LVQ and single trial EEG classification

Abstract

Electroencephalogram (EEG) signals, which reflect the electrical activity of the brain, can be classified on a single trial basis. An internally or externally paced event results not only in the generation of an event-related potential (ERP), but also in a change of the ongoing bioelectrical brain activity in form of either an event-related desynchronization (ERD), or event-related synchronization (ERS). In contrast to the ERP, which is a time- and phase-locked response of the brain, the ERD/ERS is time-locked but not phase-locked. While the ERP can be extracted from the ongoing EEG directly with averaging techniques, the evaluation of ERD and ERS requires a preceding filtering and squaring of the samples. The resulting signals can then be analyzed with averaging techniques or with learning classification methods. In this chapter, the classification of EEG signals during preparation of different types of movement with learning vector quantization (LVQ) is addressed. A good separability for left and right finger movements and also foot and tongue movements is shown in the chapter. Furthermore, the problem of feature selection is addressed. A modified version of LVQ, distinction sensitive learning vector quantization (DSLVQ), is discussed and applied for the selection of optimal EEG parameters.