Research On Decoding Of Natural Hand Movement Based On EEG

Natural human-computer interaction is the core of the integration of humans and robots.Brain computer interface(BCI)systems such as intelligent prostheses,exoskeleton robots and rehabilitation robots based on motor imaginary,steady-state visual evoked potentials and event-related potentials have achieved certain research results in assistance and motor function rehabilitation for the disabled.However,these are still have the problem of unnatural brain-computer interaction,such as using of foot movement imagination to control the function of neuroprosthetics.Natural and intuitive BCI control can promote process of users and the BCI system,which can enable users to actively participate and improve rehabilitation effects.EEG-based natural action decoding provides a new way for realizing intuitive and natural BCI control.Decoding reach-and-grasp actions from electroencephalograms(EEG)is of great significance for the realization of intuitive and natural neuroprosthesis control,and the recovery or reconstruction of hand functions of patients with motor disorders.However,there are few studies on using EEG decoding natural hand movements and are not in-depth enough.Most studies only tried to classify two or three natural hand movements,and the decoding performance is limited.Relevant studies have shown that the movement information contained in time-domain movement-related cortical potentials(MRCPs)can improve the performance of EEG decoding for natural movements.This research is oriented to natural human-computer interaction,and develops research of decoding reach-and-grasp actions based on EEG.In this study,five types of reach-and-grasp actions of unilateral limbs were taken as the decoding objects,namely palmar grasp,pincher grasp,push grasp,twist grasp and plug grasp.The core content of this research includes:(1)Facing natural human-computer interaction,the EEG experimental platform for reach-and-grasp actions is designed.The experimental data acquisition system was built based on the design of software and hardware,and a VC++ software was developed for synchronous data recording.(2)Designed an experimental paradigm based on motor-related natural movement EEG,and recruited volunteers to carry out experiments;we preprocessed the collected EEG data,and extracted effective features,and designed decoding models.(3)The results of neurophysiological analysis and decoding performance are reported.The results of MRCPs and EEG Source Imaging analysis show that significant differences appears between the five different grasping and non-movement conditions,and different grasping actions show different activation patterns,which prove the feasibility of decoding reach-and-grasp actions.MRCPs amplitude were selected as the features for offline analysis.The binary classification results show that the average peak classification accuracy of grasping condition and non-movement condition is 75.06±6.8%,and the average classification accuracy between grasping condition and grasping condition is 64.95±7.4%.The multi-class decoding results show that the average peak classification accuracy of five different natural hand continuous grasping actions is 36.7±6.6%.The results of binary and multi-class classification are better than the chance level.Five different natural grasping movements of a unilateral hand can be successfully decoded based on the low-frequency time-domain EEG.In this paper,we successfully decoded five different reach-and-grasp actions based on MRCPs.This study has provided a new idea for the realization of natural and intuitive BCI control,which increased the possibility of intuitive and natural neuroprosthesis or rehabilitation robot control.This work is helpful for future research on decoding hand movement information and is important and significant for future application of BCI.