Multi-modal driver drowsiness detection: A feasibility study

Abstract

Driver drowsiness is a significant contributing factor to road accidents and can lead to severe physical injuries, deaths, and significant economic losses. Earlier research primarily concentrated on estimating the level of drowsiness using a single measure such as lane or steering monitoring, behavioral measures or physiological ones. Hybrid systems employing multiple measures provide more reliable solutions because they minimize the number of false alarms and maintain a high recognition rate, which promote their acceptance. In this paper, we present an unobtrusive system that combines a dashboard-mount camera alongside with a watch-like wearable system recording optical (PPG) signals. Statistical analysis performed on 17 subjects demonstrates that our hybrid drowsiness detection system, which leverages both physiological and behavioral measures can reliably determine the drowsiness level with an accuracy ranging from 86% to 96% depending on the number of levels and the classification method used.