Research On MI-BCI Algorithm And Its Application Based On Particle Swarm Optimization

Since the beginning of the 21 st century,brain-computer interface(BCI),as the focus of brain science research,has attracted more and more attention.It has been successfully applied in motor function rehabilitation,speech signal recognition and automatic control system.And with the continuous development of this technology,it is expected to be more widely used in clinical medicine,with broad prospects for development.Brain-computer interface technology based on motor imagery(MI-BCI)can decode and recognize the signal by collecting the motor imagery-based EEG(MIEEG)during imaginary movement,and then realize the control function of external equipment.It provides a new way to reconstruct the function for patients who have lost movement or other functions.Due to the high data dimension and low signal-to-noise ratio of EEG,it is difficult to guarantee the classification accuracy under the existing signal processing framework,and it takes a long time to train and test,which limits the practical application of related technologies.Therefore,an efficient signal processing framework has become one of the current research hotspots.Aiming at the bottleneck in the development of MI-BCI technology,this paper proposes an efficient signal processing framework based on multi-level particle swarm optimization(MLPSO)to reduce the dimension of feature vector,improve the accuracy of algorithm classification,and enhance the applicability of the system.Improved Stransform and Bayesian linear discriminant analysis are used for feature extraction and classification,and MLPSO is used for feature optimization.Three optimization schemes of channel selection,feature selection and joint selection of channel and feature based on MLPSO are designed,and the effectiveness of the related algorithms of this framework is verified by using common standard data sets.After optimization,the accuracy rate of the best classification is 99%,which is 10% higher than that of the method before optimization.Kappa value is 0.2 higher than that of the method before optimization,reaching 0.98,and the specificity and accuracy are 100%.In the case of using less than 10.5% of the original features for classification,the classification accuracy is still up to 99%,but the classification time is reduced by more than 90%.By comparing the optimized classification indexes,the effectiveness of this framework is confirmed,which can be used as a reference for the research of BCI real-time system.The classification accuracy of this algorithm is also higher than that of the current literature using the same data set.In addition,this paper explores the location distribution of the best channel combination based on MLPSO channel selection optimization,and compares it with the related research of event-related potential distribution map and brain function analysis in physiology.The experimental results are highly consistent,which confirms that the channel combination is the most relevant channel distribution for motor imagery task,and illustrates the proposed method The proposed channel optimization scheme is scientific.In addition,this study also uses the self-collected data to verify the feasibility of this framework in the actual system.Stroke is an acute cerebrovascular disease.Due to the lack of effective treatment,it is a serious threat to the health of Chinese people and has become the first cause of death in China.Therefore,the establishment of quantitative and objective evaluation criteria is of great significance for all kinds of stroke rehabilitation treatment.At present,MI-BCI technology as a supplementary treatment for stroke rehabilitation has been applied to a certain extent in clinical,but it has not been reported as a treatment evaluation method.The evaluation of rehabilitation treatment of stroke by functional magnetic resonance imaging(f MRI)has objective and quantitative characteristics and is of great clinical significance.But f MRI has high requirements for equipment and personnel,and the cost is high,so f MRI cannot be used as a conventional clinical evaluation method.MI-EEG signal can objectively reflect the activity of the brain nervous system,and it is low cost and high clinical popularization rate.Therefore,we try to use MI-EEG signal as the evaluation method of clinical rehabilitation treatment of stroke,and do some exploratory research.This paper proposes that the channel location combination distribution which is most related to motor imagination obtained by channel selection based on MLPSO is used as an evaluation method for stroke nerve remodeling.The feasibility of this method in practice is verified by self-collected data,and the most common data set of left and right hand dysfunction is completed,which provides a reference for the subsequent research.