Research About Hybrid Brain Computer Interface Based On P300 And SSVEP

In recent years, scientists have proposed a new communicative means between brain and outside environment, in order to help disabled patients with a normal brain to live independently. The new mean is brain computer interface. Brain computer interface avoids the pathological peripheral nervous system or muscle tissue, control equipment or communicate with the external environment by detecting the brain electrical activity. At present, a single category of EEG signals is used as the input signal of BCI. However, Brain thinking is highly complex. Single category of EEG signals can only be used to identify a simple thinking task, and can not meet the needs of practical applications. Hybrid BCI has become an important direction of BCI research.The traditional character spelling based on P300 is one of the classic BCI, which determines the row and column of the character through the recognition of P300 feature. In order to improve performance of the character spelling BCI, this paper attempts the hybrid BCI. The SSVEP is introduced into the traditional character spelling system, mixed with P300 parallel control the hybrid BCI. More details as below:(1) This paper designs a hybrid character spelling BCI based on P300 and SSVEP. The hybrid BCI has two forms flashing, one of them is the block under the character periodically flashing and different column has a different frequency, the other is red block random flashing in the row, thus can induce P300. At the same time, a contrast paradigm based on P300 is designed.(2) Respectively feature extraction and classification of EEG in Hybrid BCI and P300 BCI. Wavelet decomposition and time domain entropy can extract P300 feature, then the feature can be classified by SVM, thus can identify the row of target characters; The reference signals are constructed with the same frequency of stimulation, then CCA is used to classify SSVEP, thus can identify the column of target character. Through row and column, the target character can be calculated. The contrast BCI based on P300 use the same P300 feature extraction and classification method with the hybrid BCI, thus can identify target character. The performance is analyzed by recognition accuracy and the information transfer rate, the results show the hybrid BCI based on SSVEP and P300 reduce stimulation time of single character, improve the recognition accuracy and the information transfer rate compared with the BCI based on P300. Hybrid BCI based on a variety of EEG signals can improve the performance of traditional BCI system. Hybrid BCI has a good application prospect.(3) ICA-EMD method is proposed to automatically eliminate EOG artifacts from EEG signals. The simulation results shows, ICA-EMD can remove EOG artifacts and get better denosing effects. The progress in removal EOG can be realized without EOG reference signals. The method removal EOG artifacts proposed in the paper is suitable for online BCI system.