Microcontroller-implemented artificial neural network for electrooculography-based wearable drowsiness detection system

Abstract

Various methods have been explored to develop an effective drowsiness detection system to give drivers a warning of impending drowsiness. The present work has successfully developed an electrooculagraphy-based wearable drowsiness detection system in the form of a visor cap by implementing an artificial neural network (ANN) into an Arduino LilyPadUSB microcontroller. As a result, a stand-alone and wearable system that does not require a computer was achieved. The performance of the system for drowsiness detection has an overall accuracy of 90.00%, precision of 88.00%, sensitivity of 91.67% and a training mean squared error (MSE) of 2.70 × 10−3