Preventive Detection of Driver Drowsiness from EEG Signals using Fuzzy Expert Systems

Abstract

Currently, the percentage of traffic accidents has increased, and according to statistics, this percentage will continue to increase every year, so it is necessary to develop new technologies to prevent this kind of accidents. This paper presents a drowsiness detection system based on electroencephalogram (EEG) signals using a pair of channels (Fp1 and Fp2) applied to drivers before entering their vehicles. First, this model detects the relationship between the area under the curve (AUC) of alpha brain waves, an effective parameter for detecting drowsiness. Then, the extracted information is passed to a fuzzy expert system (FES) that classifies the subject's state as "alert" or "sleepy"; the criterion used was a threshold and training with subjective levels. The proposed system was compared with neural network models, such as support vector machine (SVM), K nearest neighbors (KNN), and random forest (RF). Measurements of one hundred and twenty minutes were performed on each of the ten drivers for two days to test the system. The tests confirm that this system is suitable for preventive measures and that the fuzzy system is superior to traditional neural network methods.