Simultaneous and proportional estimation of hand kinematics from EMG during mirrored movements at multiple degrees-of-freedom

This paper proposes and tests on able-bodied subjects a control strategy that can be practically applied in unilateral transradial amputees for simultaneous and proportional control of multiple degrees-of-freedom (DOFs). We used artificial neural networks to estimate kinematics of the complex wrist/hand from high-density surface electromyography (EMG) signals of the contralateral limb during mirrored bilateral movements in free space. The movements tested involved the concurrent activation of wrist flexion/extension, radial/ulnar deviation, forearm pronation/supination, and hand closing. The accuracy in estimation was in the range 79%-88% (r(2) index) for the four DOFs in six able-bodied subjects. Moreover, the estimation of the pronation/supination angle (wrist rotation) was influenced by the reduction in the number of EMG channels used for the estimation to a greater extent than the other DOFs. In conclusion, the proposed method and set-up provide a viable means for proportional and simultaneous control of multiple DOFs for hand prostheses.     

An accurate forearm EMG signal classification method using two‐channel electrode

An accurate electromyography (EMG) classification algorithm to control a virtual hand prosthesis with 12 degrees of freedom using two surface EMG electrodes is presented in this paper. We propose the application of independent component analysis (ICA) for blind‐source separation of the EMG signals obtained from two electrodes. One of the problems affecting the EMG classification accuracy is the location dependence of the EMG signal due to the superposition of signals from multiple sources. ICA is used to separate the two signals obtained from two surface electrodes into two independent EMG signals prior to the feature extraction and classification processes. We demonstrate that the EMG classification accuracy can be improved using the ICA algorithm. We also propose a novel eigen‐based feature that is extracted from the short‐time Fourier transform (STFT) magnitude spectrum. Our new feature not only decreases feature dimensions but also performs better than other well‐known features. We also implement the EMG classification scheme on the virtual robot arm. The performance shows promising result as indicated by a decrease in the Davies–Bolden (DB) index after applying the ICA © 2013 Institute of Electrical Engineers of Japan. Published by John Wiley & Sons, Inc.     

A Deep Neural Network Model for Upper Limb Swing Pattern to Control an Active Bionic Leg

Leg amputations are common in accidents and diseases. The present active bionic legs use Electromyography (EMG) signals in lower limbs (just before the location of the amputation) to generate active control signals. The active con-trol with EMGs greatly limits the potential of using these bionic legs because most accidents and diseases cause se-vere damages to tissues/muscles which originates EMG signals. As an alternative, the present research attempted to use an upper limb swing pattern to control an active bionic leg. A deep neural network (DNN) model is implemented to recognize the patterns in upper limb swing, and it is used to translate these signals into active control input of a bionic leg. The proposed approach can generate a full gait cycle within 1082 milliseconds, and it is comparable to the normal (a person without any disability) 1070 milliseconds gait cycle.     

Wind speed prediction using the hybrid model of wavelet decomposition and artificial bee colony algorithm-based relevance vector machine

In this paper, the hybrid model of wavelet decomposition and artificial bee colony algorithm-based relevance vector machine (WABCRVM) is presented for wind speed prediction. Here, wind speed can be regarded as a signal and decomposed into four decomposed signals with different frequency range, which can be obtained by 2-layer wavelet decomposition for wind speed data, and the prediction models of the four decomposed signals can be established by RVM with their each appropriate embedding dimension. Artificial bee colony algorithm (ABC) is used to select the appropriate kernel parameters of their RVM models. Thus, each decomposed signal’s RVM model of wind speed has appropriate embedding dimension and kernel parameter. Finally, the experimental results show that it is feasible for the proposed combination scheme to improve the prediction ability of RVM for wind speed.     

