Improving Movement Onset Detection in EEGbased BCIs with ‘Weakly’ Supervised Learning
As EEG based BCI systems that are expected to function in everyday situations are developed, the challenge to deal with noise due to the environment or the subject itself becomes more important. In this project it was explored whether the ‘weakly’ supervised learning techniques for movement onset detection with an EEG-based BCI used by Awwad Shiekh Hasan et al. might improve the classification accuracy in the experiment conducted by Verbaarschot et al. This experiment, in which participants played a self-paced BCI-game, was carried out during the InScience festival; a noise-rich ecologically valid environment. It was found that methods proposed by Awwad Shiekh Hasan et al. performed better than the linear classifier Verbaarschot et al. used however both were not very accurate in detecting movement onset.
Faculteit der Sociale Wetenschappen