基于表面肌电信号的人体动作识别算法研究进展

表面肌电信号(Surface Electromyography,sEMG)是通过相应肌群表面的传感器记录下来的一维时间序列非平稳生物电信号,不但反映了神经肌肉系统活动,对于反映相应动作肢体活动信息同样重要。而模式识别是肌电应用领域的基础和关键。为了在应用基于表面肌电信号模式识别中选取合适算法,本文拟对基于表面肌电信号的人体动作识别算法进行回顾分析,主要包括模糊模式识别算法、线性判别分析算法、人工神经网络算法和支持向量机算法。模糊模式识别能自适应提取模糊规则,对初始化规则不敏感,适合处理s EMG这样具有严格不重复的生物电信号;线性判别分析对数据进行降维,计算简单,但不适合大数据;人工神经网络可以同时描述训练样本输入输出的线性关系和非线性映射关系,可以解决复杂的分类问题,学习能力强;支持向量机处理小样本、非线性的高维数据优势明显,计算速度快。比较各方法的优缺点,为今后处理此类问题模式识别算法选取提供了参考和依据。     

基于定量分析方法的动作表面肌电信号分析

介绍了非线性数据处理方法递归图法(recurrence plots,RP)及其定量分析方法(recurrence quantifi-cation analysis,RQA),并利用RP和RQA研究了动作表面肌电信号。研究发现,表面肌电信号在不同动作模式下其所对应的RP图在结构上差异明显,通过计算两通道肌电信号的RQA指标递归率,发现不同动作信号的RQA指标递归率值具有不同的聚类分布。该方法为肌电信号的动作模式分类提供了一种新的思路。     

基于支持向量机的表面肌电信号和加速度融合跌倒识别方法

目的:针对老人易跌倒和跌倒过后可能产生严重后果这一现实问题,通过将表面肌电信号和加速度融合,进一步优化采用支持向量机分类器下的包含跌倒在内的几种不同动作的分类效果。方法:提出基于表面肌电和加速度信号融合的跌倒识别算法,首先采集股直肌,股内侧肌,胫骨前肌和腓肠肌的表面肌电信号以及位于腰部的三轴加速度信号作为实验数据,然后利用滑动窗口法提取表面肌电和加速度信号的均方根值,最后针对人体日常活动和跌倒的运动特征,构建了支持向量机的分类器。结果:实验数据共计320组数据,包括3种日常活动和向前跌倒,其中160组数据作为训练集,另外160组数据作为测试集。对4种动作进行识别实验,算法的准确度为93.23%、灵敏度为92.4%、特异度为100%,达到了良好的分类效果。结论:基于支持向量机的表面肌电信号和加速度融合的跌倒识别算法分类效果良好,对于老人跌倒防护具有现实意义。     

An Anthropomorphic Robot Hand Developed Based on Underactuated Mechanism and Controlled by EMG Signals

When developing a humanoid myo-control hand,not only the mechanical structure should be considered to afford a highdexterity,but also the myoelectric (electromyography,EMG) control capability should be taken into account to fully accomplishthe actuation tasks.This paper presents a novel humanoid robotic myocontrol hand (AR hand Ⅲ) which adopted an underac-tuated mechanism and a forearm myocontrol EMG method.The AR hand Ⅲ has five fingers and 15 joints,and actuated by threeembedded motors.Underactuation can be found within each finger and between the rest three fingers (the middle finger,the ringfinger and the little finger) when the hand is grasping objects.For the EMG control,two specific methods are proposed:thethree-fingered hand gesture configuration of the AR hand Ⅲ and a pattern classification method of EMG signals based on astatistical learning algorithm-Support Vector Machine (SVM).Eighteen active hand gestures of a testee are recognized ef-fectively,which can be directly mapped into the motions of AR hand Ⅲ.An on-line EMG control scheme is established basedon two different decision functions:one is for the discrimination between the idle and active modes,the other is for the recog-nition of the active modes.As a result,the AR hand Ⅲ can swiftly follow the gesture instructions of the testee with a time delayless than 100 ms.     

The role of EMG awareness in EMG biofeedback learning

Underlying most research on biofeedback learning is a theoretical model of the processes involved. The current study tested a prediction from the Awareness Model: High initial EMG awareness should facilitate response control during EMG biofeedback training. Seventy-two undergraduates were assessed for forehead EMG awareness by asking them to produce target responses from 1.0 to 5.0 µV every 15 s for 16 trials. Based on this assessment, two groups (high and low awareness) were trained for 64 trials to produce these target levels with either EMG biofeedback, practice (no feedback), or noncontingent EMG feedback. A transfer task was identical to the initial assessment. During training, the biofeedback group deviated less from target than the practice and noncontingent groups. The biofeedback group was the only group to improve from initial EMG awareness activity. During transfer, only the low awareness biofeedback group remained below initial EMG awareness level. These findings can be interpreted in terms of the Two-Process Model.     

