基于表面肌电信号的人体下肢动作识别及关节力矩分析研究

发布时间：2018-03-14 18:37

本文选题：表面肌电信号　切入点：模式识别　出处：《苏州大学》2016年硕士论文　论文类型：学位论文

【摘要】：表面肌电信号是指从肌肉表面通过电极引导而记录下来的神经肌肉系统活动时的生物电信号,它与肌肉的活动状态和功能状态之间存在着不同程度的关联性,不同的肢体动作由不同的肌肉收缩模式产生,肌电信号特征也存在差异,通过对表面肌电信号特征进行分析就可以区分出不同的动作模式。表面肌电信号已经广泛应用到临床医学、运动医学、生物医学与工程等诸多领域,在智能假肢方面已经成为功能性电刺激的理想控制信号。本文通过对采集的表面肌电信号(sEMG)信号进行分析处理,识别出运动者的运动意图,研究肌电信号与关节力矩之间的关系,为智能假肢控制做基础。主要工作包括如下四个方面：(1)利用表面电极获取下肢运动肌电信号,设计采集10位测试人员人体腿部的股直肌、股内侧肌、股外侧肌、股二头肌、半腱肌、胫骨前肌、内侧腓肠肌、外侧腓肠肌、比目鱼肌九块肌肉在上下台阶和上下斜坡时的表面肌电信号。通过采用小波阈值去噪方法,对小波分解后的各层系数中模大于和小于某阈值的系数分别处理,然后对处理完的小波系数再进行反变换,重构出经过去噪后的信号。(2)基于DB、Dmey和Bior三种不同的母小波,采用离散小波变换通过不同的分析方法对表面肌电信号进行多尺度分解,通过对特征值的分析比较,对单一特征值进行融合分析构成特征向量,使各单一特征的优点互补,增强了特征值的表征能力,从而有效提高了模式分类的正确率。(3)运用BP神经网络和Elman神经网络对特征矢量进行人体运动模式的识别,通过将单一特征值和融合特征值分别输入BP神经网络和Elman神经网络,得出多特征融合与单特征集的表征能力比较。(4)本文以表面肌电信号为生物学输入信号计算膝关节肌肉力,从正反动力学两方面进行研究,逆动力学使用运动学数据、测力台数据等通过力及力矩平衡方程得到关节合力矩。正动力学首先使用表面肌电信号计算得到肌肉激活度,使用运动学数据和个性化的人体骨骼肌肉模型计算得到肌肉长度、收缩速度和力臂。然后将肌肉激活度、肌肉长度、收缩速度代入肌肉收缩模型得到肌肉力代入力臂后得到肌肉力矩。最后将逆动力学得到的关节合力矩与正动力学得到的关节肌肉合力矩进行比较分析。
[Abstract]:Surface electromyography (EMG) is a bioelectric signal recorded from the surface of a muscle guided by an electrode. It is related to the active and functional states of the muscle to varying degrees. Different limb movements are produced by different muscle contraction patterns, and the characteristics of EMG signals are also different. Different action patterns can be distinguished by analyzing the characteristics of surface electromyography (EMG), which has been widely used in clinical medicine, sports medicine, biomedicine and engineering, etc. The intelligent prosthesis has become an ideal control signal for functional electrical stimulation. In this paper, we analyze and process the collected surface EMG signal to identify the motion intention of the motor. To study the relationship between EMG signal and joint torque, which is the basis of intelligent prosthesis control. The main work includes the following four aspects: 1) using surface electrode to obtain motor EMG signal of lower extremity, designing and collecting the rectus femoris muscle of human leg of 10 testers. Surface electromyography of medial femoral muscle, lateral femoral muscle, biceps femoris muscle, semitendinosus muscle, anterior tibial muscle, medial gastrocnemius muscle, lateral gastrocnemius muscle and soleus muscle. The coefficients whose modulus is greater than or less than a certain threshold in the coefficients of each layer after wavelet decomposition are processed separately, and then the wavelet coefficients after processing are inversely transformed to reconstruct the de-noised signal. (2) based on three different mother wavelets, the Dmey wavelet and the Bior wavelet are used to reconstruct the de-noised signal. The discrete wavelet transform is used to decompose the surface EMG signal by different analysis methods. Through the analysis and comparison of the eigenvalues, the single eigenvalue is fused to form the eigenvector, which makes the advantages of each single feature complement each other. The feature value representation ability is enhanced, and the correct rate of pattern classification is improved effectively. BP neural network and Elman neural network are used to recognize the human motion pattern. By inputting the single eigenvalue and the fused eigenvalue into BP neural network and Elman neural network respectively, the comparison of the characterizing ability of multi-feature fusion and single feature set is obtained. (4) in this paper, the surface EMG signal is used as the biological input signal to calculate the muscle force of knee joint. From two aspects of positive and negative dynamics, inverse dynamics uses kinematics data and force table data to obtain joint joint force moment by force and torque balance equation. Positive dynamics first uses surface electromyography to calculate muscle activation. Using kinematics data and personalized skeletal muscle models to calculate muscle length, contraction speed, and force arm. Then muscle activation, muscle length, The muscle torque is obtained after the muscle force is replaced by the force arm. Finally, the joint resultant moment obtained by inverse dynamics is compared with the joint muscle force moment obtained by positive dynamics.
【学位授予单位】：苏州大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TN911.7;R318

【相似文献】