초록

용어보기논문에서 용어와 풀이말을 자동 추출한 결과로, 시범 서비스 중입니다. AI-Helper

Abstract ▼ AI-Helper

In this paper, we describe a method for off-line recognition of Hangul handwritten in sammool style with statistical feature extraction scheme. In order to make the best use of characteristics of Hangul which consist of straight line components for the most part, directional components-horizontal, ...

In this paper, we describe a method for off-line recognition of Hangul handwritten in sammool style with statistical feature extraction scheme. In order to make the best use of characteristics of Hangul which consist of straight line components for the most part, directional components-horizontal, vertical, right up diagonal and left up diagonal lines - were extracted from input pattern. Then, in order to extract feature vectors which are invariant to touching stroke variations, the mesh was determined dynamically according to the density of black pixels. As this type of feature extraction methods do not need the size normalization and/or the thinning process, and as the number of classes to be matched in the classification process is significantly reduced by the restrictions in writing style, very high speed of processing and recognition rate can be obtained. In order to verify the performance of the proposed method, experiments with 2,350 classes of handprinted Hangul in sammool style were performed, and more than 90% of correct recognition rate and 0.17 second per one character on the PC 486(33MHz) of recognition speed were obtained. These results reveal that the development of a practical high speed OCR system is feasible.