Recognition of online handwritten Gurmukhi characters using recurrent neural network classifier

Handwriting recognition is one of the challenging tasks in the area of pattern recognition and machine learning. This paper presents the recognition of Online handwritten basic characters of Gurmukhi, an Indian script used by more than 100 million individuals. There are 41 basic characters (i.e., Consonants) in Gurmukhi and we have used a primary dataset of 52,570 Gurmukhi words, written by 175 different individuals. A total of 81 stroke-classes have been identified to represent the 41 basic characters. The handwritten data is stored and annotated at stroke-level. In order to train the classifier, we have taken 150–170 average samples of each identified stroke-class. After performing the training/testing experiments, an accuracy of 98.67% has been achieved for stroke classification using RNN classifier and 90.93% obtained as testing accuracy.