Extraction of EEG signals during L/R hand motor imagery based on ERD/S

Abstract

Electroencephalogram (EEG) based on brain computer interfaces (BCIs) provides new channels between human brain and the outside world. An EEG feature, event-related desynchronization/synchronization (ERD/S) caused by motor imagery (MI), is broadly used to analyze the brain activity and estimate human motor intention. In this research, our purpose is to extract the features based on ERD/S, and determine left/right (L/R) hand side movements through Support Vector Machine (SVM). In the past, raising the accuracy of MI classification is always the main objective of research teams. Hence, we propose a novel method to extract features providing better classification accuracy. After feature extraction, linear discriminant analysis (LDA) was used to perform dimension reduction. Results came from the classification of SVM (RBF kernel) with leaveone-out cross-validation (LOOCV). Approximately 97.62% classification accuracy is achieved to determine L/R hand movements.