Research On Brain Computer Interface Technology Based On Time Frequency Space Feature Fusion

Brain-Computer Interface(BCI)(Brain-Computer Interface,BCI)is a new type of system that establishes a connection path between the brain and the external environment,and is the focus of future scientific and technological devel opment.BCI technology is of great significance in the field of rehabilitation engineering and pattern recognition.At present,BCI technology is widely used in various fields,such as environmental control and life entertainment.Among them,the research of Motor Imagery(MI)is a research hotspot.Therefore,it is of great scientific significance to study the classification of EEG signals based on motor imagination.In this paper,based on two sets of four-category motor imaging EEG data sets,in-depth research on feature extraction methods and classification algorithms based on motor imaging EEG signals,and at the same time realize the online EEG signal recognition system based on wavelet transform as the core alg orithm,the paper's The main contents are as follows:1.Based on BCI competition iv data set Data Set 2a and BCI competition iii data set Data Set iiia two sets of multi-class sports imaging EEG signal data sets,a sliding window-based time period selection algorithm is designed,based on the difference between different subjects Difference,select the time period that is most suitable for the subjects to deal with;2.In view of the current research in the field of motor imaging EEG signals,most of which are single feature extraction and more binary classification algorithms,this paper proposes a time-frequency space multi-class feature extraction algorithm based on wavelet transform,and extracts the time domain based on the conduction process of EEG Features;ERD/ERS phenomenon based on motion imagination to extract frequency domain features: second-order moment energy;based on the spatial correlation of EEG,through the Common Spatial Pattern(CSP)algorithm to extract spatial domain features,while the CSP algorithm in In multi-classification extension,a selection algorithm based on the best individual features is proposed to realize comprehensive analysis of EEG signals from multiple angles of time domain,frequency domain and space domain,thereby improving the recognition rate of EEG signals.3.For the current EEG signal multi-classification method in the One-Versu s-One strategy and One-Versus-Rest strategy problems: when multiple categoriesvote the same,it is not correct Recognizing the problem,in view of this phenomenon,this paper proposes a Wighted-Score mechanism based on Support Vector Machine(SVM).By using the results of the two classifiers in the training phase as the weighted score items in the test phase,finally The highest comprehensive score is the label of this category,and compared with the OVO and OVR strategies.The results show that the WS mechanism is better than other methods as a whole,and the average recognition rate has increased by 5%.At the same time,combining the feature extraction method and the classificatio n algorithm,the two data sets are compared.The results show that in data set one,the best average Kappa value of the method in this paper reaches 0.69,which is significantly improved from 0.57,the first place in the competition;On data set two,the average Kappa value of this paper reaches 0.77,and the a verage recognition rate reaches 0.83.The experimental results verify the superiority of this method in feature extraction and classification.4.Based on the feature extraction algorithm and classification algorithm proposed above,an online EEG signal recognition system is realized.This paper implements real-time transmission and reception of EEG data,online display of results,and Bluetooth transmission based on the Qt framework and the Lab Streaming Layer protocol widely recognized in the field of EEG signals.After offline experiment testing,BCI implemented in this paper The system has good performance and lays a technical foundation for the application of EEG in the real-time field.