Thai Handwritten Segmentation Using Proportional Invariant Recognition Technique

Abstract

Handwritten character recognition has been an on-going research in many languages for many years. Most techniques focus on isolate recognition. However, if a writer writes at fast speed, two or more characters can be written within one stroke. This input cannot be used to recognition. Previously I present the method to separate two characters connected Thai handwriting. This paper presents the method to separate connected Thai handwriting based on a modified water-reservoir and Proportional Invariant Recognition technique to separate two or more characters. The proposed technique find candidate segmentation points (include correctly segmentation points) at 92.8% and segment correctly at 87.6%.