Studies On EEG-Based Brain-Computer Interface

A brain–computer interface (BCI) is a communication system that does not depend on the brain's normal output pathways of peripheral nerves and muscles. It is an active topic in brain function research.A new method for BCI construction is proposed in the dissertation: using steady-state visual evoked potential (SSVEP). The BCI system based on this method has the advantages of noninvasive signal recording, little training required for use, and high information transfer rate. To improve the performance of the system, the visual stimulation parameters and translation algorithm are studied. Three parameters that may influence the amplitude of SSVEP are studied by experiments, i.e., modulation mode, color, and modulation depth. The optimal stimulation parameters are provided. In translation algorithm aspect, slide-window least square lattice combined with adaptive notch filter is used to realize spectral line enhancement, which provides notable improvement in signal-to-noise ratio (SNR) and high convergence speed. A periodogram-based method is established to extract the stimulation frequency of SSVEP, in which the threshold for frequency component extraction is determined according to the probability of false alarm. The average information transfer rate of the BCI system designed according to these methods is higher than 50 bits/min, superior to the data reported in references, 5 to 25 bits/min.The biofeedback phenomenon of evoked potential is studied in detail. It is found that subjects can learn to regulate the amplitude of SSVEP by biofeedback training, and can use that regulation to generate binary control. This regulation is realized by changing the amplitude difference and/or phase difference of the SSVEP between left and right hemisphere of brain. Further studies show that the training effect is effective on other stimulation frequencies, and is continuable. These studies provide more information on the biofeedback phenomenon of evoked potential.The mu rhythm-based BCI, which belongs to independent BCI, is the system with potential applications. The translation algorithm for mu rhythm-based BCI is studied in the dissertation, in which the mu rhythm is enhanced and extracted from spatial, frequency and time domain. The algorithm is applied to the data from 3 subjects in a four-class classification problem, and the accuracy is comparable to current best results.Two applications of SSVEP-based BCI are introduced in the dissertation, i.e., environmental control and rehabilitation robot control. The custom-made environmental controller integrates multiple electric device controllers into a single unit, which brings great convenience to the motion-disabled. BCI-based rehabilitation robot control provides a new approach for clinical paralysis rehabilitation.