Wearable Driver Distraction Identification On-The-Road via Continuous Decomposition of Galvanic Skin Responses.

Area of Science:

• Human-computer interaction
• Physiological computing
• Road safety engineering

Background:

• Distracted driving is a major cause of fatal road accidents, necessitating reliable detection methods.
• Camera-based driver distraction detection raises privacy concerns.
• Physiological signals offer a privacy-preserving alternative, but wearable technology can be intrusive.

Purpose of the Study:

• To investigate the use of phasic Galvanic Skin Response (GSR) measured by a wristband for continuous driver distraction detection.
• To develop a privacy-preserving and efficient method for identifying driver distraction in real-time.

Main Methods:

• Raw GSR signals were decomposed into phasic and tonic components using Continuous Decomposition Analysis (CDA).
• Spectro-temporal transformations of phasic GSR were analyzed for distracted (calling, texting) and non-distracted driving scenarios.
• Support Vector Machine Recursive Feature Elimination (SVM-RFE) was used for feature selection.
• Support Vector Machine (SVM) with 10-fold cross-validation (10-CV) was employed for performance evaluation.

Main Results:

• High-resolution spectro-temporal features were extracted to identify patterns associated with driver distraction.
• SVM-RFE identified a reduced feature subset, maintaining high identification accuracy.
• Cross-validation accuracy reached 94.81% with all features and 93.01% with the reduced feature set.

Conclusions:

• Phasic GSR measured by a wristband is a reliable indicator of driver distraction.
• The proposed method effectively detects distraction with high accuracy and improved efficiency.
• This approach offers a privacy-preserving solution for real-time driver distraction monitoring, enhancing road safety.