Analysis And Research On Algorithm For Motor Imagery-based Brain-Computer Interfaces System

As a brand new communication option,brain-computer interface technology has been a hot topic in the field of science and engineering.It can analyze and identify the type of EEG signals generated by the human brain through the computer to complete the exchange of information between people and the outside world.For motor imagery-based brain-computer interfaces system,this paper is a study of the algorithm of feature extraction,selection and classification.Analyze the characteristics of motor imagery EEG from the time domain,the frequency domain and the time-frequency domain respectively by using the statistical analysis method.Combined with event-related potential and movement-related potential analysis methods,the ERD/ERS and movement readiness potentials of the mu and beta rhythm of motor imagery EEG signals are found.To distinguish different types of motor imagery EEG signals,this paper presents a multi-feature classification fusion algorithm,the common spatial pattern and wavelet packet decomposition method are used to extract the features of EEG signals,and the support vector machine is used to classify the EEG signals.In addition,this paper uses neural network algorithm to optimize the weights of three kinds of decision-making value,aiming for the best weight values and the fusion decision value.Comparison with several different algorithms,the correct recognition rate of the algorithm proposed in this paper is higher than other recognition algorithms.Aiming at the problem of feature redundancy and parameter optimization of SVM,feature selection algorithms and intelligent optimization algorithms are studied,respectively.Relief and gravitational search algorithm-based feature selection of EEG signals and parameter optimization of SVM are presented to obtain the best features and the best parameter and to improve the accuracy.Compared with particle swarm optimization algorithm and improved particle swarm optimization,gravity search algorithm and improved gravity search algorithm can find the optimal parameter value of SVM faster.And the feature extraction and recognition of multi-category EEG signals are completed through the "hierarchical" decision-making method.An experimental platform consisting of a Emotiv EPOC+ real-time raw EEG acquisition device,a computer,14 EEG signal sensors and a testee is set up,and experimental stimulation interface is designed.And the ERD/ERS of the mu and beta rhythm of two-class motor imagery EEG data are completed.Moreover,according to two experimental circumstances consisted of two-classification and four-classification,the feature extractions and classifications of EEG signals acquired by the EPOC+ are completed.