Recognition of EEG patterns during mental intentions: a comparative study

Recognition of specific oscillatory patterns in human electroencephalograms (EEGs) is an important problem that has attracted significant attention for creating brain-computer interfaces (BCIs). Some of these patterns are easily identified by various numerical methods. However, it is much more difficult to recognize mental intentions that can be further transformed into control commands for hardware, and the choice of the appropriate numerical tool becomes very important. In this study, we compare several numerical methods applied to multichannel EEGs recorded in untrained volunteers who imagined arm and leg movements. We show that the quality of recognition varies between different methods and depends on the subject. We discuss the possibilities of reliable separation between imaginary movements of various types.