提出了一种基于流形保持投影的驾驶疲劳识别方法.利用光流技术计算人脸皮层的运动速度,并以此作为疲劳特征;为了有效地进行疲劳特征提取(即特征降维),在保局投影的基础上,将数据的非近邻信息引入目标函数中,提出了流形保持投影方法,有效地保持了疲劳数据的局部流形结构和全局流形结构,同时利用格拉姆—施密特正交化过程解决了保局投影非正交问题.实验结果表明该方法具有很好的识别效果.
A new method for driver fatigue recognition based on manifold preserving projections(MPP) is presented. Facial velocity information,which is determined using optical flow techniques,is used to characterize fatigue.The manifold preserving projections are proposed to extract the effective fatigue feature(viz.dimensionality reduction).The method incorporates the non-locality information of data into the objective function of locality preserving projection(LPP),which preserve the local and global structure of the data manifold,and Gram-Schmidt orthogonalization is used to solve non-orthogonal problem of LPP.The experiments show that the proposed method attains a satisfactory recognition effect.