Research On Multi-modal Brain-computer Interface Based On EEG-NIRS

The traditional brain-computer interface(BCI)research mainly focuses on the analysis of single-mode brain signals.The single-modal signals contain limited information,and there are problems such as difficulty in improving classification accuracy and unstable classification.In order to improve the classification performance,this paper mainly carried out the following research work:(1)Direct brain-controlled multi-robot cooperation task based on EEG.Using the steady-state visual evoked potential(SSVEP)with good stability,accuracy and rapidity,select the typical correlation analysis(CCA)algorithm,and design two kinds of brain switches to realize the brain-controlled multi-robot collaboration(MRC)task,We obtained an effective brain control MRC method.(2)The above research is based on multi-channel EEG-BCI,which is not conducive to the development of portable applicability of BCI.Therefore,the EEG-BCI study with less channel is carried out.Two schemes are used: Wavelet Packet Decomposition(WPD)method,Hilbert Huang Transform(HHT)method.The original 3-channel data is expanded to 15 channels,and the features are extracted by the common space model(CSP),and the two classifiers are used for identification.The results show that both schemes are effective and the HHT’s effect is better than WPD.(3)Due to the problem of low accuracy and insufficient stability reflected by EEG-BCI,we carry out another modal BCI technology research--BCI based on fNIRS.Near-infrared spectroscopy(NIRS)signal was used as the input signal of BCI to obtain the corresponding concentration change of oxyhemoglobin(HbO).After preprocess and feature extraction,three classifiers were used for pattern recognition.This study proved that NIRS-BCI can obtain better accuracy and stability compared with EEG-BCI.(4)In order to solve the limitations of each single-mode BCIs,a BCI study combining EEG and NIRS modes is carried out.Considering the acquisition characteristics of the two signals,design a joint acquisition experimental paradigm,conduct experimentsand collect data.The EEG and fNIRS signals are processed separately,and then the feature fusion strategy is used to fuse the two features into a new feature vector,and then the classification is verified.The result shows that the BCI classification performance of the two modes is better than that of the two single modes.