前臂肌电信号实时智能模式识别系统研究
|
摘要
随着康复医学的发展,康复机器人的研制越来越受到人们的重视,肌电信号模式识别系统是康复机器人的重要组成部分,通过肌电信号对病人的运动意识进行快速、准确地识别,是康复机器人完成康复训练的必要前提。本文通过对特征提取方法与模式识别方法的研究,结合肌电信号自身的特点,提出用于肌电信号模式识别的具体算法,并且使用真实的肌电数据对算法进行仿真和验证。根据康复机器人的需求,具体实施了该算法,设计并开发了一套肌电信号模式识别系统,通过对前臂肌电信号的检测与处理,实时地对腕部动作进行识别,并通过实验对算法进行验证。
本文首先介绍当前国内外对肌电信号模式识别研究的现状,研究应用于肌电的各种特征提取和处理方法以及模式识别方法,比较各自的优缺点。然后介绍肌电采集系统的结构和各部分需要考虑的问题,叙述采集系统的实现过程,该部分是进行肌电信号模式识别的前提。然后文章详细叙述模式识别系统的实现,根据肌电信号的特性和各种算法的特点选择小波包变换作为特征提取的方法,针对小波包变换缺乏时不变性的缺点,用结点能量代替原始系数构成特征向量,从而克服该缺点;确定特征空间降维方法,针对其实施过程中遇到的问题提出解决方案,并选择人工神经网络作为模式分类器。再具体描述上述算法的实现方法和过程,详述了识别系统的结构和基本工作原理。最后叙述了实验过程和结果,并对其结果进行简要分析与总结。在本实验中,对测试样本的识别正确率达到了94.17%.
本文通过本系统的研究与开发,提出了一套可以对肌电信号进行实时识别的算法和实现方案,并且针对实现过程中遇到的问题提出解决方案。在未来的工作中,将继续对各种特征处理的方法进行研究,找出更适合肌电信号的方法,同时也尝试不同分类器对分类性能的影响,提高系统性能,并将其应用于康复机器人中。
By the development of convalescence medicine, the research of rehabilitation robot is attracting more and more attention of people, and electromyographic (EMG) plays an important role in it by recognizing the sports consciousness of patients, which is an indispensable procedure of rehabilitation training. According to the specific characteristic of EMG, this paper proposes an algorithm used to recognize the pattern of EMG by studying a series of algorithms of feature extraction and pattern recognition, and tested this algorithm using actual EMG signal. This algorithm is also implemented by designing and developing an EMG pattern recognition system according to the requirement of rehabilitation robot, which is also used to certificate the effect of the proposed algorithm.
Firstly, this paper introduces the current condition of the research of EMG pattern recognition, studies and compares a series of approaches of feature extraction and pattern recognition. Then the structure of the EMG collecting system which is necessary for pattern recognition is introduced with some problems to consider. After that, this paper describes the procedure of the implementation of the EMG pattern recognition system in detail. According to the characteristic of EMG, wavelet package analysis is used to extract the feature, and the node energy is used to construct the feature vector instead of the original coefficients of wavelet package decomposition to resolve the time-invariance problem encountered. Then the algorithm of feature reduction is determined and a method to solve the problem encountered when calculating the reduction matrix. Then the implementation of the algorithm and the construct of the system were described. At last, the process and result of the experiment were described and a brief analysis was presented. In this experiment, the accuracy of recognition reaches up to 94.17%.
By the development of this system, we proposed a series of algorithms and implementing approaches of EMG pattern recognition. In the future, we will continue to study the various algorithms of pattern recognition and find which one is more suitable for the EMG. We will also try different kinds of classifiers and compare the performance of them and apply our system to the rehabilitation robot.
本文首先介绍当前国内外对肌电信号模式识别研究的现状,研究应用于肌电的各种特征提取和处理方法以及模式识别方法,比较各自的优缺点。然后介绍肌电采集系统的结构和各部分需要考虑的问题,叙述采集系统的实现过程,该部分是进行肌电信号模式识别的前提。然后文章详细叙述模式识别系统的实现,根据肌电信号的特性和各种算法的特点选择小波包变换作为特征提取的方法,针对小波包变换缺乏时不变性的缺点,用结点能量代替原始系数构成特征向量,从而克服该缺点;确定特征空间降维方法,针对其实施过程中遇到的问题提出解决方案,并选择人工神经网络作为模式分类器。再具体描述上述算法的实现方法和过程,详述了识别系统的结构和基本工作原理。最后叙述了实验过程和结果,并对其结果进行简要分析与总结。在本实验中,对测试样本的识别正确率达到了94.17%.
