EEG-based emergency braking intention detection during simulated driving.

Area of Science:

• Neuroscience and Human-Computer Interaction
• Driver monitoring systems
• Automotive safety technology

Background:

• Current electroencephalogram (EEG)-based driver's emergency braking intention detection primarily distinguishes emergency from normal driving.
• Limited research exists on differentiating emergency braking from normal braking using EEG.
• Existing methods rely on traditional machine learning with manually extracted features.

Purpose of the Study:

• To propose a novel EEG-based strategy for detecting a driver's emergency braking intention.
• To compare traditional, Riemannian geometry-based, and deep learning methods for this detection task.
• To utilize raw EEG signals as input, bypassing manual feature extraction.

Main Methods:

• Experiments conducted on a simulated driving platform with normal driving, normal braking, and emergency braking scenarios.
• Analysis of EEG feature maps to identify differences between braking modes.
• Application of traditional, Riemannian geometry-based, and deep learning algorithms to raw EEG signals for intention prediction.

Main Results:

• Both Riemannian geometry-based and deep learning methods outperformed traditional methods.
• The deep learning-based EEGNet algorithm achieved an Area Under the Curve (AUC) of 0.91 and an F1 score of 0.85 for distinguishing emergency from normal braking at 200 ms prior.
• Significant differences in EEG feature maps were observed between emergency and normal braking.

Conclusions:

• Detecting emergency braking intention from EEG signals is feasible, differentiating it from both normal driving and normal braking.
• Accurate identification of emergency braking intention allows for earlier activation of vehicle braking systems.
• This technology offers a user-centered framework for human-vehicle co-driving, potentially preventing serious collisions.