应用于智能轮椅控制的头部姿态识别  

Head gesture recognition applied to intelligent wheelchair

作　　者：翁磊[1] 沈天飞[1] 张贺[1] 

机构地区：[1]上海大学机电工程与自动化学院,上海200072

出　　处：《电子测量技术》2013年第12期45-48,59,共5页Electronic Measurement Technology

摘　　要：针对适用于肢体残疾人士的智能轮椅控制,提出了一种基于侧面人脸的头部姿态识别方法。首先利用肤色分割和边缘分割相结合的方法找出人脸区域,然后计算人脸连通区域的质心坐标和面积,以标准正向姿态时的人脸区域质心坐标和面积为参考值,根据活动人脸连通区域质心坐标与参考点的相对位置、欧式距离及面积比特征判别头部姿态。实验结果表明,该方法可以很好的识别向左倾斜、向右倾斜、抬头、低头等头部姿态,平均识别率为92.2%。A head gesture recognition method based on the face profile of the intelligent wheelchair for the limb disabled people is proposed. First, the skin color segmentation and edge segmentation method is combined to detect the face region to calculate the centroid position and the area of the connected region of the face. The centroid position and area of the face region in the standard forward gesture is set as the reference value. The head gesture is determined on the euclidean distance and the area ratio between the centroid position of the moving face connected region and that of the reference point. Experimental results show that the head gestures like tilted to the left , tilted to the right, turn up and turn down can be well recognized by this method. The average recognition is 92.2%.

关 键 词：头势 智能轮椅 肤色分割 人脸检测 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]