Brain computer interface eeg signal processing with new approaches of feature extraction and classification

Abstract

An efficient processing approach is essential for increasing identification accuracy since the Electroencephalogram (EEG) signals produced by the Brian-Computer Interface (BCI) apparatus are non-linear, non-stationary, and time varying. The interpretation of scalp EEG recordings can be hampered by non-brain contributions to Electroencephalographic (EEG) signals, referred to as artifacts. This is particularly accurate when the artifacts have significant amplitudes such as movement artifacts or appear repeatedly like eye-movement artifacts. Common disturbances in the capture of EEG signals include Electrooculogram (EOG), Electrocardiogram (ECG), Electromyogram (EMG) and other artifacts, which have a significant impact on the extraction of meaningful information. This study suggests integrating the Singular Spectrum Analysis (SSA) and Independent Component Analysis (ICA) methods to pre-process the EEG data. In this research, Higher Order Linear Moment based SSA (HOL-SSA) is used to first decompose the EEG signals into multivariate signals after which the source signals are extracted from the multivariate data using Online Recursive ICA (ORICA). Thus, the proposed HOL-SSA and ORICA based pre-processing approach has shown improved results in artifact rejection. The experimental findings demonstrate that the suggested technique can identify and eliminate EOG, ECG, EMG and other artifacts from EEG data while still preserving brain activity that is ignored by the noise component. The characteristics of the denoised EEG data are then extracted using the Common Spatial Pattern (CSP) technique. newline