Eye Tracking in Driving Environment Based on Multichannel Convolutional Neural Network
Xiaolei Cha, Xiaohui Yang, Yingji Zhang, Zhiquan Feng, Tao Xu, Xue Fan · 2020
Gaze is the most important way for human to obtain information from the outside world, and it is the most direct and significant cue to analysis human behavior and intention. In driving environment, eye tracking is usually applied to model driver's fixations and gaze allocations, which is important in advanced driver assistance system (ADAS). In this paper, we have proposed a new eye tracking method in driving environment, which is based on multichannel convolutional neural network. Firstly, we establish the dataset for driver's eye tracking, which includes the left eye region image, the right eye region image and the face region image. After that, the multi-channel convolutional neural network is training using the dataset. Finally, the driver's gaze zone will be estimated using the pre-trained network. Experimental results show that the accuracy of the proposed method is 94.60% for seven gaze zone estimation, and it can be used in ADAS to analysis the driver's behavior and detect driver distraction.