A fuzzy and rough sets approach for recognition of handwritten Thai characters

  • Authors:

  • Pisit Phokharatkul;Darunee Chatchawalanonth;Chom Kimpan

  • Affiliations:

  • Department of Computer Engineering, Faculty of Engineering, Mahidol University, Thailand;Faculty of Information Technology, Rangsit University, Patumtani, Thailand;Faculty of Information Technology, Rangsit University, Patumtani, Thailand

  • Venue:
  • ISPRA'06 Proceedings of the 5th WSEAS International Conference on Signal Processing, Robotics and Automation
  • Year:
  • 2006

Quantified Score

Hi-index 0.00

Visualization

Abstract

This paper presents both fuzzy logic and rough logic as applied to Thai character recognition. Both fuzzy and rough sets have been introduced as tools to deal with vagueness and uncertain data in artificial intelligence applications. They both attack the problem of impreciseness, but in different ways. Whereas fuzzy logic is concerned with vagueness, rough logic concerns itself with indiscernability. Hence rough logic views impreciseness as lack of knowledge and not as a feature of the problem itself. In a recognition system, the rough set theory can only deal with discrete values. This means that the value attributes need to be discrete before they are provided to the rough set theory. Then attempt were to use the concept of fuzzy as a tool of discretization of the features of handwritten Thai characters. The features used for the method are heads of Thai character (loop contours), end points, peripheral shape features, stroke density features, width-height ratio, feature code, number of island, characteristic head, character ripple, and chain code respectively. Rough logic offers a variety of desirable features, most notably the ability of reduction of knowledge. This enables us to be able to extract the essential information from a given rule base, and thus it offers a simple way of obtaining recognition with a minimum set of rules. The system is tested with unknown samples, composed of 53,400 handwritten Thai characters. Experimental results have shown that both fuzzy logic and rough logic are powerful tools in successfully classifying handwritten Thai characters. The recognition rate by this method is about 84%.