The theoretical development of a multichannel time-seriesmyoprocessor for simultaneous limb function detection and muscle forceestimation

This paper details the theoretical development and simulation of a complete time-series myoprocessor which provides reliable and economical predictions of both the magnitude and direction of limb motion from the spectral content of the surface EMG. Treating multiple channels of surface EMG as a vector-valued autoregressive process incorporates spatially distributed information which extends the operating range of parallel filtering limb function classifiers and reduces their sensitivity to modeling conditions. Active joint moment is estimated simultaneously from the pooled variance of the prewhitened EMG generated during the classification procedure. Estimation from the prewhitened sequence imposes no additional computational requirements and extends optimal myoprocessors to include multiple channels of serially dependent data. Such a system may be applied to the control of actively powered prostheses or orthoses.     

Assessment of average muscle fiber conduction velocity from surface EMG signals during fatiguing dynamic contractions

In this paper, we propose techniques of surface electromyographic (EMG) signal detection and processing for the assessment of muscle fiber conduction velocity (CV) during dynamic contractions involving fast movements. The main objectives of the study are: 1) to present multielectrode EMG detection systems specifically designed for dynamic conditions (in particular, for CV estimation); 2) to propose a novel multichannel CV estimation method for application to short EMG signal bursts; and 3) to validate on experimental signals different choices of the processing parameters. Linear adhesive arrays of electrodes are presented for multichannel surface EMG detection during movement. A new multichannel CV estimation algorithm is proposed. The algorithm provides maximum likelihood estimation of CV from a set of surface EMG signals with a window limiting the time interval in which the mean square error (mse) between aligned signals is minimized. The minimization of the windowed mse function is performed in the frequency domain, without limitation in time resolution and with an iterative computationally efficient procedure. The method proposed is applied to signals detected from the vastus laterialis and vastus medialis muscles during cycling at 60 cycles/min. Ten subjects were investigated during a 4-min cycling task. The method provided reliable assessment of muscle fatigue for these subjects during dynamic contractions.     

Electromyogram amplitude estimation with adaptive smoothing windowlength

Typical electromyogram (EMG) amplitude estimators use a fixed window length for smoothing the amplitude estimate. When the EMG amplitude is dynamic, previous research suggests that varying the smoothing length as a function of time may improve amplitude estimation. This paper develops optimal time-varying selection of the smoothing window length using a stochastic model of the EMG signal. Optimal selection is a function of the EMG amplitude and its derivatives. Simulation studies, in which EMG amplitude was changed randomly, found that the "best" adaptive filter performed as well as the "best" fixed-length filter. Experimental studies found the advantages of the adaptive processor to be situation dependent. Subjects used real-time EMG amplitude estimates to track a randomly-moving target. Perhaps due to task difficulty, no differences in adaptive versus fixed-length processors were observed when the target speed was fast. When the target speed was slow, the experimental results were consistent with the simulation predictions. When the target moved between two constant levels, the adaptive processor responded rapidly to the target level transitions and had low variance while the target dwelled on a level.     

Multiple site electromyograph amplitude estimation

Temporal whitening of individual surface electromyograph (EMG) waveforms and spatial combination of multiple recording sites have separately been demonstrated to improve the performance of EMG amplitude estimation. This investigation combined these two techniques by first whitening, then combining the data from multiple EMG recording sites to form an EMG amplitude estimate. A phenomenological mathematical model of multiple sites of the surface EMG waveform, with analytic solution for an optimal amplitude estimate, is presented. Experimental surface EMG waveforms were then sampled from multiple sites during nonfatiguing, constant-force, isometric contractions of the biceps or triceps muscles, over the range of 10-75% maximum voluntary contraction. A signal-to-noise ratio (SNR) was computed from each amplitude estimate (deviations about the mean value of the estimate were considered as noise). Results showed that SNR performance: 1) increased with the number of EMG sites, 2) was a function of the sampling frequency, 3) was predominantly invariant to various methods of determining spatial uncorrelation filters, 4) was not sensitive to the intersite correlations of the electrode configuration investigated, and 5) was best at lower levels of contraction. A moving average root mean square estimator (245-ms window) provided an average ± standard deviation (A±SD) SNR of 10.7±3.3 for single site unwhitened recordings. Temporal whitening and four combined sites improved the A±SD SNR to 24.6±10.4. On one subject, eight whitened combined sites were achieved, providing an A±SD SNR of 35.0±13.4     

