Detection of steering direction using EEG recordings based on sample entropy and time-frequency analysis

Area of Science:

• Neuroscience
• Transportation Safety
• Machine Learning

Background:

• Preventing traffic accidents is a major societal goal.
• Monitoring driver intentions in real-time is crucial for proactive safety measures.

Purpose of the Study:

• To investigate the feasibility of predicting driver steering intentions using electroencephalography (EEG).
• To analyze brain dynamics preceding steering actions and classify steering direction.

Main Methods:

• Recorded high-resolution EEG and accelerometer data from 5 subjects driving in real conditions.
• Applied sensor-level analyses: sample entropy and time-frequency analysis.
• Utilized machine learning (Principal Component Analysis and Support Vector Machine) for classification.

Main Results:

• Observed increased sample entropy and theta/alpha band power approximately 100ms before steering onset.
• Achieved higher classification accuracy for steering direction using sample entropy (73.5% ±6.8) compared to power spectral analysis (62.6% ±5.6).

Conclusions:

• EEG signals, particularly sample entropy, can predict steering direction with significant accuracy.
• This approach shows promise for developing advanced driver-assistance systems to prevent accidents.