Research On Visual Brain-Computer Interface Based On C-VEP And SSVEP Hybrid Coding Modulation

Brain-computer interface(BCI)is a new communication system that establishes a direct path between the human brain and external devices.It directly control external electronic devices,which does not require the participation of peripheral nerves and muscle tissues.In recent years,visual evoked potentials based BCI have gradually become hotspot of research on BCIs.In this paper,visual brain-computer interface based on code modulation and frequency modulation is studied.For the problems of recognizable targets limitation of c-VEP BCI and SSVEP BCI,a new BCI paradigm based on c-VEP and SSVEP hybrid code modulation is proposed in this paper.This proposed BCI paradigm can effectively increase the number of targets.In this paper,a hybrid visual brain computer interface system with 32 targets is designed,each target is modulated by c-VEP and SSVEP.On this basis,three different modes of a single stimulus target are proposed:internal-external mode,up-down mode and left-right mode.Combining template matching method and canonical correlation analysis algorithm,by introducing a confidence pointer,a target recognition algorithm based on hybrid visual brain-computer interface is proposed.Nine subjects were invited to participate in the brain-computer interface experiment,and the experimental data were processed and analyzed in detail.Firstly,the signal characteristics of the reference template are analyzed.The results show that the stimulus response generated by the pseudo-random sequences and the frequency sequences used in the system show good characteristics,make each target easy to be discriminated.Then the performance of three models are analyzed,the results of the experimental data show that the average classification accuracy of the internal-external mode hybrid BCI system is 87.09%,and the average information transfer rate is 49.28bit/min,which shows better performance than the up-down mode and left-right mode.Meanwhile,the influence of different data length and confidence pointer values on system performance is studied,and the results show that increasing the data length and selecting proper confidence pointer can effectively improve the classification rate.