Algorithm Design Of BCI Based On Frequency And Phase Feature Extraction Of Steady-State Visual Evoked Potential

With the rapid development of computer science, signal processing technology, and the increasing consciousnesss for the disabled in improving life quality, a lot of clinical rehabilitation treatments using brain-computer interface(BCI) have sprung up in recent years. Among these, the BCI based on steady-state visual evoked potential(SSVEP) is usually chosen as a bridge carrying brain commands for its advantages of no invasion, simple system configuration and high information conversion rate. One performance index of SSVEP-BCI is the amount of realizable commands(i.e., target stimulus blocks). The larger this amount is, the more diverse the extended actions are and also the more perfect the system will be.At present, the most commonly used SSVEP command recognition method is based on extracting the frequency information only. However, for the SSVEP stimuls signals produced by liquid crystal display(LCD), due to the excitation frequency is obtained by integer dividing the refresh rate of LCD, then the amount of stimulus is restricted obviously. Furthermore, those frequencies related to non synchronous sampling have to be discarded. Moreover, due to the fact that EEG signals collected by the electrodes are weak and also suffer from some artifact constituents, the accuracy of detection will be heavily influenced. Therefore, some novel measures of signal processing are required.In order to increase the amount of stimulus within the permissible detectable narrow band, in the coding module, this paper proposes one BCI hybrid design combining frequency coding with phase difference coding. This is realized by entiling incentives block frequency with the feature of phase, thus the efficiency of available frequency band is greatly enhanced.In decoding module, to improve the target detection accuracy, this paper synthesizes the spectrum correction theory, all-phase point pass filter design theory, all-phase FFT spectral analysis theory and pattern classification techniques for frequency and phase information extraction. Among them, with the aid of spectrum correction, the proposed scheme can easily detect the frequency and phase values of SSVEP under non synchronous sampling; With the help of all-phase point pass filter, those interferences outside the stimulus frequency band will be attenuated and thus the signal quality will be enhanced accordingly; In utilization of all-phase FFT’s and inhibiting spectral leakage, the interference between neighboring frequency can be suppressed. Morover, the all-phase FFT’s propery of ‘phase invariant’ also helps to enhance phase estimation precision and to reduce the computational complexity; Finally, target recognition can be completed through combining the nearest neighbor classification method with the above frequency and phase based decoding measures.The above methods’ feasibility in extracting frequency and phase characteristics of SSVEP is also proved through theoretical analysis and simulation. In addtion, by means of acquisiting of field data and the further experimental analysis, this disseration will show that the proposed scheme can remove the incapability in extracting SSVEP phase information under non synchronous sampling, which greatly eases the restrictions on target stimulus frequency and increases the amount of realizable targets. Hence, the proposed SSVEP-BCI has a wide range of potential applications.