Thesis supervision | Anti Ingel: Machine Learning in VEP-based BCI
In this thesis, a classification method for SSVEP-based BCI is proposed. The classification method is based on simple comparisons of extracted feature values and thresholds and it involves a way of optimizing the thresholds. Optimizing the thresholds is formalized as a maximization task of the information transfer rate of BCI, but instead of using the standard formula for calculating ITR, more general formula is derived. This allows the thresholds to be automatically optimized and avoids calculating incorrect ITR estimate. The proposed method shows good performance in classifying targets of a BCI and achieves ITR as high as 60 bit/min. The proposed method also provides a way to reduce false classifications, which is important in real-world applications. BCIs have high potential to be used in the field of medicine as they provides a way for severely disabled people to control external devices.