Compliant grasp in a myoelectric hand prosthesis

A novel myoelectric hand prosthesis consisting of electromyogram (EMG) signal processing units, a microprocessor-based dc motor servo system, and a 1 degree-of-freedom (DOF) end effector has been developed. The flexion angle and compliance of the finger of this prosthesis can be voluntarily controlled with EMG signals. The biomimetic controller for the myoelectric hand incorporated a model of the neuromuscular control system constructed from an analysis of the force response to length perturbation of the flexor pollicis longus muscle, processing of EMG signals, and the configuration of the hand. Basic functions of the human neuromuscular control system are realized by using position control, force feedback, and variable gain, modulated by EMG signal amplitude. A limb-absent person and four healthy subjects were able to voluntarily control the finger angle and compliance of the prosthesis and were able to easily grasp a soft object after a short training period.     

Changes in the surface EMG signal and the biomechanics of motion during a repetitive lifting task

The analysis of surface electromyographic (EMG) data recorded from the muscles of the back during isometric constant-force contractions has been a useful tool for assessing muscle deficits in patients with lower back pain (LBP). Until recently, extending the technique to dynamic tasks, such as lifting, has not been possible due to the nonstationarity of the EMG signals. Recent developments in time-frequency analysis procedures to compute the instantaneous median frequency (IMDF) were utilized in this study to overcome these limitations. Healthy control subjects with no history of LBP (n=9; mean age 26.3/spl plusmn/6.7) were instrumented for acquisition of surface EMG data from six electrodes on the thoraco-lumbar region and whole-body kinematic data from a stereo-photogrammetric system. Data were recorded during a standardized repetitive lifting task (load=15% body mass; 12 lifts/min; 5-min duration). The task resulted in significant decreases in IMDF for six of the nine subjects, with a symmetrical pattern of fatigue among contralateral muscles and greater decrements in the lower lumbar region. For those subjects with a significant decrease in IMDF, a lower limb and/or upper limb biomechanical adaptation to fatigue was observed during the task. Increases in the peak box acceleration were documented. In two subjects, the acceleration doubled its value from the beginning to the end of the exercise, which lead to a significant increase in the torque at L4/L5. This observation suggests an association between muscle fatigue at the lumbar region and the way the subject manipulates the box during the exercise. Fatigue-related biomechanical adaptations are discussed as a possible supplement to functional capacity assessments among patients with LBP.     

融合肌电信号与 A 型超声的新型肌肉疲劳检测方法

为了提高肌肉的疲劳检测效果,提出了一种双传感融合的方式来弥补单传感模式下信息容易丢失的不足。 该方式将表 面肌电信号的时频域特征与 A 型超声信号的肌肉厚度特征多维度融合,实现了双传感疲劳检测新模式。 采用支持向量机和神 经网络多模型训练,表面肌电信号与 A 型超声双传感融合在 3 种疲劳状态下的检测准确率可以达到 85%以上。 相较于仅仅使 用表面肌电信号的时频域特征(76. 99%)与 A 型超声的肌肉厚度(74. 87%)进行疲劳检测,准确率提升了 8% ~ 13%。 结果表明 对于疲劳检测,表面肌电信号与超声信号双传感融合模式比单传感模式更加准确有效。     

基于AD8232的表面肌电信号采集系统设计

表面肌电信号是人体表面肌肉通过收缩产生的生物电，能够反应人体肌肉的活动状况，在医学领域有重要的研究价值。由于表面肌电信号是微弱信号，较小的噪声就能产生较大的干扰；针对其这一弱点，本文设计了一款体积小、功耗低、抗干扰能力强的表面肌电信号采集系统。设计使用AD8232作为采集芯片，结合放大、滤波、模数转换电路，使用STM32进行数据处理，最后将数据通过低功耗蓝牙发送到上位机。测试结果表明：表面肌电信号能量主要集中在50~150 Hz，300 Hz以上明显衰减，受低频信号干扰较大。经过硬件滤波与小波降噪算法处理后，干扰得到较好抑制，尤其对低频干扰有显著衰减。系统有较强的抗干扰能力和可靠性，有着良好的市场前景。     

