Research On Feature Extraction And Recognition Of Mental Tasks Based On Electroencephalogram Recordings

Brain-computer interface (BCI) is a brand-new interface between the human brain and the computer. BCIs give their users communication and control channels that do not depend on the brain's normal output channels of peripheral nerves and muscles,but accepts commands encoded in neurophysiological signals. Current interests in BCI development come mainly from the hope that this technology could be a new valuable augmentative communication option for those with severe motor disabilities that prevent them from using conventional augmentative technologies. BCI technology also has potential applications in other fields such as autocontrol and defense. Because of the enormous prospects of its applications, BCI technology has provoken interests of international scientists. BCI research has drawn attention of scientists in the brain-science research, rehabilitation engineering, biomedical engineering and human machine automatic control.Most of the existing brain-computer interface technologies are based on electroencephalogram (EEG) signals. EEG signals are electrical impulses produced by the brain,and can be recorded directly from the scalp and the palladium. Reseachs of communication interchange methods based on EEG are managing to transform EEG or some of its elements into a new output pathway, through which the brain can perform communication interchange and control with its envirament.In BCI systems, feature extraction and recognition is the crucial work for EEG signals analysing. If this work is incorrectly done, the BCI system will also incorrectly recognize the special instruction in the user's mental activities, and neither can the system send out correct controlling instruction which coincides with the user's intention to the devices .Our study in this dissertation is mainly about this work.An adaptive autoregressive model (AAR) is proposed to extract features from EEG signals. As a non-stationary stochastic model, the AAR model is very suitable to discribe EEG signals which are also non-stationary stochastic signals. The application...