Study On Brain Computer Interface Based On Visual Evoked Potentials Of Different Modulation Methods

The brain computer interface(BCI)is a bridge built between the brain and the external environment, one can communicate with the outside world directly via the bridge instead of the peripheral neuromuscular. With the number of people demanding for BCI technology increasing and the maturity of neuroscience and engineering technology, more and more institutions and scientists are working on BCI research. Of all kinds of BCIs, the steady-state visual evoked potential(SSVEP) based BCI and code modulated visual evoked potential based BCI get more attention and better performance, in order to further improve the performance of these two BCIs, the paper carried out depth research and exploration on SSVEP based BCI and code modulated BCI.In SSVEP based BCI, the target stimulus is generally modulated by frequency. Because the response frequency band of SSVEP is narrow and some subjects have blind frequencies, therefore, the number of targets presented is limited. To solve this problem, we studied frequency and phase mixed coding SSVEP based BCI, by this modulation method, multiple targets can be modulated by the same frequency but different phase, the utilization of frequency is greatly improved, increasing the number of stimulus targets. Targets can be recognized by carrying out fast Fourier transform on EEG at each stimulation frequency, then projecting the Fourier coefficients for each stimulation frequency onto the reference phases of each target, the target with maximum projection value is the target to be identified. When the length of time window is 2s, the average recognition rate of all subjects reached 89.30%.In code modulated BCI, the modulation performance of pseudo-random code has a direct infulence on recognition accuracy rate, so this paper studied the modulation performance of the M sequence, nearly perfect sequence and Golay complementary sequence, template matching algorithm is applied to recognize targets. The results show that when the length of time window is one stimulation cycle, target recognition accuracies of three sequences are 89.71%, 94.29%, 93.20% respectively. According to the autocorrelation function waveforms of three sequences, compared with the M sequence, nearly perfect sequence and Golay complementary sequence have sharper peak and lower overall sidelobes. so the modulation performance of these two sequences is better. In order to increase the number of targets while ensuring recognition accuracy and recognition speed, we propose different codes packet modulation method, and combine multi-channel EEG to recognitize target. In the multi-lead system that employs nearly perfect sequence and Golay complementary sequence to modulate two groups of targets, the average recognition accuracy reaches 92.34%.