Description of Invention
Drowsiness is thought to be the cause of 20% of all road accidents. Therefore, this study proposed a behavioural-based driver drowsiness detection system using deep learning approaches. PERCLOS, eye aspect ratio, mouth aspect ratio, yawning, blink duration, and frequency were extracted from the video frames. After that, 2 deep learning models: Artificial Neural Network (ANN) and Long Short-Term Memory (LSTM) were used to classify the driver's drowsiness. The results demonstrated that ANN had an 89% accuracy with a faster training time compared to LSTM. This study is anticipated to minimise road deaths, particularly those caused by sleepy drivers.