An Analysis of Driver Drowsiness Detection Using Electromyography (EMG) Facial Muscles

Drowsiness during driving can lead to fatal vehicle crashes and deaths. Facial responses when drowsy such as eyes closures and yawning are useful signals to detect driver drowsiness. This project aims to analyze the effectiveness of using electromyography (EMG) signals from the facial muscles (Masseter and Orbicularis Orris) to detect driver drowsiness. 12 healthy male participants took part in the data collection process by driving for 1 hour in a driving simulator where the EMG electrodes were placed on the targeted facial muscles. 7-time domain features were extracted from the raw EMG and k-Nearest Neighbor (kNN) classifier was used as the signal processing model to detect driver drowsiness. The highest classification accuracy for two classes (drowsy and non-drowsy) problem achieved is 85.71%. This result acquired by setting these parameters; i) 70:30 training-test data ratio, ii) the number of neighbors, k values set to 2 or 4, iii) all seven-time domain features, iv) and both Masseter and Orbicularis Orris muscles. The study concludes that the facial muscles are a good predictor to detect drowsiness in drivers during simulated driving.