Generation of driver monitoring ground truth dataset – drowsiness

Domain: Interior sensing
Service: Methodology preparation, dataset collection and analysis

Challenge

Our Tier 2 Partner needed ground truth dataset for the development of novel driver drowsiness detection algorithms as part of their driver monitoring product. Our team had to combine multiple measures including: driving performance, detailed eye-tracking and physiological data (EEG and ECG) to reference drowsiness levels. As driver drowsiness is a safety-critical state, our team proposed a safe and standardised solution that also enabled to generate relevant dataset.

The Robotec.ai team developed proprietary methodology for collecting and referencing drivers’ drowsiness data while driving using a high-end driving simulator as well as a dedicated EEG and ECG setup. The study was conducted with human drivers from multiple ethnicities and age groups to meet requirements of international automotive standards.

Solution

• Methodology of inducing drowsiness through sleep deprivation and accounting for circadian rhythm as well as applying monotonous driving conditions.
• Using Robotec.ai proprietary signal processing algorithms for driving performance and physiological data analysis.
• Setting up dedicated 500 Hz physiological setup EEG and ECG and constant monitoring of signal quality.
• Programming the high-end driving simulator.
• Synchronising of multiple data acquisition systems.

Expertise fields

• Biomedical physics
• Signal processing
• Data analysis
• Cognitive and transport psychology
• Software engineering (driving simulator and synchronising)