本文通过本系统的研究与开发,提出了一套可以对肌电信号进行实时识别的算法和实现方案,并且针对实现过程中遇到的问题提出解决方案。在未来的工作中,将继续对各种特征处理的方法进行研究,找出更适合肌电信号的方法,同时也尝试不同分类器对分类性能的影响,提高系统性能,并将其应用于康复机器人中。
By the development of convalescence medicine, the research of rehabilitation robot is attracting more and more attention of people, and electromyographic (EMG) plays an important role in it by recognizing the sports consciousness of patients, which is an indispensable procedure of rehabilitation training. According to the specific characteristic of EMG, this paper proposes an algorithm used to recognize the pattern of EMG by studying a series of algorithms of feature extraction and pattern recognition, and tested this algorithm using actual EMG signal. This algorithm is also implemented by designing and developing an EMG pattern recognition system according to the requirement of rehabilitation robot, which is also used to certificate the effect of the proposed algorithm.
Firstly, this paper introduces the current condition of the research of EMG pattern recognition, studies and compares a series of approaches of feature extraction and pattern recognition. Then the structure of the EMG collecting system which is necessary for pattern recognition is introduced with some problems to consider. After that, this paper describes the procedure of the implementation of the EMG pattern recognition system in detail. According to the characteristic of EMG, wavelet package analysis is used to extract the feature, and the node energy is used to construct the feature vector instead of the original coefficients of wavelet package decomposition to resolve the time-invariance problem encountered. Then the algorithm of feature reduction is determined and a method to solve the problem encountered when calculating the reduction matrix. Then the implementation of the algorithm and the construct of the system were described. At last, the process and result of the experiment were described and a brief analysis was presented. In this experiment, the accuracy of recognition reaches up to 94.17%.
By the development of this system, we proposed a series of algorithms and implementing approaches of EMG pattern recognition. In the future, we will continue to study the various algorithms of pattern recognition and find which one is more suitable for the EMG. We will also try different kinds of classifiers and compare the performance of them and apply our system to the rehabilitation robot.
引文
[1]加玉涛,罗志增.肌电信号特征提取方法综述.电子器件, 2007, 30(1): 326-330.
[2]罗志增,王人成.基于表面肌电信号的前臂手部多运动模式识别.仪器仪表学报, 2006, 27(9): 996-999.
[3] UNSER M, AIDROUBI A. A review of wavelets in biomedical applications[C]. Proc. Of the IEEE, 1996, 84(4): 626-638.
[4] JANG G CH. Using time frequency analysis technique in the classification of surface EMG signals[A]. Annual Int. Conf. of the IEEE Engineering in Medicine & Biology Society. 1994, 4: 1242-1243.
[5]钱政,杨莉,严璋.组合神经网络模型中典型训练样本集的选取.高电压技术, 1999, 25(4): 2-5
[6]王人成,黄昌华,李波等.基于BP神经网络的表面肌电信号模式分类的研究.中国医疗器械杂志, 1998, 22(2): 63-66.
[7]蔡立羽,王志中,张海虹.基于小波变换的肌电信号识别方法研究[J].数据采集与处理, 2005, 15(2): 255-258.
[8] Yong-Sheng Yuaq, F.K. Lam, F.H.Y. Chan, Yuan-Ting Zhang, Philip. A. Parker. A New Fuzzy Approach for Pattern Recognition with Application to EMG classification. Neural Networks, 1996, 2:1109-1114.
[9] Katsutoshi Kuribayashi, et al. A Discrimination System Using Neural Network for EMG-Controlled Prostheses-Integral type of EMG Signal Processing[C]// Proc. Of the 1993 IEEE/RSJ Int. Conference on Intelligent Robots and Systems, 1983, pp. 1750-1755.
[10] GRAUPE D. Function separation of EMG signals via ARMA identification methods for prosthesis control purposes [J]. IEEE Trans. Syst. Man Cybern. 1975, 5: 252-259.
[11] Stulen F B, De Luca C J. Frequency Parameters of the Myoelectric Signal as a Measure of Muscle Conduction Velocity [J]. IEEE Trans. Biomed Eng, 1981, BME-28: 515-523.
[12] Euljoon Park, Sanford Meek G. Fatigue Compensation of the Electromyographic Signal for Prosthetic Control and Force Estimation [J]. IEEE Trans Biomed Eng,1993, 40(10).
[13] Kevin Englehart, Bernard Hudgins, Philip A. Parker. A Wavelet-Based Continuous Classification Scheme for Multifunction Myoelectric Control. Biomedical Engineering, 2001, 48(3): 302-311.
