| Title |
Target Image Detection using Extended Covariance Matrix of EEG Signals |
| Authors |
이기배(Kibae Lee) ; 고건혁(Guhn Hyeok Ko) ; 이종현(Chong Hyun Lee) |
| DOI |
https://doi.org/10.5573/ieie.2023.60.8.58 |
| Keywords |
EEG signal; RSVP; Extended covariance matrix; Riemannian geometry; LSTM AE |
| Abstract |
Image interpretation using brain waves generated by visual stimuli has recently been studied. Existing studies have limitations in practical applications assuming the stationarity of the invariant stochastic properties of electroencephalogram (EEG) signals. To overcome these limitations, we proposes the feature extraction algorithm of EEG signals using the extended covariance matrix. The algorithm proposed generates the extended covariance matrix that includes the covariance of EEG signal and the cross-covariance between different signals and extracts the projected feature vector based on Riemannian manifold. We also propose the long short-term memory autoencoder (LSTM AE) model to adress the issue of frequent false alarms in multi-image target detection. In the rapid serial visual presentation (RSVP) experiment, where images are presented at 2 and 10 frames per second (FPS), the extended covariance matrix based feature shows single image detection performance of 85.42% and 77.38%, respectively. The target detection results within a chunk utilizing the extended covariance matrix based feature and the LSTM AE model, show performance of 77.34% and 74.66% for images presented at 2FPS and 10FPS, respectively. |