画像情報を利用した指文字の認識

概要

論文の詳細を見る
This research is aiming to develop the recognition system of the "sign language" which is one of the communications with the deaf-and-dumb and the persons without a handicap. As the first phase, this paper considers constructing the algorithm that translates "manual kana" into Japanese. The recognition algorithm is built and verified by recognizing actually about 41 characters which are stationary without gesture. Its algorithm adopts the regularization of the size of subject's hand image which is input into a computer, and the classification tree for feature data. And manual kana "て" of each subject is used for the regularization as the standard image, which set up the recognition range and the various conditions for extracting feature data. The recognition of the images of 41 kinds of manual kana, for respectively 20 subjects, has brought a recognition rate of an average of 77%.
拓殖大学の論文
2004-10-20