Comparative Research On Decomposition Algorithms Of High-density EMG For Blind Source Separation

Electromyogram(EMG)signal is a superposition of Motor Unit Action Potential(MUAP)generated by motor neuron discharge events,which encodes the neural activity of the spinal cord circuit and can reflect high-level neural control from the brain signal.Motor neuron discharge activity can provide a theoretical reference for the neural control of movement,can provide clinical insights into neuromuscular injury,and can also be used as an interface signal for human-computer interaction.Therefore,it is important to decompose and extract the discharge information of individual Motor Unit(MU)from the EMG signal,and it is also an important research field of computer information processing.Through the improvement of high density electromyography(HD EMG)decomposition algorithm and a series of experiments,this thesis systematically evaluated three different EMG decomposition algorithms based on Independent Component Analysis(ICA)in each The accuracy and yield of decomposition under various conditions,the algorithm is tested on the HD EMG signal obtained by simulation and experiment,the signal has a series of steady-state muscle contraction levels,and has different contraction levels,similar to the dynamic characteristics of muscle contraction.It provides effective technical support for the selection of ICA algorithm in HD EMG decomposition algorithm.The main work is as follows:(1)Explain the research status and basic principles of the HD EMG decomposition algorithm,analyze the latest research on the decomposition algorithm,summarize the shortcomings of the current decomposition algorithm research,and provide ideas for the subsequent research algorithm improvement.(2)In view of the single problem of ICA algorithm in HD EMG decomposition algorithm,Infomax and RobustICA algorithms were introduced into EMG decomposition,HD EMG decomposition algorithm was improved,and high-density EMG signal data was obtained through simulation and actual human experiment analysis.And the technology implements three different ICA-based HD EMG decomposition algorithms of FastICA,Infomax and RobustICA,obtains the performance data of these three algorithms under different signal conditions,and analyzes the characteristics of the three algorithms to enrich the EMG decomposition algorithm ICA algorithm selection method.(3)Focusing on the problem of lack of systematic performance evaluation in the HD EMG decomposition algorithm,by processing the analysis data,the performance of the three algorithms under different signal complexity and signal quality was systematically evaluated,including accuracy,algorithm consistency rate,Forming a systematic performance evaluation method of HD EMG decomposition algorithm.(4)In view of the lack of reference selection of the blind source separation ICA algorithm under the specific conditions in the HD EMG decomposition algorithm,the expansion factor R of the three algorithms was verified,and the analysis of the three algorithms under simulated data and real data was analyzed.Performance,provides technical support for the algorithm selection of some specific applications.