A Study On Brain Computer Interface Based On The Steady-State Visual Evoked Potential Phase

Patients suffering from severe motor diseases, such as amyotrophic lateral scleroses (ALS), multiple sclerosis, cerebral palsy and head trauma, is incapable of communi-cating with the external world, because the neural pathway of controlling muscles are damaged so that the voluntary muscle control partly or totally loses. In order to improve the patients'quality of life and decrease the burden of society and family, brain computer interfaces (BCIs) provide an efficient and promising solution. Among all BCI modalities, BCIs based on the steady state visual evoked potential (SSVEP) can provide higher accuracy and a faster information transfer rate. It is significant to study SSVEP based BCIs.The SSVEP refers to the activity of the cerebral cortex that results from attend-ing to a repetitive visual stimulus oscillating at a constant frequency higher than 4 Hz and is characterized by peaks at the stimulation frequency and/or its harmonics in the power spectral density of electroencephalogram (EEG) signals. Most of current SSVEP based BCIs use frequencies between 4 Hz and 30 Hz. These frequencies lead to a higher probability of causing visual fatigue, a higher risk of photic-and pattern-induced seizure. Furthermore, these frequencies locate in the frequency range of spon-taneous EEG signals, which might cause false positive. Frequencies higher than 30 Hz, therefore, are more preferable, because they are closer to the critical fusion frequency. But only a limited number of frequencies above 30 Hz can elicit a sufficiently strong SSVEP for BCI purposes. Current studies show the SSVEP phase can be used to iden-tify the target stimulation.In order to extract the SSVEP phase efficiently, a method based on spatial filtering and phase synchrony analysis was proposed in this study. First, a one-dimensional signal resulted from multiple electrode signals with the average maximum contrast combination (aMCC) which is a variation on not only the maximum contrast combi-nation but also the common spatial pattern. Second, the phase synchrony analysis based on the Hilbert transform was used to extract the instantaneous phase difference of the SSVEP and stimulation. This phase difference was defined as the SSVEP phase. Based on this method, a high performance four-command realtime BCI system was de-veloped to achieve accuracy 94.47%±6.02% and information transfer rate 36.03±5.42 bits/minute. Comparing with current methods for extracting the SSVEP phase, the proposed method requires a shorter data and provides a more stable detection.Aiming at the limited number of available phases at one frequency, a cyclic class error-correcting code was proposed based on the cyclic code. The synthetic analysis and real code analysis demonstrated that this error-correcting code can efficiently in-crease the number of stimulation patterns and consequently the executable commands of a BCI, and the code having (word length minus 1) times phase transitions should (?)excluded.A method based on parameter-tuned stochastic resonance (SR) for enhancing the SSVEP was proposed. This method first used an autoregressive (AR) based whitening filter to remove the 1/f trend in EEG spectrum, and then linearly converted the SSVEP frequency to one smaller than 1 Hz to meet the conditions where SR could occur. Fi-nally, SR effect was achieved by tuning the parameters of a bistable nonlinear system so that the SSVEP could be enhanced. The result showed that a peak at the SSVEP frequency appears in EEG spectrum after SR processing and the signal-to-noise ra-tio of the SSVEP increases significantly. This method is helpful in detecting SSVEP. In addition, this study introduced noise at different strength into the stimulation and observed the corresponding SSVEP response. It was found that the SSVEP response shows an SR phenomena.Since the EEG signal is nonlinear, sample entropy is suitable for EEG analysis. The first was to identify the SSVEP from the background EEG. The results show that the SSVEP entropy is smaller and that the entropy of background EEG is close for all subjects at different frequencies which demonstrates that sample entropy can be a gen-eral criterion to detect the SSVEP. The second was to study the pain relief mechanism of transcutaneous electrical acupoint stimulation (TEAS). The analysis of the data of rest phase, pain phase and therapy phase shows that the degree of disorder and the complexity increase in pain and become lower after electrically stimulating Hegu acu-point. The results show that sample entropy can be used to potentially explain the trend of pain degrees and that TEAS may modulate pain mechanism in brain as the variation of sample entropy is consistent with the reported pain degrees. The third was that sample entropy of eyeblink component after ICA is smaller than the others so that it can be used to automatically identify this artifact component.