Processing movement related cortical potentials in EEG signals for identification of slow and fast movements

The extraction of intended kinetic information from an EEG signal can have several applications related to the rehabilitation for subjects with various neurological disorders. However, the task is mainly constrained by the low signal-to-noise ratio for the EEG signals. It is well known that the cortical activity takes place at a very low frequency since it is characterized by the dropping of movement related cortical potential (MRCP) across the sampled EEG signal. The strong variations in the MRCP is indicative of the noise due to various sources. The aim of this work is to remove this noise from the EEG signals using empirical mode decomposition, which decomposes a signal into harmonics (intrinsic mode functions - IMF) of various frequencies. The IMFs pertaining to small frequencies are later used for features extraction where we extract the spatial and spectral features from the selected IMFs. The features are later used for classification using support vector machines (SVM). Our experiments show superior results to the benchmark method for the underlying dataset that has been used in this research.