Auxiliary Force EEG Recognition Research Of Upper Limb Rehabilitation Robot For Stroke Patients

System engineering integrates many disciplines including systematics,cybernetics and informatics,which is widely used in rehabilitation medicine.Based on clinical rehabilitation medicine and assistive rehabilitation,the upper limb rehabilitation robot provides a new technical path to patient's rehabilitation.Its research involves the intersection of rehabilitation medicine,robotics,biomedicine,artificial intelligence and other subjects.Especially,the bioelectric signals provide a new auxiliary program for stroke rehabilitation therapy.The rehabilitation stage of stroke patients is mainly divided into disease stage,soft palate stage,hemorrhoid stage and rehabilitation stage.Rehabilitation robots are required to provide different optimal assisting forces at different stages.The rehabilitation training process of most rehabilitation robots is based on the fixed parameter values set by the rehabilitation physician.Combined with the patient's active consciousness EEG signal,the new rehabilitation robot control system makes it possible to optimize the assisted motion control of the rehabilitation robot.The current EEG-based rehabilitation aids can provide basic direction identification of discrete logic control instructions,such as left and right.The application of auxiliary forces need to identifiy magnitude of the force in addition to the direction.The analysis and identification of auxiliary forces based on EEG provides a new way for the rehabilitation robot assist in the patient's different recovery stages.This thesis aims at the recognition of three-category auxiliary force EEG during the rehabilitation training process.Firstly,design the experimental paradigm of different auxiliary force training.Secondly,collect the auxiliary power EEG data to determine the key frequency bands associated with the auxiliary force.Then extract the effective features.Finally use SVM to perform the three-class assisted force identification on EEG data.The main research results are as follows:(1)According to the requirements of auxiliary forces in different rehabilitation periods,the experimental paradigms of the three training states are designed to obtaindifferent auxiliary EEG data,and the data is pre-processed to serve as the data basis for identification.(2)The approximate entropy and sample entropy characteristics of the assisted EEG are extracted and statistically analyzed.On the basis of the analysis of model algorithms,it is verified that there are certain specific differences in the three types of auxiliary EEG complexity.(3)Identified the frequency bands associated with the rehabilitation training assistance.The auxiliary force reflecting in EEG has the characteristics of the highly non-stationary low signal-to-noise ratio.The frequency bands of the auxiliary features of the EEG associated with the rehabilitation process are used as the entry point for EEG signal processing.Analyze the activated brain regions and identify the key channels.Finally,use WPT and power spectrum fusion methods to analyze the energy spectrum distribution of the key channel EEG.The key frequency bands with auxiliary force characteristics are 7.33-7.81 Hz and 8.30-8.79 Hz.(4)Analyze the activation and non-activation states of brain signals during the auxiliary force training.The wavelet energy threshold method and the two-level CSP algorithm are adopted to analyze the non-active state detection respectively.The final recognition accuracy rate of the two detection methods is obtained,and the best identification algorithm is found by comparing the correct rate.Based on CSP,the activated state is treated as a separate state,and the energy features are extracted.Then conduct multi-stage classification and recognition together with the auxiliary force 1 and auxiliary force 2 in the active state.The result is an average of more than80% of the classification accuracy.(5)Identify different auxiliary force EEG modes.Use the approximate entropy and sample entropy as features to classify the three types of auxiliary EEG,compare them with the energy features,determine the signal energy as the optimal classification feature of the auxiliary force EEG,and use the signal energy as the feature to identify the frequency bands of the assisting force EEG.The experiment results show that as the auxiliary force increasing,the EEG energy value will gradually move to the high frequency band,and when the auxiliary force reaches the maximum value,the EEG energy value reaches the highest,and the auxiliary forceEEG has the highest classification accuracy in the frequency bands of 4-8Hz and8-12 Hz,respectively reaching 78.84% and 80.31%.The ant colony algorithm is used to optimize the selection of the channel,which is proved that the channel optimization algorithm can further reduce the number of key channels and achieve reliable classification accuracy.