Two-stage classification approach for the recognition of handwritten atomic Odia character
by Ramesh Kumar Mohapatra; Banshidhar Majhi; Sanjay Kumar Jena
International Journal of Applied Pattern Recognition (IJAPR), Vol. 4, No. 1, 2017

Abstract: This paper presents a two-stage classification mechanism for the identification of isolated (segmented) Odia basic handwritten character. Some preprocessing techniques have been applied for the removal of noise and image enhancement of each sample in the database. The database has been divided into two groups considering some essential characteristics of the shape of each character. A linear support vector machine (SVM) classifier is utilised for this purpose and results in an error rate of 0.0014%. In the second phase, the structural features extracted followed by PCA-based feature reduction technique to derive the discriminant features from each character in the groups for classification. In this phase, two three-layer back propagation neural networks (BPNN) are used to recognise handwritten Odia segmented character in each group. A relative investigation has been made on a sensibly generous dataset with the competent schemes. From test outcomes, we reason that our proposed method outmanoeuvres different schemes on OHCS v1.0 database and provides an overall accuracy of 98.87%.

Online publication date: Sat, 04-Mar-2017

The full text of this article is only available to individual subscribers or to users at subscribing institutions.

 
Existing subscribers:
Go to Inderscience Online Journals to access the Full Text of this article.

Pay per view:
If you are not a subscriber and you just want to read the full contents of this article, buy online access here.

Complimentary Subscribers, Editors or Members of the Editorial Board of the International Journal of Applied Pattern Recognition (IJAPR):
Login with your Inderscience username and password:

Username:        Password:          Forgotten your password?

Want to subscribe?
A subscription gives you complete access to all articles in the current issue, as well as to all articles in the previous three years (where applicable). See our Orders page to subscribe.

If you still need assistance, please email subs@inderscience.com