基于预测型人脸跟踪的人机接口

Human Face Predicting Based on Tracking of Man-machine Interface

实现了一种新颖人机接口方法。首先,利用Adaboost算法进行快速人脸检测和定位,然后通过Canny滤波获取精确的人脸区域建立肤色直方图模型。其次,使用Camshift算法计算人脸跟踪窗口的位置和大小。然后用二维Kalman滤波预测人脸的位置,增加跟踪速度。最后,人脸的三维自由度运动分别被用作控制电脑游戏角色的6种运动。通过实验,验证了此方法的实用性和有效性。

This paper achieves a novel method of man-machine interface.Firstly, Adaboost algorithm is used to detect and focus on the face area. Then a histogram model is constructed after the more precise area of the face was got due to the Canny filter.Secondly,Camshift(Continuously Adaptive Mean Shift) estimates the center and size of human face tracking window.To increase the tracking speed,a two-dimensional Kalman filter(stochastic estimator) is used to track the object region centroid(human face).Finally,the three degrees of freedom of human face are used to control over the role of computer games in six sports.The experimental results show that the method is robust and efficient.