Fiber parametric amplifiers for wavelength band conversion

By using a loop configuration formed by a polarization beam splitter, we experimentally demonstrate that the existing wavelength-division multiplexing (WDM) sources in C-band can be wavelength converted to the S-band with low polarization sensitivity and low crosstalk. Using a fiber parametric amplifier as a band converter, we achieve experimentally <0.65-dB polarization sensitivity and >4.7-dB conversion efficiency over 30-nm conversion bandwidth in 315 m of fiber. Compared to the conventional straight fiber wavelength conversion scheme, a more than 2-dB improvement in polarization sensitivity is measured. In addition to the polarization insensitivity, channel crosstalk is measured to be <-27 dB in 315 m of high nonlinearity fiber. In a detailed experimental study, the pattern of crosstalk in longer fiber lengths and the coupling between the polarization sensitivity and crosstalk are measured. For example, with a 430-m fiber length, we measure the degradation in polarization sensitivity to be ~4 dB for 12-dB increased signal power. The experimental results are also confirmed by theoretical calculations. Moreover, in a 32 channels systems simulation, the signal-to-noise ratio (SNR) of the converted signals after 800-km propagation is calculated to be only 0.8-dB degraded compared to using laser diodes with the same initial SNR values. Furthermore, we calculate the effect of pump noise and show that the relative intensity noise of the pump is transferred to the converted signals with an additional 8-dB/Hz degradation     

Crosstalk in multilayer printed circuit boards

Experimental measurements were conducted to study the problem of crosstalk in multilayer printed-circuit-board design. Experiments using both pulse and sinusoidal signals were designed to study crosstalk as a function of coupling line lengths, line spacings, and shielding versus nonshielding (lines in different layers). The mutual inductance and capacitance between two similar transmission lines were also extracted from simple models of crosstalk. It is found that when a sinusoid-driven signal was used as the voltage source, the coupled transmission lines acted as a transformer. The induced voltage on the passive line is approximately proportional to the square root of the coupling line length in the range of 2-20 in. These experiments can be used to enhance students' quantitative understanding of the problem of crosstalk in modern electronic circuit design     

A novel neural network controller and its efficient DSP implementation for vector-controlled induction motor drives

An artificial neural network controller is experimentally implemented on the Texas Instruments TMS320C30 digital signal processor (DSP). The controller emulates indirect field-oriented control for an induction motor, generating direct and quadrature current command signals in the stationary frame. In this way, the neural network performs the critical functions of slip estimation and matrix rotation internally. There are five input signals to the neural network controller, namely, a shaft speed signal, the synchronous frame present and delayed values of the quadrature axis stator current, as well as two neural network output signals fed back after a delay of one sample period. The proposed three-layer neural network controller contains only 17 neurons in an attempt to minimize computational requirements of the digital signal processor. This allows DSP resources to be used for other control purposes and system functions. For experimental investigation, a sampling period of 1 ms is employed. Operating at 33.3 MHz (16.7 MIPS), the digital signal processor is able to perform all neural network calculations in a total time of only 280 /spl mu/s or only 4700 machine instructions. Torque pulsations are initially observed, but are reduced by iterative re-training of the neural network using experimental data. The resulting motor speed step response (for several forward and reverse step commands) quickly tracks the expected response, with negligible error under steady-state conditions.     

