System and Method for Driver Drowsiness Detection Using Behavioral and Sensor-Based Physiological Measures

Area of Science:

• * Road safety and artificial intelligence
• * Human-computer interaction
• * Biomedical engineering

Background:

• * Driver drowsiness is a significant cause of road accidents, leading to injuries and financial burdens.
• * Existing drowsiness detection methods, either intrusive (physiological) or non-intrusive (behavioral), often lack accuracy and generalizability.
• * Single-measure approaches have proven insufficient for reliable driver drowsiness detection.

Purpose of the Study:

• * To propose and evaluate a hybrid model for driver drowsiness detection.
• * To combine non-intrusive behavioral measures with intrusive physiological measures for enhanced accuracy.
• * To address the limitations of single-approach methods in detecting driver drowsiness.

Main Methods:

• * A hybrid model integrating AI-based Multi-Task Cascaded Convolutional Neural Networks (MTCNN) for facial feature analysis (non-intrusive).
• * Galvanic Skin Response (GSR) sensors to collect physiological skin conductance data (intrusive).
• * Model efficacy evaluated in a simulated driving environment.

Main Results:

• * The proposed hybrid model achieved 91% efficacy in detecting the transition from awake to drowsy states.
• * The combined approach demonstrated capability in identifying drowsiness under all tested conditions.
• * Integration of behavioral and physiological measures significantly improved detection accuracy.

Conclusions:

• * The hybrid model offers a robust and accurate solution for driver drowsiness detection.
• * Combining non-intrusive and intrusive measures is effective in overcoming limitations of single-method approaches.
• * This research contributes to improved road safety by providing a reliable drowsiness detection system.