Detection of Motor Imagery (MI) Event in Electroencephalogram (EEG) Signals using Artificial Intelligence Technique

With the recent development of technology and acquisition devices, the research of detection and classification utilizing EEG signals is rapidly increasing. One of the critical research in the field of the brain-computer interface includes an accurate detection of motor neuron behavior called motor imagery (MI) events. Due to the increased number in people with inabilities (e.g. paralyzed people, autism, and elderly people), accurate detection of MI events can of great help. In this work, a method for the detection of the MI events using the electroencephalogram (EEG) signal is proposed. Data from thirteen random subjects from a publicly available dataset was utilized. Firstly, the EEG signals were preprocessed and then a combination of time domain and frequency domain features were extracted from the signals. The number of features was reduced and selected using a minimum-redundancy-maximum-relevance (MRMR) algorithm and forward feature selection. On the subject level with leave-one-subject-out cross-validation, MI events were recognized with an average F1-score of 68.69% using the Support Vector Machine classification model. The best individual performance was obtained with an F1-score of 79%. These results suggest that the proposed approach is able to identify MI events in the EEG signal and thus the method may potentially be integrated into devices that can assist people with inability. Further improvement in the performance of the method can be done by carrying out testing in a wider population.