A thesis submitted in partial fulfillment of the requirements for the degree of Master of Engineering in Computer Science, School of Engineering and Technology

Thesis (M.Eng.) - Asian Institute of Technology, 2009

Sketching on paper is a simple method to design an idea, to communicate with other people by mixing words with drawings and to compress a lot of information into a useful image. The strong advantage of sketching is convenient, fast and flexible to create. In order to extend the capability of this advantage, the sketching interfaces were implemented for storing information into digital format which is able to search and retrievable. The simplest solution for keeping all different sketching information is to record in a flat image but this approach leads to inefficiently retrieval because the system can not understand the meaning of an image. A better solution is to separate the sketching into three types such as model, handwriting and annotation. The purpose of this thesis is focusing on annotation recognition in medical patient records. The process of recognition is divided into two steps, first is to convert strokes into graph for analyzing them as low-level information. The second, the specific-domain recognizer, uses the power of constraint logic programming to match the configuration of strokes with annotation configuration. The prototype was implemented to prove this approach. The experiment on the set of general medical annotations demonstrates the flexibility for adding new symbols, fast recognition and robustness to order, orientation, and size of drawing.