基于频域分离解调的复合信源无线电能与信号并行传输技术

提出了一种并联LC隔离型电场耦合式无线电能与信号并行传输电路拓扑。该拓扑依靠自身阻抗特性实现了绝大部分电能串扰的隔离。同时,该拓扑具有较宽的信道通频带宽,克服了目前已有电能与信号并行传输系统中信道通频带宽窄的问题。充分利用该带宽,以多路复合信源作为信道输入,采用快速傅里叶变换算法对信道输出信号进行频谱分析,在频域中实现了低频串扰与多路复合高频信号的分离和电能串扰频谱干扰下的多路复合信号解调,从而提升了信号传输速率与抗干扰能力。通过仿真与实验验证了该方法的正确性与有效性。     

基于TiCNN-DRSN模型的sEMG手势识别算法的研究

基于表面肌电信号和模式识别的手势识别方法在康复手领域中具有广阔的应用前景。提出一种基于表面肌电信号的手部姿势识别方法,以预测手部的52种动作。为解决表面肌电信号易受干扰的问题,提高对表面肌电信号的分类效果,提出了TiCNN DRSN网络,主要作用是在拥有噪声的情况下能够更好的识别率,减少滤除噪声的时间。TiCNN网络使用卷积核Dropout和极小批量训练,为卷积神经网络引入训练干扰并且增加了模型的泛化性;DRSN网络可以有效的剔除sEMG信号中的冗余信号,减少信号噪声干扰。TiCNN DRSN网络在不需要任何降噪预处理的前提下,取得了很高的抗噪与自适应性能。本模型在Ninapro数据库上的识别率达到97.43%±0.8%。     

超声波法进行变压器局部放电模式识别的研究

在线识别局部放电模式可以有效地判断局部放电对变压器绝缘的危害程度。文中根据局部放电超声波信号存在的非线性和非平稳特性，提出利用分形理论对局部放电超声波信号的时域脉冲形进行分析，对分形理论及其参数计算方法进行了简单的介绍，在自制的几种典型变压器局部放电模型上进行实验，计算了所得到的局部放电超声波信号的分型参数（分维数和空缺率），得到不同放电形式的分形特征，利用人工神经网络对所得到的分形特征进行模式识别，结果表明利用超声波信号可以有效地判断变压器局部放电模式，为变压器局部放电信号的特征提取和模式识别提供了一种新的研究方法。     

Flexible noncontact electrodes for comfortable monitoring of physiological signals

Physiological signals such as electrocardiography (ECG), electromyography (EMG), and electroencephalography (EEG) could objectively reflect the functioning status of the human body and the monitoring of these signals is useful for various applications including brain computer interface, neurological rehabilitation, and long‐term healthcare monitoring. Currently, wet electrodes are commonly used for the monitoring of physiological signals and it usually requires conductive gels to achieve high quality recordings, which may cause discomfort to the patient and increase risk of skin allergy. In this study, a noncontact electrode made of a multilayer flexible printed circuit without any rigid electronic components on either side was proposed. The flexible noncontact electrode was capable of measuring physiological signals without any direct skin contact or conductive gels and could be bent freely according to the local shape to achieve optimal capacitive coupling with the skin surface. The results showed that the proposed flexible noncontact electrode could obtain different physiological signals with good quality compared with traditional wet electrodes. The ECG signals could be reliably measured with different insulation materials between the skin and the electrode, with up to five layers of insulation materials. It was also found that flexible electrode could achieve higher signal‐to‐noise ratio and therefore had better performance than traditional hard printed circuit board electrode, when measuring EMG signal through the cloth and EEG signals over the hair. The proposed method of this study might provide a novel and comfortable way to measure physiological signals for neurological rehabilitation, wearable devices, and other healthcare applications.     

Direct neural sensory feedback and control of a prosthetic arm.

Evidence indicates that user acceptance of modern artificial limbs by amputees would be significantly enhanced by a system that provides appropriate, graded, distally referred sensations of touch and joint movement, and that the functionality of limb prostheses would be improved by a more natural control mechanism. We have recently demonstrated that it is possible to implant electrodes within individual fascicles of peripheral nerve stumps in amputees, that stimulation through these electrodes can produce graded, discrete sensations of touch or movement referred to the amputee's phantom hand, and that recordings of motor neuron activity associated with attempted movements of the phantom limb through these electrodes can be used as graded control signals. We report here that this approach allows amputees to both judge and set grip force and joint position in an artificial arm, in the absence of visual input, thus providing a substrate for better integration of the artificial limb into the amputee's body image. We believe this to be the first demonstration of direct neural feedback from and direct neural control of an artificial arm in amputees.     

