Abstract
The ability of recognizing driving actions could help building a more advanced driving assitance system, and could even be applied in automated driving to improve the driving safety. In this paper, we investigate the offline recognition of three classes of driving actions (turning left, turning right and braking), based on electroencephalography (EEG) signals. A simulated experiment was conducted to collect EEG data of participants. The proposed algorithm includes Wavelet Analysis and Common Spatial Patterns (CSP), to extract the discriminative features. The classification results were obtained using the Linear Discriminant Analysis (LDA). The results yielded an average single trial classification accuracy of 70.25% for all subjects, showing the discrimination of different actions and the correlation between driving actions and EEG signals.
Gan Huang
My research interests include Neural Modulation, Brain Computer Interface and Neural Prosthetics.