[14] Jun-Uk Chu, Inhyuk Moon, Mu-Seong Mun. A Real-Time EMG Pattern Recognition based on Linear-Nonlinear Feature Projection for Multifunction Myoelectric Hand. 9th International Conference on Rehabilitation Robotics, 2005, 295-298.
[15] Gupta V, et al. Fractal Dimension of Surface EMG during Isokinetic Contractions [C]// IEEE-EMBC and CMBEC. 1995, 1331-1332.
[16] Wolf A, Swift J B, et al. Determining Lyapunov Exponents from a Time Series [J]. Physical D, 1985, 16:285-317.
[17]罗炯,金季春.表面肌电的处理方法及在体育科研中的应用前景.山东体育学院学报, 2005, 21(2): 56-59.
[18]许晶莉,李林.表面肌电图在脑性瘫痪诊治中的应用.中国伤残医学, 2008, 16(5): 136-137.
[19]俞晓杰,王颖.表面肌电图在腰痛康复评价中的应用.中国康复医学杂志, 2008, 23(10): 954-956.
[20]余洪俊.表面肌电图评价肌肉的功能状况.中国临床康复, 2002, 6(23): 3514-3515.
[21]王健,金德闻.康复医学领域的表面肌电应用研究.中国康复医学杂志, 2006, 21(1): 6-7.
[22]王人成,杨巍,黄昌华,李波.肌电控制及肌电信号的分析处理.康复医学工程, 1996, 11(6): 261-263.
[23] Peter Konrad. The ABC of EMG: A Practical Introduction to Kinesiological Electromyography. Noraxon INC, 2005.
[24] DelSys Incorporated. Surface Electromyography: Detection and Recording. Delsys Incorporated, 1996.
[25] K. Englehart, B. Hudgins, P.A. Parker, M.Stevenson. Classification of the Myoelectric Signal using Time-Frequency Based Representations. Special Issue of Medical Engineering and Physics on Intelligent Data Analysis in Electromyography and Electroneurography, Summer 1999, 21: 431-438.
[26]边肇祺,张学工等.模式识别(第二版).北京:清华大学出版社, 2000.
[27]飞思科技产品研发中心.小波分析理论与MATLAB7实现.北京:电子工业出版社, 2005.3.
[28] Gary G. Yen, Kuo-Chung Lin. Wavelet Packet Feature Extraction for Vibration Monitoring. IEEE Transactions on Industrial Electronics, 2000, 47(3):650-667.
[29] I. Daubechies, The wavelet transform, time-frequency localization and signal analysis. IEEE Transactions Inform. Theory, 1990, 36: 961-1005.
[30] Stephane G. Mallat. A Theory for Multiresolution Signal Decomposition: The Wavelet Representation. IEEE Transactions on Pattern Analysis and Machine Intelligence 1989, 11(7): 674-693.
[31] Donald B.Percival, Andrew T. Walden. Wavelet Methods for Time Series Analysis.北京:机械工业出版社, 2004.5.
[32] J. Kittler. Mathematical methods of feature selection in pattern recognition. Int. J. Man-Mach. Studies, 1975, 7(5):609-638.
[33] K. Fukunaga. Introduction to Statistical Pattern Recognition. New York: Academic, 1992.
[34] P. A. Devijver, J. Kittler. Pattern Recognition- A Statistical Approach. London, U.K: Prentice-Hall, 1982.
[35] Christopher M. Bishop. Pattern Recognition. Clarendon Press. Oxford, 1995.
[36] Bor-Chen Kuo, David A. Landgrebe. Nonparametric Weighted Feature Extraction for Classification. IEEE Transactions on Geoscience and Remote Sensing, 2004, 42(5): 1096-1105.
[37] M. Bressan, J. Vitria. Nonparametric discriminant analysis and nearest neighbor classification. Pattern Recognition Letters, 2003, 24: 2743-2749.
[38] M. Loog, R. P. W. Duin, R. Haeb-Umbach. Multiclass Linear Dimension Reduction by Weighted Pairwise Fisher Criteria. IEEE Trans. Pattern Analysis and Machine Intelligence, 2001, 23: 762-766.
[39] M. Aladjem. Multiclass discriminant mappings. Signal Processing, 1994, 35: 118-121.
[40] Bor-Chen Kuo, Li-Wei Ko, Chia-Hao Pai. Regularized Feature Extractions for Hyperspectral data Classification. International Geoscience and Remote Sensing Symposium. Toulouse, France, July 21-25, 2003, 3: 1767- 1769.
[41] J. P. Marques de sa著,吴逸飞译.模式识别——原理、方法及应用.北京:清华大学出版社, 2002.
