Title: Classification of handwritten Odia basic character using Stockwell transform

Authors: Ramesh Kumar Mohapatra; Banshidhar Majhi; Sanjay Kumar Jena

Addresses: Department of Computer Science and Engineering, National Institute of Technology, Rourkela, 769008, India ' Department of Computer Science and Engineering, National Institute of Technology, Rourkela, 769008, India ' Department of Computer Science and Engineering, National Institute of Technology, Rourkela, 769008, India

Abstract: In this paper, we have proposed a scheme for the recognition of handwritten basic isolated Odia characters. Our method utilises chain code histogram (CCH) to split the overall dataset into two different groups. Preprocessing is carried out for noise removal and enhancement of the character images in each group. Discrete orthonormal S-transform (DOST) features are extracted followed by a PCA-based feature reduction scheme to derive discriminant features to be used by the ANN for classification. Comparative analysis has been performed on a reasonably large dataset with the competent schemes. From experimental results, we conclude that our proposed scheme outperforms other schemes on the dataset that we have designed and provides an overall accuracy of 98.55%.

Keywords: Odia script; chain code histogram; CCH; discrete orthonormal S-transform; DOST; principal component analysis; PCA; artificial neural networks; ANNs; classification; handwritten characters; Stockwell transform; character recognition.

DOI: 10.1504/IJAPR.2015.073854

International Journal of Applied Pattern Recognition, 2015 Vol.2 No.3, pp.235 - 254

Received: 22 Jun 2015
Accepted: 29 Jul 2015
Published online: 24 Dec 2015 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article