EEG Signal Categorization Performance: Influence of ERD position within classification window

Much research has been done in the field of cue-paced movement (imagery) classification performance. But cue-paced classification greatly narrows BCI bandwidth and possibilities. This research aims to uncover the extent to which signal classification performance is lost when the Event Related Desynchronisation (ERD) is not placed optimally in the classification window. This also denotes the primary difference between cueand self-paced BCI. In self-paced situations cues are not known, resulting in windows where ERD placement differs every sample. Raw electroencephalography data was obtained from one test subject and was categorized for several different window sizes and ERD placements. This was achieved by sliding the classification window backwards and forwards in time. Markers were used as a reference point. We found that the placement of windows during training has great influence on classification performance. Larger windows seem to result in higher overall performance but are prone for multiple hits per ERD. We also found that classification performance rises steeply when more bandwidth power of the ERD enters the classification window, but rises steeper for smaller windows. When classification windows are slid further past the marker performance starts to drop, but drops slower than it rises.