Human-Centric Spatial Cognition Detecting System Based on Drivers' Electroencephalogram Signals for Autonomous...

Area of Science:

• Neuroscience
• Human-Computer Interaction
• Robotics

Background:

• Autonomous driving systems struggle to align with driver intentions due to limitations in detecting cognitive states.
• Existing systems often focus on intention or hazard recognition, neglecting spatial cognition.

Purpose of the Study:

• To develop a human-centric system for detecting drivers' spatial cognition using electroencephalogram (EEG) signals.
• To enable autonomous driving systems to make contextually aligned decisions by understanding drivers' focus on relative distance and orientation.

Main Methods:

• Developed a system with EEG signal preprocessing and spatial cognition decoding.
• Introduced a novel Dual-Time-Feature Network (DTFNet) integrating multi-scale temporal features and an attention mechanism for EEG decoding.
• Investigated temporal dynamics of spatial cognition perception.

Main Results:

• The DTFNet achieved 65.67% and 50.65% accuracy in three-class tasks, and 84.46% and 70.50% in binary tasks.
• Demonstrated superior performance compared to existing EEG decoding methods.
• Observed that relative distance perception occurs slightly later than relative orientation perception.

Conclusions:

• The proposed EEG-based spatial cognition detection system enhances autonomous driving decision-making.
• DTFNet offers an effective method for decoding complex cognitive states from EEG signals.
• Findings provide insights into the temporal sequencing of spatial cognitive processes in drivers.