An Improved Fatigue Detection System Based on Behavioral Characteristics of Driver

Gupta, Rajat; Aman, Kanishk; Shiva, Nalin; Singh, Yadvendra · 2017 · arXiv

archive: archived pipeline: cataloged verified

Get this paper ↗ (full text — opens at the source; we link to it, we don't host it)

Summary

This paper addresses the critical safety issue of driver fatigue, a major contributor to the approximately 1.25 million annual global road fatalities reported by the WHO. Existing fatigue detection systems are categorized into three types: vehicle attribute-based (e.g., steering movement), physiological signal-based (e.g., EEG, ECG), and behavioral characteristic-based (e.g., eye blinking, yawning). The authors identify significant limitations in prior approaches, noting that vehicle-based systems are constrained by road conditions, physiological systems require intrusive skin sensors that cause user discomfort, and many commercial behavioral systems trigger abrupt alarms that may startle drivers into dangerous reactions. Consequently, the authors propose a practical, subject-independent, and calibration-free system that monitors driver behavior using non-intrusive computer vision to detect fatigue levels and issue graduated alerts. The proposed system utilizes a camera mounted on the car dashboard to capture real-time video of the driver. The methodology involves four subsystems: video capture, face detection and feature extraction, fatigue detection, and an alert unit. The face detection unit employs the Viola-Jones algorithm with Haar-like features to locate the face, followed by low-light image enhancement and noise elimination to ensure accuracy in varying lighting conditions. The system extracts specific regions of interest—eyes (80x30 pixels) and mouth (40x40 pixels)—from the detected face. To manage high-dimensional data, Principal Component Analysis (PCA) is applied to reduce feature vectors while minimizing information loss. These compressed features are then classified using a Support Vector Machine (SVM), chosen for its efficiency with high-dimensional data and ability to handle both linear and non-linear separability. The SVM outputs a binary classification (+1 for fatigued, -1 for alert). The alert unit processes the SVM outputs through a running sum mechanism to determine the cumulative fatigue level over time. This approach utilizes two threshold levels to distinguish between no/low fatigue and high fatigue. For low fatigue levels, the system triggers a 10-second alarm, after which it re-evaluates the driver’s state. For high fatigue levels, the system initiates more aggressive accident-prevention measures, such as automatically reducing vehicle speed, stopping the vehicle, or deploying a water spray. The design aims to avoid sudden, startling alerts that could cause abrupt driver reactions, instead providing a safe margin for detecting sleep onset. The significance of this work lies in its cost-effective, non-intrusive design that relies solely on a camera, eliminating the need for expensive hardware or uncomfortable physiological sensors. By focusing exclusively on eye and mouth features and utilizing robust image processing techniques, the system optimizes both time and accuracy. The authors conclude that this behavioral-based approach overcomes the drawbacks of previous methods by providing a robust, scalable solution for real-time fatigue monitoring that can effectively prevent accidents through graduated intervention strategies.

Key finding

The authors propose a calibration-free, subject-independent fatigue detector that combines eye and mouth feature extraction with PCA dimensionality reduction and SVM classification, but report no empirical validation, accuracy metrics, or participant data.

Methodology

theoretical

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed. Discovered via discover_arxiv on 2026-05-04 (4 acquisition events logged).

Summary generated by qwen3.6-27b-prismaquant on 2026-06-10; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• drowsiness detection algorithms
• distraction detection algorithms

Information type

What kind of knowledge this paper contributes, grouped by family — independent of topic (what it is about) and method (how it was studied).

• Empirical Findings: physiological data
• Methodological Resource: tool software