| Author | Chomtip Pornpanomchai |
| Call Number | AIT Diss. no.CS-00-03 |
| Subject(s) | Pattern recognition systems
|
| Note | A dissertation submitted in partial fulfilment of the requirement for the degree of Doctor of Technology Science, School of Advanced Technologies |
| Publisher | Asian Institute of Technology |
| Series Statement | Dissertation ; no. CS-00-03 |
| Abstract | An automatic handwritten character recognition system is one of the most difficult tasks in image processing. A handwritten character recognition task has two main problems, an unlimited handwritten character style and an uncertain handwritten character geometric shape. This thesis uses a feature-based, fuzzy logic and object-oriented approach (FBFLOOA) to recognize Thai handwritten characters. A feature-based approach is used to extract Thai handwritten character features. Fuzzy logic technology is used to justify the uncertain geometric shape of a handwritten character. An object-oriented approach is used to analyze and design a Thai handwritten character recognition model. The study concludes that all Thai handwritten characters consist of three basic geometric shapes or basic sub from features, a line, an arc and a circle. These three basic sub from features are connected with an appropriate coordination or precision feature to construct a Thai character. A basic sub from feature is a terminal node of a form feature hierarchy that inherits some properties from their super-classes. In term of form feature hierarchy and properties in heritance, it implies that an object-oriented approach is an effective approach that can analyze and design a Thai handwritten character recognition model. Moreover, The study uses the Unified Modeling Language (UML) standard diagrams, which are 1) a use case diagram, 2) class-relationship diagram, 3) dynamic diagram, 4) activity diagram, 5) sequence diagram and 6) collaborate diagram to represent a Thai handwritten character model. The study tested five hundred Thai handwritten words, which consist of four thousand one hundred and fifty one characters. The experiment was divided into two phases. The first phases used only the FBFLOOA to recognize a handwritten character and the second phases used the FBFLOOA combined with a Thai dictionary file to search for correct answer for a non-recognition character. The first phases experimental results demonstrated a recognition accuracy of 91.21%, 7.51% non-recognition and 1.28% mismatch. The second phases precision results were 98.21%, 0.51% non-recognition and 1.28% mismatch. On everage, both phases had a recognition speed of 6.71 seconds per character. The FBFLOOA executed program size was 189 KB and the Thai dictionary file was 853 KB. The small size of the Thai handwritten character recognition program is suitable for portable computers, non-keyboard calculators and handheld computers. Based on the results, this study concludes that the FBFLOOA is an effective handwritten character recognition approach, which can be applied to identify not only Thai handwritten characters but also the handwritten characters of any language because all language characters consist of form and precision features. |
| Year | 2000 |
| Corresponding Series Added Entry | Asian Institute of Technology. Dissertation ; no. CS-00-03 |
| Type | Dissertation |
| School | School of Advanced Technologies (SAT) |
| Department | Department of Information and Communications Technologies (DICT) |
| Academic Program/FoS | Computer Science (CS) |
| Chairperson(s) | Batanov, Dentcho N.; |
| Examination Committee(s) | Sadananda, Ramakoti;Vatcharaporn Esichaikul;Dimmitt, Nicholas;Burrell, Phillip; |
| Scholarship Donor(s) | Mahidol University; |
| Degree | Thesis (Ph.D.) - Asian Institute of Technology, 2000 |