A Driving Fatigue Dataset of Driver's Facial Features and Heart Rate Features

Aiming at the problems of single eigenvalue type and not-real fatigue evaluation in the current published driving fatigue dataset, and conforming to the research development trend of facial and heart rate features fusion, this paper studies a set of simulated fatigue driving experiment scheme using low intrusive camera and smart watch as fatigue information collection equipment, and combined with self-evaluation and expert-evaluation of fatigue state. According to this experiment, a fatigue driving dataset including driver's facial features, heart rate features and more realistic fatigue evaluation is obtained. Facial features are five categories of 51 dimensions, including head posture, gaze direction, facial micro expression, eye opening and mouth opening. Heart rate features include 7 dimensions such as instantaneous-HR, mean-HR, SDNN, RMSSD, LFP, HFP and LFP/HFP. The dataset is composed of 2400 sample data of 20 experimenters, and the training set and test set are randomly divided into 7:3, which can be used for the training and test of the corresponding driving fatigue detection model.