Classifying direction of the right index finger movement from delta band activity using HMM

This contribution examines the usage of low frequency components (<; 5 Hz) in single trial EEG recordings obtained during right index finger movement for classification of reaching and grasping movements. These components contain delta band activity and Movement Related Potentials (MRPs) associated with the movements. Time-frequency development is used to classify the movements using Hidden Markov Model based classifier. It is shown that in some cases the utilization of these components can lead to a better classification score than the utilization of the previously used oscillatory activity in the μ and β bands, which are used as the reference here. The classification score has changed on average by -1.3% (-11.7% to +16.1%) compared to the referenced 5-40 Hz band. By choosing the newly examined band only for subjects where there is a benefit in it, a score of 90.9% was obtained (+2.9% improvement on reference itself). The examined frequency band is optimized for each subject as the inter-subject variability of EEG plays a role here