基于表面肌电信号上肢运动意图识别
|
摘要
基于表面肌电信号的上肢运动意图识别的研究是上肢康复机器人及智能假肢精确控制的基础。在Fourier M2上肢力反馈康复平台上进行了4种上肢动作意图的识别研究。针对上肢表面肌电信号存在的高频噪声、基线漂移以及50 Hz工频干扰采用交叉去噪方法对噪声进行有效的滤除。为进一步提高上肢意图识别的准确率,分别在时域和频域对肌电信号进行特征值提取,构造特征向量。选用BP神经网络作为模式分类器实现4种上肢意图动作的分类。实验验证该分类器对10位测试者右肢的4个意图动作(向左、向右、向前、向后)的识别率为90%。
The research on upper limb motion intent recognition based on surface EMG signals is the basis for accurate control of upper limb rehabilitation robots and intelligent prostheses. And identification exploration of the four kinds of upper limb movement intentions is done on the Fourier M2 upper limb force feedback rehabilitation platform. For the high frequency noise, baseline drift and 50 Hz power frequency interference of the upper limb surface EMG signal, the cross noise denoising method is used to effectively filter out the noise. In order to further improve the accuracy of upper limb intent recognition, the eigenvalues of the EMG signals are extracted and constructed in the time domain and the frequency domain respectively.The BP neural network is selected as the pattern classifier to classify the four kinds of upper limbs. The experiment indicates that the correct recognition rate of the classifier reaches 90% on the four intended actions(left, right, forward, backward) of the right limb for the 10 testers.
The research on upper limb motion intent recognition based on surface EMG signals is the basis for accurate control of upper limb rehabilitation robots and intelligent prostheses. And identification exploration of the four kinds of upper limb movement intentions is done on the Fourier M2 upper limb force feedback rehabilitation platform. For the high frequency noise, baseline drift and 50 Hz power frequency interference of the upper limb surface EMG signal, the cross noise denoising method is used to effectively filter out the noise. In order to further improve the accuracy of upper limb intent recognition, the eigenvalues of the EMG signals are extracted and constructed in the time domain and the frequency domain respectively.The BP neural network is selected as the pattern classifier to classify the four kinds of upper limbs. The experiment indicates that the correct recognition rate of the classifier reaches 90% on the four intended actions(left, right, forward, backward) of the right limb for the 10 testers.
引文
[1]Chu J U,Moon I,Mun M S.l.A Supervised Feature Projection for Real-time Multifunction Myoelectr Hand Control[C]//ConfProc IEEE Eng Med Biol Soc.2006,1(24):17-20.
[2]Reddy NP,Gupta V.Toward direct Biocontrol Using Surface EMG Signals:Control of Finger and Wrist Joint Models[J].Med Eng Phys,2007,29(3):398-403.
[3]Tsuji T,Fukuda O,Kaneko M,et al.Pattern Classification of Time-series EMG Signals Using Neural Networks[J]International Journal of Adaptive Control and Signal Processing,2000,14(8):829-848.
[4]Englehart K,Hudgin B,PA Parker.A Wavelet-based Continuous Classification Scheme for Multifunction Myoe-lectric Control[J].IEEE Transactions on BME,2001,48(3):302-311.
[5]Martinez Alajarin,Ruiz Merino.Wavelet and Wavelet PacketCompression of Phonocar Diograms[J].Electronics Letters,2004,40(17):1040-1041.
[6]Gui Zhong-hua,Han Feng-qin.Wavelet Packetmaximum EntropySpectrum Estimation and Its Application in Turbine s Fault Diagno-sis[J].Automation of Electric Power Systems,2004,28(2):62-66.
[7]Nasiri,Poshtan,Kahaei,et al.Wavelet Packet Decomposition as a Proper Method for Fault Detection in Three Phase Induction Mo-tor[C]//Proceedings of the IEEE International Conference on Mechatronics2004,7(3):13-18.
[8]Stulen F B,De Luca C J.Frequency Parameters of the Myo-ele ctric Signal as a Measure of Muscle Conduction Velocit[J].IEEE Trans Biomed Eng,1981,28(7):515-523.
[9]GEIJI O,SHOZO S,YUKIHIRO A,et al.Effects of Back Propagation Operation for Fresnel Hologram Reconstruction Images[J].ITE Technical Report,1982,6(23):7-12.
[10]RUMERLHAR D.E.Learning Representation by Backpropagating Errors[J].Nature,1986,323(3):533-536.
[11]GIROD L,ESTRIN D.Robust Range Estimation Using Acoustic and Multimodal Sensing[C]//Proceedings of2014 IEEE/RSJ International Conference on Intelligent Robots and Systems.Maui:IEEE,2014:1312-1320.
[2]Reddy NP,Gupta V.Toward direct Biocontrol Using Surface EMG Signals:Control of Finger and Wrist Joint Models[J].Med Eng Phys,2007,29(3):398-403.
[3]Tsuji T,Fukuda O,Kaneko M,et al.Pattern Classification of Time-series EMG Signals Using Neural Networks[J]International Journal of Adaptive Control and Signal Processing,2000,14(8):829-848.
[4]Englehart K,Hudgin B,PA Parker.A Wavelet-based Continuous Classification Scheme for Multifunction Myoe-lectric Control[J].IEEE Transactions on BME,2001,48(3):302-311.
[5]Martinez Alajarin,Ruiz Merino.Wavelet and Wavelet PacketCompression of Phonocar Diograms[J].Electronics Letters,2004,40(17):1040-1041.
[6]Gui Zhong-hua,Han Feng-qin.Wavelet Packetmaximum EntropySpectrum Estimation and Its Application in Turbine s Fault Diagno-sis[J].Automation of Electric Power Systems,2004,28(2):62-66.
[7]Nasiri,Poshtan,Kahaei,et al.Wavelet Packet Decomposition as a Proper Method for Fault Detection in Three Phase Induction Mo-tor[C]//Proceedings of the IEEE International Conference on Mechatronics2004,7(3):13-18.
[8]Stulen F B,De Luca C J.Frequency Parameters of the Myo-ele ctric Signal as a Measure of Muscle Conduction Velocit[J].IEEE Trans Biomed Eng,1981,28(7):515-523.
[9]GEIJI O,SHOZO S,YUKIHIRO A,et al.Effects of Back Propagation Operation for Fresnel Hologram Reconstruction Images[J].ITE Technical Report,1982,6(23):7-12.
[10]RUMERLHAR D.E.Learning Representation by Backpropagating Errors[J].Nature,1986,323(3):533-536.
[11]GIROD L,ESTRIN D.Robust Range Estimation Using Acoustic and Multimodal Sensing[C]//Proceedings of2014 IEEE/RSJ International Conference on Intelligent Robots and Systems.Maui:IEEE,2014:1312-1320.