Handwritten Tamil Character Recognition in Palm Leaf Manuscripts using BiLSTM Classifier

Abstract

Recognizing and reading the Tamil characters which were written many centuries ago in the palm leaf manuscripts is a tough task. As the character has taken a different form over the centuries, the shape of the characters has not been known to contemporary Tamil readers. Neural network classifiers are the tool to unlock the treasure of knowledge paralyzed for such reasons. Only those who are familiar with the shape and strokes of Tamil characters can read palm leaf manuscripts. The knowledge gained from the Tamil literatures have learned about the palm leaf manuscripts has been prepared through computer so that all the people can know the written ideas. This method works in three stages like text line segmentation, character segmentation, character recognition. The final stage of recognition has done by the Bidirectional Long Short Term Memory (BiLSTM) classifier that produces a better result than other conventional CNN methods in Tamil character recognition.