An intelligent system to detect drowsiness at the wheel

Abstract

Drowsiness is one of the causes of road accidents with higher fatality rate and even though it has been studied over the years, there is still no solution that can mitigate this problem from a low cost perspective. Thereby, artificial intelligence will be used to predict drowsiness and develop a non-intrusive and low-cost system using a wearable device that is capable of alerting the driver before they exhibit any signs. The aim of this work is to present the results achieved in the first stage, where driving simulations were conducted and video information, subjective reporting and physiological-based data were collected. Drowsiness levels and eye blinks were extracted from analysis of the participants' videos, in addition to simulation based information, such as speed variations and road accidents, allowing to classify the driver's state. Multivariate statistical process control was the method implemented, considering the Heart Rate Variability (HRV) and Electroencephalography (EEG) information. Using this methodology, it was able to detect the transition between the different drowsiness phases. Although the results are promising, there are still missing analyses, such as the application of Machine Learning (ML) techniques to classify the drowsy state and identify the best subset of features capable of detecting this state. Besides this, data analysis must be done to understand how drowsiness can be predicted. Finally, the proposed solution must be implemented in a real environment.