The Method Of Motion Imaging EEG Signal Recognition And Its Application In Wheeled Robot

Brain-computer Interface(BCI)is a kind of does not depend on the brain communication control system composed of peripheral nerves and muscles,with the continuous development of BCI technology,has been gradually implemented by using the BCI technology makes the Brain directly communicate with the outside world,the major breakthrough to can't normal communication or paralysed patients brought the gospel.In order to strengthen the self-independence ability of such people and improve the convenience of communication,it is very important to carry out classification research on electroencephalon in motor imagination and design an online BCI platform based on electroencephalon motor imagination.In order to quantitatively describe the BCI system,this paper will study in depth four kinds of motion imaginary EEG signals:forward,backward,left and right.Firstly,the experimental scheme was designed to collect the motion imaginary EEG signal,and the original EEG signal was denoised by the improved threshold method of wavelet.Secondly,the multi-feature fusion method of EEG signal feature extraction was used for feature extraction,and then Principal Component Analysis(PCA)was used to reduce the dimension of multi-dimensional feature vectors.Finally,Class Driven Feature Selection and Classification Support Vector Machine(CFSC-SVM)was used for classification processing,and the offline analysis results were applied to the online BCI platform based on EEG imagination,so as to realize the control of the wheeled robot.The specific research contents of this paper are as follows:Based on the BCI technology,the experimental scheme was designed by ourselves to collect the cerebral motor imagination EEG signals,and then convert them into digital signals and input them to the computer for subsequent analysis and processing.Considering that there are some interference factors in the process of acquisition,this paper adopts the improved threshold method of wavelet to preprocess the acquired EEG original signal for noise reduction,and verifies the effect of noise reduction by the improved signal to noise ratio and mean square deviation.The wavelet packet energy,the wavelet packet energy of different frequency bands extracted by the autoregressive model algorithm and the correlation coefficient of time series are sequentially connected to each other to form the EEG signal characteristics of multi-fusion features,so as to ensure the accuracy and authenticity of data information.PCA dimensionality reduction method was used to reduce the dimensionality of these high-dimensional data,which laid a foundation for the subsequent classification and recognition.The CFSC-SVM algorithm was used to classify and identify the motion imaging EEG signals in different states.The superiority of CFSC-SVM in recognition performance was verified by classification accuracy,accuracy standard deviation and area under ROC curve.Meanwhile,the superiority of multi-fusion feature extraction method was verified by comparing the recognition accuracy of single feature vector and multi-fusion feature vector by CFSC-SVM classification.The offline analysis results were applied to the online BCI design based on the imagination of electroencephalon motion.The control module and the driving system were combined to realize the control of the wheeled robot.The experiment was conducted on different subjects,which verified the feasibility of the online BCI platform designed in this paper.