Mahidol University Logo
Faculty of ICT, Mahidol University
 

Admissions

Printable Version

 

THAI HANDWRITTEN CHARACTER RECOGNITION SYSTEM (THW-CR) IMPROVING FEATURE EXTRACTION PROCESS BY THE ANALYSIS OF CONTOUR CHARACTERISTICS

 

TITLE THAI HANDWRITTEN CHARACTER RECOGNITION SYSTEM (THW-CR) IMPROVING FEATURE EXTRACTION PROCESS BY THE ANALYSIS OF CONTOUR CHARACTERISTICS
AUTHOR URAIRAT LIMKONGLAP
DEGREE MASTER OF SCIENCE PROGRAMME IN COMPUTER SCIENCE
FACULTY FACULTY OF SCIENCE
ADVISOR JARERNSRI L. MITRPANONT
CO-ADVISOR SUKANYA PHONGSUPHAP
 
ABSTRACT
This research proposes an approach to the analysis of contour characteristics for the feature extraction process of an off-line Thai handwritten character recognition program (THW-CR). The program aims at increasing the accuracy of both feature extraction rate and character recognition rate. We introduce the contour to capture the movement of features. The start contour point is initiated from the location of the region. Then, it is chained along the feature shape until it meets the termination of the region. Subsequently, the contour points along the contour are retrieved and fit by the fit function which is compatible with the feature shape. The characteristic of the fit function is finally interpreted to feature type. The research framework consists of four major components: preprocessing, feature extraction process, recognition and postprocessing. However, this work focuses on the part of feature extraction, which consists of three parts: feature extraction algorithms, feature representation and feature conflict resolution rules. Specifically, the feature representation is the main focus for the improvement in which the contour is established to represent a more detectable structure and accurate feature of the character. In addition, the feature conflict resolution rules are proposed to adjust the conflict features of either position or feature type compared with the possible structure. For the implementation part, the prototype of the THW-CR was developed to validate our proposed solution. All proposed techniques were integrated into the HWR system, which was the prototype of the original work extended by this research. The 230 documents written by 32 people and containing 103,000 characters, approximately, were tested. The results were evaluated by two evaluation criteria: Feature Extraction Rate (FER) and Character Recognition Rate (CRR). The FER was tested to verify the accuracy of detected features of each character while the CRR was tested to measure the accuracy of recognized characters of the whole paragraphs or sentences. From the experiment consisting of three test scenarios of the classification of writing types, the average FER was 95.96% and the maximum FER reached 97.73%. Meanwhile, the average CRR was 90.76% and the highest CRR reached 95.35%. Additionally, the result of the overall evaluation showed that the THW-CR system generated reliable results for improving the accuracy of the feature extraction rate and the character recognition rate compared with the previous research by 3.62% and 8.33%, respectively.
KEYWORD THAI HANDWRITTEN CHARACTER RECOGNITION / CONTOUR ANALYSIS / CONTOUR FITTING / CURVE FITTING

 

Go to Top

 

ICT Building, Mahidol University, 999 Phuttamonthon 4 Road, Salaya, Nakhonpathom 73170 Tel. +66 02 441-0909 Fax. +66 02 849-6099
Mahidol University Computing Center, The Faculty of ICT, Mahidol University , Rama 6 Road, Rajathevi, Bangkok 10400 Tel. +66 02 354-4333 Fax. +66 02 354-7333