The relationship between the surface EMG and force transients in muscle: Simulation and experimental studies

The surface electromyogrmn (EMG) is an easily measured signal which when quantified by present techniques is a reliable measure of whether a muscle is active, a fairly reliable measure of steady state force and a rather unreliable measure of force transients in muscle. There is a real need for a reliable indicator of dynamic changes in muscle activity for the control of prosthetics, in diagnosis of neuromuscular diseases, in studies of the motor control system and in fundamental studies of muscle mechanics. This paper outlines the principles underlying the development of force and the EMG in muscle. The EMG is a poor estimate of muscle force since it is the result of the linear superposition of biphasic action potentials which result in an interference pattern. This in turn is dependent on the details of the firing intervals for each motor unit, whereas the force is quite insensitive to these details. Experimental results for the human biceps brachii are described in which it was not possible to obtain a consistent estimate of muscle active state from the EMG. An extensive computer simulation was used to explore the relationship between EMG and force under a variety of assumptions. The conclusion is that it is technically impossible to obtain consistent estimates of muscle force (or active state) unless a filter with a time constant of 300 ms is applied to the rectified EMG. This is inconsistent with the estimation of active state for voluntary contractions with observed rise time constants of 30-70 ms. It is susgested that the only solution is to repeat an experiment many times and average the rectified EMG. Unfortunately, in practice it is difficult to repeat certain aspects of voluntary muscular contractions.     

Sequential Multistate EMG Signal Processor

A feature of multifunction EMG control is that the operator is always aware when a decision is made. Thus, a sequential processor which observes the EMG signal for the length of time necessary to make a reliable decision would be a logical one. In this paper, two sequential processors are developed for the EMG signal. Exact expressions for the processor performance, error probability, and the average number of samples required to make a decision are derived. The two processors are modeled on a computer and verified experimentally. Both the theoretical and experimental results show that compared to an optimal Bayes receiver the sequential processor, for a given efror probability, requires, on the average, approximately half the number of samples. Finally, a proof is given for the optimum signal set and this set is used in the paper.     

A multiple-layer finite-element model of the surface EMG signal

The effect of skin, muscle, fat, and bone tissue on simulated surface electromyographic (EMG) signals was examined using a finite-element model. The amplitude and frequency content of the surface potential were observed to increase when the outer layer of a homogeneous muscle model was replaced with highly resistive skin or fat tissue. The rate at which the surface potential decreased as the fiber was moved deeper within the muscle also increased. Similarly, the rate at which the surface potential decayed around the surface of the model, for a constant fiber depth, increased. When layers of subcutaneous fat of increasing thickness were then added to the model, EMG amplitude, frequency content, and the rate of decay of the surface EMG signal around the limb decreased, due to the increased distance between the electrodes and the active fiber. The influence of bone on the surface potential was observed to vary considerably, depending on its location. When located close to the surface of the volume conductor, the surface EMG signal between the bone and the source and directly over the bone increased, accompanied by a slight decrease on the side of the bone distal to the active fiber. The results emphasize the importance of distinguishing between the effects of material properties and the distance between source and electrode when considering the influence of subcutaneous tissue, and suggest possible distortions in the surface EMG signal in regions where a bone is located close to the skin surface.     

Influence of smoothing window length on electromyogram amplitudeestimates

A systematic, experimental study of the influence of smoothing window length on the signal-to-noise ratio (SNR) of electromyogram (EMG) amplitude estimates is described. Surface EMG waveforms were sampled during nonfatiguing, constant-force, constant-angle contractions of the biceps or triceps muscles, over the range of 10%-75% maximum voluntary contraction. EMG amplitude estimates were computed with eight different EMG processor schemes using smoothing length durations spanning 2.45-500 ms. An SNR was computed from each amplitude estimate (deviations about the mean value of the estimate were considered as noise). Over these window lengths, average ± standard deviation SNR's ranged from 1.4±0.28 to 16.2±5.4 for unwhitened single-channel EMG processing and from 3.2±0.7 to 37.3±14.2 for whitened, multiple-channel EMG processing (results pooled across contraction level). It was found that SNR increased with window length in a square root fashion. The shape of this relationship was consistent with classic theoretical predictions, however none of the processors achieved the absolute performance level predicted by the theory. These results are useful in selecting the length of the smoothing window in traditional surface EMG studies. In addition, this study should contribute to the development of EMG processors which dynamically tune the smoothing window length when the EMG amplitude is time varying     

