Research On Brain-computer Interface Of Upper Limb Motor Imagery Based On Functional Near-infrared Spectroscopy

In brain-machine interface(BCI)research,relative to electroencephalography(EEG)and functional magnetic resonance imaging(f MRI),functional near-infrared spectroscopy(fNIRS)is a newer brain imaging technique for observing brain functions.There are potential research prospects in BCI.Motor imagery(MI),as an experimental strategy that can improve the brain mechanism,is widely used in BCI under the guidance of "mind" control.In specific research,on the one hand,because different experimental paradigms need to use more matching algorithms to identify in order to obtain better accuracy,on the other hand,the BCI experimental paradigm itself has difficulties to consider and whether it can be imagined.Therefore,this paper mainly focuses on the experimental paradigm design and algorithm application of the brain-computer interface.The specific experimental task is to perform the actual and imaginary movement of the upper limbs.The purpose is to lay the foundation for the future study of the multi-task movement of online direct brain-controlled robots.The main work carried out in this paper is as follows:(1)Based on the fNIRS signal,the right and left arm flexion and extension movement experiments are designed,and the AdaBoost.M1 algorithm in ensemble learning is used to identify the task signal.Because fNIRS measures the inherent delay property of blood oxygen metabolism,it mainly extracts the time-domain characteristics of data from 2 to 12 seconds after the start of the task for classification.At the same time,conventional SVM,LDA,KNN and NB algorithms are used for comparative analysis.The results show that it is feasible to apply the arm flexion and extension movement paradigm to the fNIRS-BCI research.The AdaBoost.M1 algorithm can effectively identify task signals,and its recognition ability is generally better than the conventional Classification algorithm.The flexion and extension of the arm in this chapter is the research basis for the future large-scale displacement movement of the brain-controlled mechanical arm,and it also lays the foundation for the further study of hand movement in chapters 3 and 4.(2)From the perspective of real life,the experimental paradigm of stone cloth scissors based on fNIRS was designed,and the time domain of task data was extracted from the two time windows of 3-8 seconds and 3-12 seconds Features,SVM and LDA were used to identify the features of the tasks,and the brain activation topographic maps corresponding to different tasks were compared and analyzed.The results show that the rock-paper-scissor game is feasible as the fNIRS experimental paradigm,and provides a control idea for the grasping movement of the brain-controlled manipulator under this paradigm.(3)From the perspective of communication methods,six types of one-hand sign language task recognition research based on fNIRS are designed.The time domain characteristics of task signals are extracted from the 3-12 th time window after the start of the task.SVM,LDA,HMM,NB,KNN and AdaBoost.M1 algorithms respectively identify the actual motion,imaginary motion,and actual motion vs imaginary motion of the task,and conduct a comparative study on the areas of the brain activated by performing six types of tasks.The results obtained show that the one-handed sign language task is feasible as the fNIRS-BCI experimental paradigm,and can provide richer control commands for the multitasking movement of the brain-controlled manipulator using this paradigm.In summary,the research content can provide a reference for the future multi-task motion control of brain-controlled robot upper limbs.