Classification of Wrist Movements Using EEG Signals

In this paper, executed wrist movements of four different kinds (Extension, Flexion, Pronation and Supination), for both left and right hands have been classified from single trial EEG (Electroencephalography). This has been done in two stages. The first stage discriminates between left and right wrist movements. In the second stage, movements of the same wrist have been classified. Two-class classification has been done on 16 channel EEG data recorded on three subjects. Independent Component Analysis (ICA) has been studied as a preprocessing tool, and the results obtained with and without the use of ICA have been compared. Several time and frequency domain features were extracted and used for classification. Out of these, energy of the EEG signal in beta and gamma bands (12-70 Hz) was found to give best results. Linear Discriminant Analysis (LDA) has been used for classification to keep the complexity minimum. Discrete Cosine Transform (DCT) has been used to reduce the number of features used for classification. For left and right wrist movements, a classification accuracy of 99±1 % has been obtained after preprocessing using ICA. However, for classification of movements of the same hand, the best accuracy obtained was 75 %.