Eeg-based Brain - Computer Interface Asynchronous System

Brain-Computer Interface (BCI) extracts the signals from human brain, processes and translates these signals into commands which can control the computer. By BCI systems, the users can operate external devices directly through their brains. This paper attempts to research and develop an asynchronous online BCI system based on electroencephalography (EEG), by which the user can control the cursor in one dimension at his/her own intent just through left hand or right hand motor imagery. Details of this work are as follows:Firstly, by analyzing the dataset in BCI Competition III, the physiological fundamental of the prospective BCI system was chosen, i.e. the Event-Related Desynchronization/Synchronization (ERD/ERS) ofμandβrhythms. Secondly, through investigating the dataset in BCI Competition IV, the feature extracting method and the classifier were determined according to the asynchronous EEG data, i.e. the Common Spatial Pattern (CSP) as the feature extracting method and the Support Vector Machine (SVM) as the classifier. Meanwhile, the algorithm of idle-state detection was also investigated, especially the technique to determine the thresholds for detecting the idle states. Then an effective tool for choosing thresholds was used specifically for the development of online asynchronous BCI system, i.e. the Receiver-Operating Characteristic Curve (ROC Curve). Thirdly, the chosen feature extracting method, classifier and idle-state detection algorithm were integrated into the software platform of an existing synchronous BCI system in our laboratory, thus leading to an asynchronous system. To make the system work in asynchronous way, a temporal mechanism was added into the online signal processing. Finally, the performance of the system was evaluated by the user, and the future work to improve the system was also proposed.