PROCESS: Projection-Based Classification of Electroencephalograph Signals

Classification of electroencephalograph (EEG) signals is the common denominator in EEG-based recognition systems that are relevant to many applications ranging from medical diagnosis to EEG-controlled devices such as web browsers or typing tools for paralyzed patients. Here, we propose a new method for the classification of EEG signals. One of its core components projects EEG signals into a vector space. We demonstrate that this projection may allow visual inspection and therefore exploratory analysis of large EEG datasets. Subsequently, we use logistic regression with our novel vector representation in order to classify EEG signals. Our experiments on a large, publicly available real-world dataset containing 11028 EEG signals show that our approach is robust and accurate, i.e., it outperforms state-of-the-art classifiers in various classification tasks, such as classification according to disease or stimulus. Furthermore, we point out that our approach requires only the calculation of a few DTW distances, therefore, our approach is fast compared to other DTW-based classifiers.