[42]闻新,周露,王丹力. MATLAB神经网络应用设计.北京:科学出版社, 2000.
[43] Charles Kitchin, Lew Counts. A Designer’s Guide to Instrumentation Amplifiers (3rd Edition). Analog Devices, Inc. 2006.
[44]康华光,陈大钦,张林.电子技术基础——模拟部分(第五版).北京:高等教育出版社, 2006.
[45] Analog Devices. Datasheet AD620: Low Cost Low Power Instrumentation Amplifier. Analog Devices, Inc. 2004.
[46] Burr-Brown Products from Texas Instruments. Datasheet OPA2277: High Precision Operation Amplifiers. Texas Instruments Inc. 2005.
[47]周志敏,周纪海,纪爱华.便携式电子设备电源设计与应用.北京:人民邮电出版社, 2007.
[48] Advantech. USB-4716 User Manual (1st Edition). Advantech, 2006.
[49] Advantech. Device Driver User Manual for USB-4716. Advantech.
[50]王险峰,刘宝宏. Windows环境下的多线程编程原理与应用.北京:清华大学出版社, 2002.
[2]罗志增,王人成.基于表面肌电信号的前臂手部多运动模式识别.仪器仪表学报, 2006, 27(9): 996-999.
[3] UNSER M, AIDROUBI A. A review of wavelets in biomedical applications[C]. Proc. Of the IEEE, 1996, 84(4): 626-638.
[4] JANG G CH. Using time frequency analysis technique in the classification of surface EMG signals[A]. Annual Int. Conf. of the IEEE Engineering in Medicine & Biology Society. 1994, 4: 1242-1243.
[5]钱政,杨莉,严璋.组合神经网络模型中典型训练样本集的选取.高电压技术, 1999, 25(4): 2-5
[6]王人成,黄昌华,李波等.基于BP神经网络的表面肌电信号模式分类的研究.中国医疗器械杂志, 1998, 22(2): 63-66.
[7]蔡立羽,王志中,张海虹.基于小波变换的肌电信号识别方法研究[J].数据采集与处理, 2005, 15(2): 255-258.
[8] Yong-Sheng Yuaq, F.K. Lam, F.H.Y. Chan, Yuan-Ting Zhang, Philip. A. Parker. A New Fuzzy Approach for Pattern Recognition with Application to EMG classification. Neural Networks, 1996, 2:1109-1114.
[9] Katsutoshi Kuribayashi, et al. A Discrimination System Using Neural Network for EMG-Controlled Prostheses-Integral type of EMG Signal Processing[C]// Proc. Of the 1993 IEEE/RSJ Int. Conference on Intelligent Robots and Systems, 1983, pp. 1750-1755.
[10] GRAUPE D. Function separation of EMG signals via ARMA identification methods for prosthesis control purposes [J]. IEEE Trans. Syst. Man Cybern. 1975, 5: 252-259.
[11] Stulen F B, De Luca C J. Frequency Parameters of the Myoelectric Signal as a Measure of Muscle Conduction Velocity [J]. IEEE Trans. Biomed Eng, 1981, BME-28: 515-523.
[12] Euljoon Park, Sanford Meek G. Fatigue Compensation of the Electromyographic Signal for Prosthetic Control and Force Estimation [J]. IEEE Trans Biomed Eng,1993, 40(10).
[13] Kevin Englehart, Bernard Hudgins, Philip A. Parker. A Wavelet-Based Continuous Classification Scheme for Multifunction Myoelectric Control. Biomedical Engineering, 2001, 48(3): 302-311.
[14] Jun-Uk Chu, Inhyuk Moon, Mu-Seong Mun. A Real-Time EMG Pattern Recognition based on Linear-Nonlinear Feature Projection for Multifunction Myoelectric Hand. 9th International Conference on Rehabilitation Robotics, 2005, 295-298.
[15] Gupta V, et al. Fractal Dimension of Surface EMG during Isokinetic Contractions [C]// IEEE-EMBC and CMBEC. 1995, 1331-1332.
[16] Wolf A, Swift J B, et al. Determining Lyapunov Exponents from a Time Series [J]. Physical D, 1985, 16:285-317.
[17]罗炯,金季春.表面肌电的处理方法及在体育科研中的应用前景.山东体育学院学报, 2005, 21(2): 56-59.
[18]许晶莉,李林.表面肌电图在脑性瘫痪诊治中的应用.中国伤残医学, 2008, 16(5): 136-137.
[19]俞晓杰,王颖.表面肌电图在腰痛康复评价中的应用.中国康复医学杂志, 2008, 23(10): 954-956.
[20]余洪俊.表面肌电图评价肌肉的功能状况.中国临床康复, 2002, 6(23): 3514-3515.