Restoration of use of paralyzed limb muscles using sensory nerve signals for state control of FES-assisted walking.

A real-time functional electrical stimulation (FES) state controller was designed that utilized sensory nerve cuff signals from the cat forelimb to control the timing of stimulation of the Palmaris Longus (PalL) muscle during walking on the treadmill. Sensory nerve signals from the median and superficial radial nerves provided accurate, reliable feedback related to foot contact and lift-off which, when analyzed with single threshold Schmitt triggers, produced valuable state information about the step cycle. The study involved three experiments: prediction of the timing of muscle activity in an open-loop configuration with no stimulation, prediction of the timing of muscle activity in a closed-loop configuration that included stimulation of the muscle over natural PaIL electromyogram (EMG), and temporary paralysis of selected forelimb muscles coupled with the use of the state controller to stimulate the PalL in order to return partial support function to the anesthetized limb. The FES state controller was tested in a variety of walking conditions, including different treadmill speeds and slopes. The results obtained in these experiments demonstrate that nerve cuff signals can provide a useful source of feedback to FES systems for control of limb function.     

术中神经监测系统设计与实现

在传统甲状腺手术中,临床医生凭借肉眼识别,很难有效判定喉返神经位置,容易导致喉返神经损伤,而喉返神经损伤会导致患者无法正常说话甚至危及生命。使用术中神经监测技术,有效识别神经信号,避免术中损伤神经,本文详细介绍术中神经监测系统的设计与实现方法。该系统主要由主控模块、电流刺激模块以及神经信号采集模块组成,主控模块采用了ARM+FPGA双核架构,实现了电流刺激模块输出脉冲电流以及信号采集模块中的A/D转化功能,同时对采集模块采集的数据进行处理;电流刺激模块以高精度D/A转化芯片为核心设计电流源电路,用来调控电流脉冲的强度和脉宽;采集模块以高精度A/D模数转化芯片为核心适配低噪放大电路,实现神经信号的采集功能。使用本系统进行小猪喉返神经模拟临床人体监测实验以及在医院进行临床实验验证,证明本系统能够有效激发喉返神经信号并识别到神经信号,满足临床医生需求。     

无线多通道表面肌电信号采集系统设计

本文设计了一种无线多通道表面肌电信号（surface electromyography,SEMG）采集系统,该系统包括多通道的无线传感器和信号接收部分。传感器可独立的穿戴于人体表面,以线形差分电极获取表面肌电信号,对其进行放大、滤波、A/D变换,并用无线的方式按本文设计的通信协议发送给接收部分。接收部分对各传感器的数据进行整合,并通过USB接口传输给电脑进行存储、显示和处理。每个传感器体积为35mm×20mm×11mm,重量仅13g（含电池）,一次充电可工作9个小时,无线通信距离达7.5m,采集到的信号噪声低于？70dB（肌电信号1mV代表0dB）。该设计大大提高了电极安放的便利性,采集设备的便携性与人体的安全性,且避免了工频干扰,能够满足基于表面肌电信号的手势或姿势识别等研究的要求。     

基于经验模态分解的牵张反射起始点检测研究

针对痉挛状态患者表面肌电信号易出现虚假的肌电峰值,引起牵张反射起始点前后的信号差异变小,提出经验模态分解去噪与改进样本熵识别的牵张反射起始点检测方法。首先用经验模态分解对肌电信号进行分解;然后以受试者静息状态下的表面肌电信号为参考,设定软阈值对分解的信号进行去噪;最后用改进样本熵识别牵张反射起始点。实验结果表明,经验模态分解算法可以有效地去除肌电信号噪声,而且在改进样本熵的最优参数下牵张反射起始点平均识别率为94%。