engleski

An Algorithm for the Recognition of Handwritten Symbols

The paper presents an algorithm for the recognition of handwritten symbols in a case when the lines drawn by the user can be presented as a well-defined data structure (e.g. an array). This typically occurs when writing with the mouse or writing pen. Symbols have to be predefined in a learning process that precedes the recognition. The recognition algorithm can be subdivided into three phases: i) scanning of the entered symbols, ii) adjoining of the separate lines into shapes, iii) recognition of the shapes. The last phase of the recognition consists of the formatting of the shapes into standardized forms, and of comparison of the standardized form bitmap to bitmaps in the pattern database. By providing more than one sample pattern for each symbol, the percentage of successful recognition can be significantly improved. The algorithm is tested in various ways, and compared to some other known solutions of the problem. It can be considered as producing good recognition results, while being relatively simple to implement. Some limitations of the algorithm in its present state are discussed, as well as the possibilities for its further improvement. The program is implemented as a self-standing application, and as a Web-based application accessible via Internet.

symbol recognition; recognition of handwritten symbols; algorithm

The paper was presented on the conference, it was included in the contents of the Conference Proceedings, and its abstract appeared in the printed Conference booklet (the scanned pdf file that is placed instead of the full text). However, due to the omission of editors, the paper was not included in the CD edition of the Conference Proceedings.