Handwriting Recognition Using Active Contour

Abstract

Handwriting recognition is the process of recognizing handwritten or printed alphabets in various documents and even old manuscript. This method also helps in preserving the texts and writings of ancient times. This paper aims to represent the work related to recognition of cursive handwriting of different languages. Cursive handwritings are connected to each other, hence segmentation is required. Segmentation is used for extraction: line segmentation to extract sentence, word segmentation to extract words and character segmentation to extract individual letters. The segmentation involves dividing based on contours by setting different kernel size. For classification, we have used classifiers—convolution neural networks. We carried our experiment using datasets collected from E-MNIST, UCI. The experimental accuracy for the E-MNIST dataset is 79.3% and for UCI Devanagari dataset is 93%.