Exploiting Features From Triangle Geometry For Digit Recognition

• Main Authors: Azmi, Mohd Sanusi, Nasrudin, Mohammad Faidzul, Omar, Khairuddin, Che Wan Ahmad, Che Wan Shamsul Bahri, Wan Mohd Ghazali, Khadijah
• Format: Conference or Workshop Item
• Language: English
• Published: 2013
• Online Access: http://eprints.utem.edu.my/id/eprint/11816/1/06689658.pdf; http://eprints.utem.edu.my/id/eprint/11816/; http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=6689658&tag=1
• Institution: Universiti Teknikal Malaysia Melaka

Triangle is a basic geometry. There are six type of triangle, but scalene triangle was chosen to be used in this research that based on coordinates of corners generated by our proposed algorithm. In this paper, nine features are proposed where six features were derived from coordinates and sides of triangle. Another three features are angle of corners. After features are identified, image will be zoned into 25 zones. The zoning processes are based on Cartesian plan, Vertical and Horizontal zones. From the zoning, from nine features will become 225 features. The features proposed will be used to HODA, MNIST, IFHCDB and BANGLA datasets. Experiments will be conducted using supervised learning that are Support Vector Machine (SVM) and Multi-layer Perceptron (MLP). Results from the experiments will be evaluated with different Cost (c) for the SVM and Learning Rate (LR) for the MLP. Then, the result will be compared to state of the art by other researches.