A New Method to Detect Driver Fatigue Based on EMG and ECG Collected by Portable Non-Contact Sensors

Recently, detection and prediction on driver fatigue have become interest of research worldwide. In the present work, a new method is built to effectively evaluate driver fatigue based on electromyography (EMG) and electrocardiogram (ECG) collected by portable real-time and non-contact sensors. First, under the non-disturbance condition for driver’s attention, mixed physiological signals (EMG, ECG and artefacts) are collected by non-contact sensors located in a cushion on the driver’s seat. EMG and ECG are effectively separated by FastICA, and de-noised by empirical mode decomposition (EMD). Then, three physiological features, complexity of EMG, complexity of ECG, and sample entropy (SampEn) of ECG, are extracted and analysed. Principal components are obtained by principal components analysis (PCA) and are used as independent variables. Finally, a mathematical model of driver fatigue is built, and the accuracy of the model is up to 91%. Moreover, based on the questionnaire, the calculation results of model are consistent with real fatigue felt by the participants. Therefore, this model can effectively detect driver fatigue.