Classification of Motor Imagery based EEG signals using Ensemble model

The Brain-Computer Interfaces (BCIs) has been a source of fascination since its discovery. Controlling items just by thinking about them is a new degree of modernity. Out of many paradigms present in electroencephalogram (EEG), motor imagery (MI) has gained significance for being a safe and non-invasive process and including cognitive engagement. The main challenge for MI-based BCIs is Feature extraction and training a classifier, which could give us better accuracy. In our research, we adhere to the standard MI-based BCI workflow, but our methodology introduces an optimized and better model for classifying EEG data. To reduce the data used for the classification, we extracted five features, spanning both frequency and wavelet domain features, from the EEG signals. These features are then dimensionally reduced by passing through Linear Discriminant Analysis (LDA), and the best two features have been selected for classification. This Ensemble classifier model is applied to the best two features that have been extracted. This result is being compared with existing traditional Machine Learning algorithms - Support Vector Machines (SVM) and Logistic Regression (LR). Our empirical findings highlight the remarkable boost in classification accuracy achieved by the ensemble model compared to the tried-and-true machine learning methods. The suggested approach can be further enhanced to create a dependable and real-time BCI application.