SVM for Classification of Ten-Finger Imagined Movements using tEEG Signals

The search for advancements and insights in Brain-Computer Interface (BCI) technology has led to the exploration of tripolar ring electrodes-electroencephalography (tEEG). While conventional electroencephalography (EEG) remains popular for its non-invasive nature and high temporal resolution, tEEG presents itself as a viable alternative with its enhanced spatial resolution, alongside the retention of temporal characteristics, artifact suppression, and noise reduction capabilities. Prior research has investigated the utility of tEEG in classifying imagined two-finger movements and physical five-finger movements, albeit with limited success. This work utilizes a Support Vector Machine (SVM) for imagined ten-finger classification using tEEG and EEG signals of seven healthy subjects. Data features were extracted solely from the unprocessed, raw time series of a single tEEG electrode. The obtained results demonstrated accuracies approaching $100 \%$ in most instances.