Driver fatigue detection method based on multi-feature empirical fusion model.
Saved in:
| Title: | Driver fatigue detection method based on multi-feature empirical fusion model. |
|---|---|
| Authors: | Qin, Yanbin1 (AUTHOR) qinyanbin1@aliyun.com, Lyu, Hongming1 (AUTHOR) lhmyg@163.com, Zhu, Kaibin1 (AUTHOR) Zkb1631@163.com |
| Source: | Multimedia Tools & Applications. Jul2025, Vol. 84 Issue 22, p25253-25270. 18p. |
| Subjects: | Facial expression, Drowsiness, Machine learning, Multisensor data fusion, Traffic safety, Real-time computing, Computer vision |
| Abstract: | As the number of long-distance commuters continues to rise, driver fatigue has become a major contributor to traffic accidents, underscoring the critical need for real-time fatigue detection and prevention. Facial expressions serve as crucial indicators that directly reflect the driver's fatigue state, which may vary due to individual differences. This paper presents a comprehensive visual monitoring approach for real-time driver fatigue detection. The proposed method uses a vehicle-mounted camera to extract the driver's facial features through an empirical multi-feature fusion model. By determining appropriate thresholds based on individual driving habits and conditions, the algorithm maps multi-dimensional facial behaviors to corresponding Karolinska Sleepiness Scale (KSS) scores and fatigue levels, accurately identifying four states: awake, mild fatigue, moderate fatigue, and severe fatigue. Evaluated on a dataset of 2,555 validated samples obtained from real-world driving conditions, the method demonstrated an impressive average accuracy of 98.35% during both training and experimentation. Furthermore, a comparative analysis against state-of-the-art fatigue detection algorithms on a self- curated dataset revealed that the proposed approach achieved the highest average accuracy of 98.9%, with lower computational requirements and a lighter weight. The mean average precision (mAP) was 1.6% higher than the lightweight Efficient Det-D2, while also having fewer model parameters and reduced computational complexity. [ABSTRACT FROM AUTHOR] |
| Copyright of Multimedia Tools & Applications is the property of Springer Nature and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.) | |
| Database: | Engineering Source |
|
Full text is not displayed to guests.
Login for full access.
|
|
| Abstract: | As the number of long-distance commuters continues to rise, driver fatigue has become a major contributor to traffic accidents, underscoring the critical need for real-time fatigue detection and prevention. Facial expressions serve as crucial indicators that directly reflect the driver's fatigue state, which may vary due to individual differences. This paper presents a comprehensive visual monitoring approach for real-time driver fatigue detection. The proposed method uses a vehicle-mounted camera to extract the driver's facial features through an empirical multi-feature fusion model. By determining appropriate thresholds based on individual driving habits and conditions, the algorithm maps multi-dimensional facial behaviors to corresponding Karolinska Sleepiness Scale (KSS) scores and fatigue levels, accurately identifying four states: awake, mild fatigue, moderate fatigue, and severe fatigue. Evaluated on a dataset of 2,555 validated samples obtained from real-world driving conditions, the method demonstrated an impressive average accuracy of 98.35% during both training and experimentation. Furthermore, a comparative analysis against state-of-the-art fatigue detection algorithms on a self- curated dataset revealed that the proposed approach achieved the highest average accuracy of 98.9%, with lower computational requirements and a lighter weight. The mean average precision (mAP) was 1.6% higher than the lightweight Efficient Det-D2, while also having fewer model parameters and reduced computational complexity. [ABSTRACT FROM AUTHOR] |
|---|---|
| ISSN: | 13807501 |
| DOI: | 10.1007/s11042-024-20115-z |
