Analysis Of Motion Imaginary Signal Based On EEG And FNIRS

Brain Computer Interface(BCI)is a technology that connects human brain with computers or external electronic devices.BCI systems recognize one’s intention through collecting and analyzing bio-signal generated by brain activity,and further control external devices accordingly.The emergence of BCI technology not only brings tremendous help to the handicapped people and stroke patient with hemiplegia,greatly enhancing the quality of their life,the widely applications in military,entertainment and other fields also provide a new method for the field of artificial intelligence in the future.Currently most traditional BCIs rely on the analysis of single EEG,which remain some limitations in the noise removal,feature extraction and poor classification effect.In order to address the above problems,the following works have been done in this study:(1)Multimodal BCI based on concurrent EEG and f NIRS techniques is employed in this study.According to the characteristics and acquisition requirements of EEG and f NIRS signals,an EEG-f NIRS compatible cap was designed,and a suitable experimental paradigm was applied for data collection.In this paper,an algorithm integrating the improved wavelet threshold and Empirical Mode Decomposition(EEMD)method was applied to denoise the EEG signal.The proposed algorithm demonstrated a higher signal-to-noise ratio(SNR)compared with the traditional denoising methods and the single denoising methods.For f NIRS signal,the changes of oxy-hemoglobin(Hb O)and deoxy-hemoglobin(Hb R)of brain tissue were computed after filtration according to Beer-Lambert law.(2)Common spatial pattern(CSP)and regularized common spatial pattern(R-CSP)were respectively employed to extract the features of f NIRS.Comparison result between CSP and R-CSP showed that CSP yielded better performance over R-CSP,which was selected as the feature selection method for the classification of the multimodal signals.(3)To address the limitations of traditional support vector machine(SVM)in the selection of penalty parameter and kernel parameter,this paper proposed artificial fish swarm algorithm(AFSA)to search for the optimal parameters in SVM classifier.Features of the signals acquired during binary motor imagery tasks were extracted by CSP method.Comparison between the proposed method and traditional SVMdemonstrated the superiority of the algorithm in this paper.(4)At last,the CSP algorithm was used to extract the features of EEG and f NIRS signals for classification of the binary motor imagery tasks,respectively.Then the combined features of EEG and f NIRS were classified by the CSP algorithm.The experimental results revealed that the classification performance of the multimodal BCI outperformed the single modal BCI.