Brain Computer Interface Using Eeg Signal

Abstract

The field of rehabilitation engineering is booming with interest in brain-computer newlineinterface (BCI). For the proper interaction between people and robots, significant newlinecharacteristics and effective classification methods are necessary. The improvement newlineof motor imagery and real movement precision is the main goal of this thesis, which newlinealso examines the impact of kinematic movement variability over the sensory-motor newlinecortex using a 10 20 EEG system. newlineThe healthy participant s EEG data were collected by preparing them for a suggested newlinetechnique. These healthy participants were instructed to use the ArmeoSpring newlinetherapy device to move their elbows while playing a computer game designed for newlinerehabilitation. Raw EEG data were considered and filtered through an 8 30 Hz newlinebandpass filter. Three modelling features are estimated and evaluated using the newlineSVM classifier: the Common Spatial Pattern (CSP), Event-Related Desynchronization newline(ERD) / Synchronization (ERS), and Auto-Regressive Moving Average (ARMA). The newlinesuggested framework can distinguish between distinct elbow kinematic motions and newlinecategorize them with an average accuracy of 84.61%, 92.77%, and 97.26% for the newlineCSP, ARMA, and ERD/ERS characteristics, respectively. newlineThe second contribution focuses on filtering and segmenting raw EEG data. A newlinecomplicated, nonlinear, non-stationary signal is the EEG. As a result, researchers newlinehave concentrated on the management of unlabeled EEG data for the categorization newlineof hand motions. The contribution suggests a method termed clustering of unlabeled newlineEEG data for solving this issue. The Enobio-20 machine and the ARMEO Spring newlinedevice were used to conduct elbow movements while the recommended procedure newlinewas being followed to capture EEG data from the sensory-motor brain. A bandpass newlinefilter is used to purify the EEG data that was recorded. A number of frequency newlinebands are created from the filtered EEG data using the discrete wavelet transform newline(DWT). The K-means method is used for each sub-band to cluster the coefficients of newlineeach frequency band. The feat