Skip links

Classification of Motor Imagery Tasks for Brain-Computer Interface Applications

Brain-computer interface (BCI), also known as the brain-machine interface (BMI), is a system that enables humans to interact with their surroundings via employing control signals generated from electroencephalographic (EEG) activity, without the intervention of peripheral nerves and muscles. An important part of a brain-computer interface is an algorithm for classifying different commands that the user may want to execute.


There are several neurological phenomena that can be used in a BCI. One of them is event-related de-synchronization (ERD), which is a temporary decrease in the power of the mu and beta brain waves. This phenomenon can be registered using electroencephalography (EEG) and occurs when a subject performs or imagines a limb movement. The goal of this research is to implement an algorithm that would be able to classify four different motor imagery tasks, in which a user imagines a movement of one of the following parts of the body: left hand, right hand, both feet, or tongue.


The EEG data used taken from BCI competition IV, dataset 2a. It contains 22-channel EEG recordings of 4 MI tasks performed by multiple subjects. Using electroencephalogram (EEG) data from the BCI Competition we test feature extraction techniques: wavelet packet decomposition features, and two classifiers: Decision Tree (DT) and support vector machines (SVM). We results compared with those present in the literature by utilizing Decision Tree (DT) and Support Vector Machine(SVM), our findings showed superior classification results, a classification accuracy of 79.2% and 85.5%, Kappa score 74% and 61% and F-measure 86.5% and 92% Sequentially, for two BCI competition datasets, with respect to results from previous studies. We test all algorithms with different parameters on the training set using cross-validation. All algorithms obtain results, which would have enabled them to win in the BCI Competition IV. This result shows that it is possible to use a simple algorithm for EEG data classification and obtain good results.

Join the Discussion

Return to top of page