[21]王健,金德闻.康复医学领域的表面肌电应用研究.中国康复医学杂志, 2006, 21(1): 6-7.
[22]王人成,杨巍,黄昌华,李波.肌电控制及肌电信号的分析处理.康复医学工程, 1996, 11(6): 261-263.
[23] Peter Konrad. The ABC of EMG: A Practical Introduction to Kinesiological Electromyography. Noraxon INC, 2005.
[24] DelSys Incorporated. Surface Electromyography: Detection and Recording. Delsys Incorporated, 1996.
[25] K. Englehart, B. Hudgins, P.A. Parker, M.Stevenson. Classification of the Myoelectric Signal using Time-Frequency Based Representations. Special Issue of Medical Engineering and Physics on Intelligent Data Analysis in Electromyography and Electroneurography, Summer 1999, 21: 431-438.
[26]边肇祺,张学工等.模式识别(第二版).北京:清华大学出版社, 2000.
[27]飞思科技产品研发中心.小波分析理论与MATLAB7实现.北京:电子工业出版社, 2005.3.
[28] Gary G. Yen, Kuo-Chung Lin. Wavelet Packet Feature Extraction for Vibration Monitoring. IEEE Transactions on Industrial Electronics, 2000, 47(3):650-667.
[29] I. Daubechies, The wavelet transform, time-frequency localization and signal analysis. IEEE Transactions Inform. Theory, 1990, 36: 961-1005.
[30] Stephane G. Mallat. A Theory for Multiresolution Signal Decomposition: The Wavelet Representation. IEEE Transactions on Pattern Analysis and Machine Intelligence 1989, 11(7): 674-693.
[31] Donald B.Percival, Andrew T. Walden. Wavelet Methods for Time Series Analysis.北京:机械工业出版社, 2004.5.
[32] J. Kittler. Mathematical methods of feature selection in pattern recognition. Int. J. Man-Mach. Studies, 1975, 7(5):609-638.
[33] K. Fukunaga. Introduction to Statistical Pattern Recognition. New York: Academic, 1992.
[34] P. A. Devijver, J. Kittler. Pattern Recognition- A Statistical Approach. London, U.K: Prentice-Hall, 1982.
[35] Christopher M. Bishop. Pattern Recognition. Clarendon Press. Oxford, 1995.
[36] Bor-Chen Kuo, David A. Landgrebe. Nonparametric Weighted Feature Extraction for Classification. IEEE Transactions on Geoscience and Remote Sensing, 2004, 42(5): 1096-1105.
[37] M. Bressan, J. Vitria. Nonparametric discriminant analysis and nearest neighbor classification. Pattern Recognition Letters, 2003, 24: 2743-2749.
[38] M. Loog, R. P. W. Duin, R. Haeb-Umbach. Multiclass Linear Dimension Reduction by Weighted Pairwise Fisher Criteria. IEEE Trans. Pattern Analysis and Machine Intelligence, 2001, 23: 762-766.
[39] M. Aladjem. Multiclass discriminant mappings. Signal Processing, 1994, 35: 118-121.
[40] Bor-Chen Kuo, Li-Wei Ko, Chia-Hao Pai. Regularized Feature Extractions for Hyperspectral data Classification. International Geoscience and Remote Sensing Symposium. Toulouse, France, July 21-25, 2003, 3: 1767- 1769.
[41] J. P. Marques de sa著,吴逸飞译.模式识别——原理、方法及应用.北京:清华大学出版社, 2002.
[42]闻新,周露,王丹力. MATLAB神经网络应用设计.北京:科学出版社, 2000.
[43] Charles Kitchin, Lew Counts. A Designer’s Guide to Instrumentation Amplifiers (3rd Edition). Analog Devices, Inc. 2006.
[44]康华光,陈大钦,张林.电子技术基础——模拟部分(第五版).北京:高等教育出版社, 2006.
[45] Analog Devices. Datasheet AD620: Low Cost Low Power Instrumentation Amplifier. Analog Devices, Inc. 2004.
[46] Burr-Brown Products from Texas Instruments. Datasheet OPA2277: High Precision Operation Amplifiers. Texas Instruments Inc. 2005.
[47]周志敏,周纪海,纪爱华.便携式电子设备电源设计与应用.北京:人民邮电出版社, 2007.
[48] Advantech. USB-4716 User Manual (1st Edition). Advantech, 2006.
[49] Advantech. Device Driver User Manual for USB-4716. Advantech.
[50]王险峰,刘宝宏. Windows环境下的多线程编程原理与应用.北京:清华大学出版社, 2002.