Work

Sleepy Drivers: Eye-state and Drowsiness Monitoring Using Face Video

Year: 2023
Type: dissertation
Abstract: Driver drowsiness has been identified as a major contributor to road accidents, accounting for up to 20% of all incidents, often resulting in fatalities and severe injuries. Even though detecting and mitigating it is crucial to improving road safety, existing datasets for drowsiness detection are limited in complexity. This dissertation addresses this critical need by undertaking two distinct processes. In the primary development process, machine learning techniques are employed to analyze face video data, with a particular focus on the eye region as an indicator of drowsi- ness. The secondary process centers on the proposal of a comprehensive data acquisition protocol, capturing multimodal data streams coupled with personal and idiosyncratic factors. The eye-state classifier achieved an accuracy of 73.75%, and a manual eye-state la- beling approach led to a binary classification model of drowsiness with 80.00% accuracy and 100.00% recall for the drowsy class. In conclusion, this work intended to contribute to future studies on effective drowsy driving alert systems based on eye-state. To further enhance this research, increasing the dataset size and subject variability is key. As to new research directions, expanding into interpretability and causality studies based on the proposed data acquisition protocol holds promise for advancing road safety initiatives.
Language: English
Field: Psychology
Open Access status: green
OpenAlex ID: W7005845847
Cites: 89
Cited by: 1
Related to: 10