Intelligent Auxiliary Control Of Manipulator Based On Human EMG Signal

With the rapid development of smart wearable devices,for people with dyskinesias or amputations,assistive devices based on bioelectric signals or natural control of prostheses can help them to better work and live.Therefore,the human-computer interaction technology of electromyogram(EMG)signal has become one of the research hotspots.In response to the need for high-precision,real-time intelligent prosthetic control based on human bioelectric signals.This paper focuses on the intelligent auxiliary control of the robotic arm based on the human EMG signal.The main research work and research content are as follows.(1)Aiming at the problems of complex circuit design,poor anti-interference ability,high power consumption,and portability of traditional EMG acquisition devices.This paper studies and implements an EMG signal high-sensitivity real-time acquisition module based on ADS 1299.First,in the hardware design of the acquisition module,the preliminary processing of the original EMG signal is completed by the design of the electrode position and the filter circuit.Then,this paper uses the integrated analog front-end chip ADS 1299 to modulate and convert the EMG signal.And through high-efficiency FPGA controller and power management,the signal is processed with low power consumption.Finally,through the analysis of the principle of EMG acquisition and the test of each module,it is verified that the acquisition module in this paper has small size,low power consumption and good anti-interference ability.(2)Aiming at the problems of current EMG signal classification algorithms that are greatly affected by noise,low recognition accuracy,and complex network models.This paper studies and implements an EMG signal denoising and classification recognition method with stacked long and short-term memory networks and residual-attention mechanism.First,the noise is classified by the combination of the stacked long and short-term memory network and the global attention mechanism,which effectively realizes the denoising processing of the EMG signal.Then,classify the preprocessed EMG signal.Using the EMG signal recognition algorithm based on the residual-attention model to solve model degradation,the attention mechanism reduces the computational burden of data,accelerates network training,and improves the accuracy of the model.Finally,the paper uses multiple public data sets and self-built EMG data sets for experimental verification.The results shown that compared with other typical algorithms,the residual-attention-based recognition model studied in this paper also has the advantages of shorter training time and higher accuracy.(3)In order to verify the effectiveness of the studied intelligent auxiliary control method of manipulator based on human EMG signal,this paper designs and implements a software module for manipulator control based on EMG signal.This module realize the functions of real-time display of original EMG signal waveform and denoising waveform,classification recognition,recognition accuracy and control command generation,etc.In this paper,the recognition algorithm based on the residual-attention model based on the software upper computer loading research of this module.Several group experiments show that the recognition method is more accurate,and the existing robotic arm platform is used to verify the real-time and reliability of the EMG signal control method based on deep learning.