Surface EMG: The issue of electrode location

This paper contributes to clarifying the conditions under which electrode position for surface EMG detection is critical and leads to estimates of EMG variables that are different from those obtained in other nearby locations. Whereas a number of previous works outline the need to avoid the innervation zone (or the muscle belly), many authors place electrodes in the central part or bulge of the muscle of interest where the innervation zone is likely to be. Computer simulations are presented to explain the effect of the innervation zone on amplitude, frequency and conduction velocity estimates from the signal and the need to avoid placing electrodes near it. Experimental signals recorded from some superficial muscles of the limbs and trunk (abductor pollicis brevis, flexor pollicis brevis, biceps, upper trapezius, vastus medialis, vastus lateralis) were processed providing support for the findings obtained from simulations. The use of multichannel techniques is recommended to estimate the location of the innervation zone and to properly choose the optimal position of the detection point(s) allowing meaningful estimates of EMG variables during movement analysis.     

A Real-Time Capable Linear Time Classifier Scheme for Anticipated Hand Movements Recognition from Amputee Subjects Using Surface EMG Signals

Anticipated hand movements of amputee subjects are considered difficult to classify using only Electromyogram (EMG) signals and machine learning techniques. For a long time, classifying such s-EMG signals have been considered as a non-linear problem, and the problem of signal sparsity has not been given detailed attention in a large set of action classes. For addressing these problems, this paper is proposing a linear-time classifier termed as Random Fourier Mapped Collaborative Representation with distance weighted Tikhonov regularization matrix (RFMCRT). RFMCRT attempts to tackle the non-linear problem via Random Fourier Features and sparsity issue with collaborative representation. The projection error of Random Fourier Features is reduced by projecting to the same dimension as the original feature space and later finding the collaborative representation, with an optional non-negative constraint (RFMNNCRT). The proposed two classifiers were tested with time-domain features computed from the EMG signals obtained from NINAPRO databases using a non-overlapping sliding window size of 256 ms. Due to the random nature of our proposed classifiers, this paper has computed the average and worst-case performance for 50 trials and compared them with other reported classifiers. The results show that RFMNNCRT (average case) outperformed state-of-the-art classifiers with the accuracy of 93.44% for intact subjects and 55.67% for amputee subjects. In the worst-case situation, RFMCRT achieves considerable performance for the same, with the reported accuracy of 91.55% and 50.27% respectively. Our proposed classifier guarantees acceptable levels of accuracy for large classes of hand movements and also maintains good computational efficiency in comparison to LDA and SVM.     

A preliminary analysis of EMG variance as an index of change in EMG biofeedback treatment of tension-type headache

The effectiveness of EMG biofeedback training for tension headache has been well established. Previous studies evaluating changes in an average EMG activity score from pre- to posttreatment have not consistently found a relationship between a reduction in average EMG activity and headache improvement at posttreatment. The current study is a preliminary analysis of the utility of EMG variance as another possible mechanism of change. Frontalis EMG average activity and variances from 6 chronic tension-type headache sufferers who demonstrated significant improvement in headache activity at posttreatment (at least 70%) and 6 chronic tension-type headache sufferers who did not demonstrate improvement (less than 30%) were examined across 6 sessions of biofeedback treatment. The improved group demonstrated larger time-specific EMG variance in relation to mean EMG amplitudes during all treatment sessions. A dramatic decline in time-specific variance was observed during the later treatment sessions for improved participants; this pattern was not observed in the group who demonstrated little or no improvement. Results from the current study suggest that the inclusion of both average EMG activity and EMG variance may provide a more comprehensive measure to evaluate possible physiological changes responsible for improvement in headache activity following EMG biofeedback training.     

Muscle fatigue and calibration of EMG measurements

Amplitude electromyography (EMG) is often used as an estimator of muscular load. Such measurements can, however, be biased by other factors, for example muscular fatigue. The aim of this study was to examine the influence of fatigue on amplitude parameters of the EMG. The test subjects raised the arm to 90⁹ of abduction in the plane of the scapula. The hand was loaded with 0, 1 and 2 kg during 5, 3 and 2 min respectively. EMG was recorded from the trapezius muscle, and spectral and amplitude parameters were calculated. There was a significant rise of the EMG amplitude as a sign of fatigue at all load levels: 7% min⁻¹ at 0 kg, 15% min⁻¹ at 1 kg, and 19% min⁻¹ at 2 kg. At 0 kg hand load there was no change of the spectral parameters but at higher load levels there was a significant decline of mean power frequency: 3% min⁻¹ at 1 kg and 11% at 2 kg. The amplitude rise due to muscle fatigue may seriously jeopardize calibration measurements unless the duration of the load is kept limited.     

