BMICA-Independent Component Analysis Based on B-Spline Mutual Information Estimation for EEG Signals

Electroencephalogram (EEG) serves as an extremely valuable tool for clinicians and researchers to study the activity of the brain in a non-invasive manner. It has long been used for the diagnosis of brain damage, for categorizing sleep stages and various central nervous system disorders like seizures and epilepsy. The EEG source signals are mixed however with other signals such as Electrooculogram (EOG) and Electromyogram (EMG) which increase the difficulty in analyzing the pure EEG and obtaining the clinical information. This paper presents a new method for denoising artifacts in mixed EEG signals. To remove these artifacts the information theoretic concept of mutual information estimated using B-Spline was used in creating an approach for Independent Component Analysis (ICA). In this paper we present a B-Spline estimator for mutual information to find the independent components in EEG signals. Tests showed that B-Spline Mutual Information Independent Component Analysis (BMICA) exhibits better performance than the standard Independent Component Analysis algorithms of FastICA, JADE, SOBI and EFICA in similar simulations. BMICA was found to be also more reliable than the well known FastICA in producing purer EEG signals for analysis.