Head pose estimation based on feature extraction, fuzzy C-means and neural network for driver assistance system
Le Zhang, Danhong Zhang, Yixin Su, Chao Wang · 2014
The driver's head pose plays a very important role in risk prediction of vehicle driving. Head pose affects the driver's ability to observe the driving environment, and determines the safety in the process of driving. This paper proposes an efficient representation and feature extraction technique for head pose estimation of the driver. This method firstly applies SIFT algorithm to extracting the feature points of the driver's head (face mostly). Then using fuzzy c-means algorithm to analyze the images contained feature points and calculate the clustering centers. Finally taking the nonlinear regression method based on neural network to map the data to the linear separable space, and the results of the nonlinear regression were carried out to estimate the head pose. The experimental results show that this method can well estimate driver's head pose, reduce the generalization error, and it has a strong practicality in the actual driving assistant system.