Research On Motor Imagery EEG Recognition Of Unilateral Upper Limb Based On Robotic Arm Control

The robotic arm instead of human being engaged in exhausting labors and dangerous work has been widely used in numerous domains of production and life,of which control mode gradually evolves to the direction based on biological signal control.As the brain computer interface(BCI)technology develops,control for the robotic arm based on motor imagery EEG has been realized.However,this technology has not been mature and presented unnatural control problem,for example,the robotic arm is operated and controlled through motor imagery EEG such as left hand,right hand,both feet and tongue.Because the robotic arm has the unique of simulating actions of human arm,the most comfortable control mode for the operator is to control motion of corresponding robotic arm by imaging actions of his arm,that is,to realize natural control for robotic arm by human.Therefore,realizing natural control for robotic arm based on motor imagery EEG is an inevitable development trend.However,a series of technical problems must be solved to realize such system,in which the key technology is to design a natural motor imagery task mode and make effective decoding.At present,motor imagery EEG based BCI mainly focuses on studying EEG processing algorithm and online BCI system.Most studies on EEG processing algorithm have used public data of BCI competition,and online BCI system is generally provided with specific motor imagery task mode to collect data,these data has a common point,that is,motor imagery based on different body parts such as hands,feet and tongue.These data has significant EEG recognition feature which is not applicable to natural control for robotic arm.The effective recognition for unilateral upper limb motor imagery EEG can realize natural control for BCI based robotic arm.However,there are few studies about recognition for unilateral upper limb motor imagery EEG at home and abroad;the accuracy of recognition shall be future improved,and few achievements have been applied to control for robotic arm.Under support of NSFC Project,this paper has,for key technology in natural control for robotic arm based on motor imagery EEG,conducted a study on recognition for unilateral upper limb motor imagery EEG.The key content is this study include:(1)An experiment of unilateral upper limb motor imagery EEG has been designed for natural control for the robotic arm.We have,first,constructed two sets of EEG acquisition systems;second,made electrode selection plans respectively for above two system based on their features and physiology of motor imagery;finally,designed the task mode for right upper limb motor imagery and experimental paradigm.(2)The processing algorithm of unilateral upper limb motor imagery EEG has been studied.In addition to typical common spatial mode algorithm,this paper has proposed to use phase synchronization information as a feature of unilateral upper limb motor imagery EEG,but also introduced CNN algorithm.(3)The experimental results of motor imagery EEG have been analyzed.For results based on non-portable and portable EEG acquisition system,we have made vertical and horizontal comparison and analysis.The vertical analysis has verified physiology of the unilateral upper limb motor imagery and compared advantages and disadvantages of different algorithms.The horizontal analysis has compared loss of different algorithms in classification performance in two EEG acquisition systems and analyzed phase correlation between motor imagery and brain regions.This study has provided a new idea for recognition of motor imagery EEG and natural control for the robotic arm,and made natural control for robotic arms based on unilateral upper limb motor imagery EEG be possible,which are important and significant for future application of BCI.