Pattern Classification of Visual Evoked Potentials Based on Parallel Factor Analysis
Visual Evoked Potentials (VEPs) reflect the brain’s mental process to specific stimuli, including perception and recognition. Feature analysis of VEPs evoked by geometric figures is of significance in understanding visual neural mechanism and has potential applications in the field of brain computer interface and biomedical engineering. We use Parallel Factor (PARAFAC) model to extract features of the VEPs triggered by three classes of geometric figures, and construct the computational model for class discrimination. PARAFAC is used to decompose the wavelet transformed VEPs. Then by the proposed computational model, we can project single trial data into the subspace spanned by channel× frequency× time of the factors to obtain the feature vectors. We further use SVM to classify the feature vectors of the selected two classes, achieving the highest classification accuracy 80%.
Unable to display preview. Download preview PDF.
- 7.Lathauwer, L., Moor, B., Vandewalle, J.: On the best Rank-1 and Rank-(r1, r2, …, rn)approximation of higher-order tensors. Society for Industrial and Applied Mathematics 21 (2000) 1324–1342.Google Scholar
- 8.Mørup, M., Hansen, L., Parnas, J., Arnfred, S.: Decomposing the time-frequency representation of EEG using non-negative matrix and multi-way factorization. Technical reports (2006).Google Scholar