form. The difference between a digitized handwritten document and a scanned document is that the prior one can be edited, and the latter cannot. Significant developments have been made on the handwritten character recognition of widely used languages like English. India is a multilingual country where there exist multiple regional languages like Kannada, Tamil, Malayalam and other Dravidian Languages with complex scripts. Kannada is spoken in most of the regions of Karnataka State, which is one of the southern regions of India. In the proposed research, a Convolutional Neural Network(CNN) is practiced to recognize Kannada handwritten Characters. The research employs densely connected-convolutional networks or DenseNet variant of CNN to recognize handwritten Kannada characters. DenseNet is preferred in this research for its known advantages such as enhanced feature propagation, improved feature reuse, and minimized vanishing gradient problem. The dataset used in the experimentation is a standard Char74k dataset. The prime objective of this research is to devise a machine learning based application to recognize Kannada handwritten characters with high accuracy and convert them into digitized characters. Digitized documents promote the growth of several other major applications like speech conversion, language translation and conversion of medieval documents. A testing accuracy of 93.87% is observed for 3285 images of handwritten Kannada characters with 5 images from each of the 657 classes. This machine learning model can also be trained to recognize characters of different Indian languages.