基于Kinect的人脸眼部状态实时检测  

Real-time Detection of Facial Eye Status Based on Kinect Sensor

作　　者：朱真真[1,2] 王伟[1] 段晓东[1] 王存睿[1] 

机构地区：[1]大连民族学院大连市民族文化数字技术重点实验室,辽宁大连116605 [2]北方民族大学计算机科学与工程学院,宁夏银川750021

出　　处：《大连民族学院学报》2015年第1期81-84,共4页Journal of Dalian Nationalities University

基　　金：国家自然科学基金(61370146);辽宁省科技计划项目(2013020026);国家民委项目(gm-2009-66)

摘　　要：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。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.

关 键 词：人眼状态 实时检测 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]