CAJ | 학술논문

利用Kinect获取人脸红外图像数据及人脸网格信息，通过眼部区域的Haar-like特征值进行眼睛状态识别，研究不同Haar-like特征对人眼状态的敏感度，以及头部旋转时眼睛区域的选取范围进行分析。实验共采集39名测试者的765个眼睛状态作为训练集和测试集，通过实验列出不同特征和组合特征的识别结果，实验结果表明构建的系统可以检测眼睛的不同状态，达到实时检测人眼状态的要求。
이용Kinect획취인검홍외도상수거급인검망격신식，통과안부구역적Haar-like특정치진행안정상태식별，연구불동Haar-like특정대인안상태적민감도，이급두부선전시안정구역적선취범위진행분석。실험공채집39명측시자적765개안정상태작위훈련집화측시집，통과실험렬출불동특정화조합특정적식별결과，실험결과표명구건적계통가이검측안정적불동상태，체도실시검측인안상태적요구。
This paper presents a new way to detect eye status using facial infrared image and grid from Kinect sensor, and uses haar-like eigenvalue of eye region to identify eye status. The sen-sitivity of human eye status is researched according to different haar-like features, and eye re-gion is analyzed by different pose of head. In this research, the dataset takes 765 eye status of 39 samples as training set and test dataset respectively. Finally, this test presents recognition re-sults by the different features and their combinations features. Experimental results indicate that this system could apply to real-time detect different eye status.