Improving Performance Of Motor Imagery Systems Using Hybrid Classifiers
Keywords:
Artificial Intelligence, hybrid Classifier, Brain Computer Interface (BCI), Logistic Regression, KNN, SVM, Random Forest Classifier, Electroencephalography (EEG), Bank Common Spatial Pattern (FBCSP)Abstract
Brain Computer Interface- Motor Imagery (BCI-MI) systems have the attention from a lot of researchers recently, especially systems based on EEG (Electroencephalography), where EEG brain signals are recorded while imagining a specific movement as an imagine moving the right hand, and recognizing it using a classifier. The preprocessing techniques and classifiers used in this recognition process have varied, but to this day they have not been sufficiently accurate to implement them in practical systems, as the accuracy of the available systems changes when tested on different data sets, in addition to the large error ratio. In this paper, we propose a hybrid recognition system for motor imagery using EEG signals, and based on a Hybrid stacking classifiers, each of it consisting of several sub-classifiers, with the use of the FBCSP (Filter Bank Common Spatial Pattern). in the signal preprocessing, where recognition ratio of 83.03%, 88.70%, 89.37% were reached on the IV2b, IV2a, AlexMI datasets, respectively.