Deficient Discrimination of EMG Levels and Overestimation of Perceived Tension in Chronic Pain Patients

Twenty chronic low back pain patients (CBP), twenty tension headache (THA) patients, and twenty healthy controls (HC) participated in a tension production task where subjects had to attain four levels (4, 8, 12, 16 V) of muscle tension at the m. frontalis and the m. erector spinae. Ratings of perceived tension, pain, and aversiveness as well as EMG, heart rate, and skin conductance levels were recorded. Signal detection and correlational methods revealed that the patients were deficient in muscle tension discrimination at high tension levels in both muscles. They generally overestimated low and underestimated high levels of muscle tension, especially in the CBP group. At low muscle tension levels, both healthy controls and patients showed deficient discrimination ability. Perceived muscle tension, aversiveness, and pain ratings during the tasks were higher in the patient groups. These data confirm and clarify previous reports of deficient tension perception and show concurrent overestimation of bodily symptoms in chronic musculoskeletal pain patients.     

The minimum number of contractions required to examine the EMG amplitude versus isometric force relationship for the vastus lateralis and vastus medialis

Studies have demonstrated that the electromyographic (EMG) amplitude versus submaximal isometric force relationship is relatively linear. The purpose of this investigation was to determine the minimum number of contractions required to study this relationship. Eighteen men (mean age = 23 years) performed isometric contractions of the leg extensors at 10–90% of the maximum voluntary contraction (MVC) in 10% increments while surface EMG signals were detected from the vastus lateralis and vastus medialis. Linear regression was used to determine the coefficient of determination, slope coefficient, and y-intercept for each muscle and force combination with successively higher levels included in the model (i.e., 10–30%, … 10–90% MVC). For the slope coefficients, there was a main effect for force combination (P < .001). The pairwise comparisons showed there was no difference from 10–60% through 10–90% MVC. For the y-intercepts, there were main effects for both muscle (vastus lateralis [4.3 μV RMS] > vastus medialis [−3.7 μV RMS]; P = .034) and force combination (P < .001), with similar values shown from 10–50% through 10–90% MVC. The linearity of the absolute EMG amplitude versus isometric force relationship for the vastus lateralis and vastus medialis suggests that investigators may exclude high force contractions from their testing protocol.     

The effect of patella taping on vastus medialis oblique and vastus laterialis EMG activity and knee kinematic variables during stair descent

The purpose of this study was to evaluate the effect of patella taping in normal subjects. Previous work has established positive effects of patella taping on patellofemoral pain syndrome patients, but the mode of action remains unclear. It has been hypothesized that taping brings about subtle changes in the internal physiological environment of the joint. It could be expected that in normal joints taping would bring about a measurable change in function, as the joint is no longer operating in an optimal physiological environment. 10 normal female subject’s (21.4 ± 1.2 years) vastus medialis oblique (VMO) and vastus laterialis (VL) EMG activity and knee kinematics (peak stance flexion angle and angular velocity) were assessed during a step descent, with and without a taped patella. The effect of taping was to significantly decrease VMO and VL EMG activity. Taping also significantly reduced peak stance phase knee flexion and peak stance phase knee flexion angular velocity. In normal asymptomatic subjects patella taping created a situation in which their performance was changed to one similar to that of the pathological patellofemoral pain syndrome population. It would appear that taping caused the joint to function sub-optimally supporting the hypothesis that taping could change the functioning of the patellofemoral joint.     

Separation of propagating and non propagating components in surface EMG

Surface electromyogram (EMG) detected by electrode arrays along the muscle fibre direction can be approximated by the sum of propagating and non propagating components. A technique to separate propagating and non propagating components in surface EMG signals is developed. The first step is an adaptive filter, which allows obtaining an estimation of the delay between signals detected at different channels and a first estimate of propagating and non propagating components; the second step is used to optimise the estimation of the two components. The method is applicable to signals with one propagating and one non propagating component. It was optimised on simulated signals, and then applied to single motor unit action potentials (MUAP) and to electrically elicited EMG (M-waves).

The new method was first tested on phenomenological signals constituted by the sum of a propagating and a non propagating signal and then applied to simulated and experimental EMG signals. Simulated signals were generated by a cylindrical, layered volume conductor model. Experimental signals were monopolar surface EMG signals collected from the abductor pollicis brevis muscle and M-waves recorded during transcutaneous electrical stimulation of the biceps muscle. The technique may find different applications: in single motor unit (MU) studies (a) for decreasing the variability and bias of CV estimates due to the presence of the non propagating components, (b) for estimating automatically the length of the muscle fibres from only three detected channels and (c) for removal of the stimulation artifact M-waves.