Research On Upper Limb Rehabilitation Technology Based On BMI Multi-channel FES

Stroke is an acute disease where brain tissue is damaged.In recent years,the incidence of stroke in China has not only shown a downward trend,but the disability rate is also increasing.Traditional upper limb rehabilitation techniques have long and boring rehabilitation cycles,expensive rehabilitation and poor rehabilitation effects.Based on the above problems,this article uses Functional Electrical Stimulation(FES)as the external control link in the Brain Machine Interface(BMI)link,and uses weak current to stimulate the muscles when rebuilding the patient’s neuronal pathways in motor imaging.Contraction improves the healing effect and promotes permanent healing.In this paper,the abovementioned BMI-FES upper limb rehabilitation technology is researched.The core part of BMI-FES upper limb rehabilitation technology is Electroencephalogram(EEG)signal preprocessing,EEG signal feature extraction and EEG signal classification.Aiming at the poor effect of traditional denoising algorithms on the artifact removal in EEG signals,this paper proposes a denoising algorithm combining empirical mode decomposition algorithm and improved independent component analysis algorithm to deal with noise and artifacts in original EEG signals.Remove traces.Aiming at the problem that the EEG feature information extracted by the traditional EEG signal feature extraction algorithm is too single,first of all,this paper uses the dual-tree complex wavelet transform to reconstruct the EEG signal of different frequency bands,and extract the energy mean feature,and then the reconstructed The EEG frequency band continues to use the filter co-space algorithm to extract spatial features to obtain multi-domain fusion information features.Finally,the Light GBM(Light Gradient Boosting Machine)integrated algorithm is used for classification,and manual adjustment of the Light GBM algorithm parameters is easy to fall into the local optimum.Based on the genetic algorithm(GA)and the Light GBM algorithm,this paper proposes a GA+Light GBM Classification algorithm to improve the robustness and adaptability of the model.The traditional method and the method in this paper are used to denoise and remove artifacts from the same EEG signal,and the signal-to-noise ratio and mean square error of the two processed signals are compared.It is found that the EMD+Picard ICA proposed in this paper has a higher signal-to-noise ratio.The mean square error is lower.To classify the EEG signal data of the left and right hand motor imaging of 8 strokes,and to compare with the traditional method,the GA+Light GBM classifier proposed in this paper has an average accuracy rate of about 88% among 8 subjects.The recognition accuracy is the highest,and the accuracy variance of different subjects is smaller.This article designs an experimental paradigm for EEG signal acquisition,upper limb rehabilitation actions and feedback experiments for stroke upper limb rehabilitation.Based on the STM32F103ZET6 SCM,a 4-channel FES lower computer is designed.Based on Open Vi BE and Python,an experimental platform for online EEG signal recognition and rehabilitation was built.This article also discusses the impact of visual feedback on accuracy.The accuracy rate without feedback is 75.34%.Compared with the situation with feedback,the accuracy of the subjects is improved by about 3%.The accuracy of online recognition with electrical stimulation and visual feedback was 82.1%.After four weeks of treatment with the brain-computer interface rehabilitation system,the number of neurons involved in the imagination increased,the FMA score increased,and the energy of the EEG signal increased compared with the initial stage of treatment,indicating the effectiveness of the system in this article.