• Daniel Bøcker Sørensen
  • Anders Jensen
  • Nicolai Lund Hasager Kirk
4. term, Computer Science, Master (Master Programme)
This project investigates the development of a BCI system using a consumer grade EEG headset. This includes signal acquisition, preprocessing, feature extraction and classification and/or regression. Riemannian geometry is taken advantage of, because of the natural EEG signals can be directly classified in this space. The Riemannian methods investigated includes Minimum Distance to Riemannian Mean (MDRM) and Tangent Space LDA (TSLDA). These methods are tested and compared against the well known methods Common Spatial Pattern (CSP), combined with Linear Discriminant Analysis (LDA), which was investigated in our previous work. Furthermore it is investigated how it is possible to combine two predictor tasks, instead of one, e.g. classification. This is done by combining classification and regression simultaneously, which opens up new ways of how a BCI system can be used. This report documents the development of a combined two-predictor-task BCI system, and concludes the found results of said methods.
LanguageEnglish
Publication date11 Jun 2015
Number of pages60
ID: 214022581