Some optimal traffic regulation schemes for ATM networks: a Markovdecision approach

Some new traffic regulation schemes are defined in terms of a relief-spacing (or spacing of the allowance for cell delivery to the network) function. The class of open-loop traffic regulators (TR's) is defined in terms of relief-spacing functions which depend on some user-state; this class may be viewed as an extension of the Spacer-Controller defined in terms of some constant (user-state independent) relief-function. The optimal open-loop TR's are derived by formulating proper optimization problems and applying a Markov decision approach. Numerical results illustrate the improved performance of the optimal open-loop TR over that of the (constant relief-spacing) Spacer-Controller. Finally, the class of closed-loop TR's is defined in terms of relief-spacing functions which depend on both some user- and some network-state information and its optimal element is derived. The improved performance under the optimal closed-loop TR over that of the optimal open-loop TR is illustrated and their difference determines the performance gain if feedback information can become available on time     

基于粒子群优化-支持向量机方法的下肢肌电信号步态识别

为提高下肢表面肌电信号步态识别的准确性和实时性,该文提出一种基于粒子群优化(PSO)算法优化支持向量机(SVM)的模式识别方法。首先对消噪后的肌电信号提取积分肌电值和方差作为特征样本,然后利用PSO算法优化SVM的惩罚参数和核函数参数,最后利用步态动作的肌电信号样本数据对构造的SVM分类器进行训练、测试。实验结果表明PSO-SVM分类器对下肢正常行走5个步态的识别率,明显高于未经参数优化的SVM分类器,优化后平均识别率达到97.8%,并兼顾了分类的准确性和自适应性。     

Probability density of the surface electromyogram and its relationto amplitude detectors

When the surface electromyogram (EMG) generated from constant-force, constant-angle, nonfatiguing contractions is modeled as a random process, its density is typically assumed to be Gaussian. This assumption leads to root-mean-square (RMS) processing as the maximum likelihood estimator of the EMG amplitude (where EMG amplitude is defined as the standard deviation of the random process). Contrary to this theoretical formulation, experimental work has found the signal-to-noise-ratio [(SNR), defined as the mean of the amplitude estimate divided by its standard deviation] using mean-absolute-value (MAV) processing to be superior to RMS. This paper reviews RMS processing with the Gaussian model and then derives the expected (inferior) SNR performance of MAV processing with the Gaussian model. Next, a new model for the surface EMG signal, using a Laplacian density, is presented. It is shown that the MAV processor is the maximum likelihood estimator of the EMG amplitude for the Laplacian model. SNR performance based on a Laplacian model is predicted to be inferior to that of the Gaussian model by approximately 32%. Thus, minor variations in the probability distribution of the EMG may result in large decrements in SNR performance. Lastly, experimental data from constant-force, constant-angle, nonfatiguing contractions were examined. The experimentally observed densities fell in between the theoretic Gaussian and Laplacian densities. On average, the Gaussian density best fit the experimental data, although results varied with subject. For amplitude estimation, MAV processing had a slightly higher SNR than RMS processing.     

Development of a Biofeedback Therapeutic-Exercise-Supporting Manipulator

Although much equipment for physical therapy has been developed, equipment to improve the quality of physical therapy is scarce. We propose a robotic biofeedback exercise device that can display human joint torque and muscle force during training without a problematic electromyogram (EMG). The purpose is to increase the therapeutic value by understanding a person's condition during exercise and to provide an incentive to improve performance. The manipulator supports lower limb rehabilitation in the sagittal plane. With its ability to adjust the maximum speed and the time constant, the manipulator provides simultaneous and safe isokinetic exercise for the knee and hip joints. This paper describes the estimation of the human joint torque and muscle force. The display of the joint torque and the muscle force is realized during exercise of the knee joint using the developed manipulator. The estimation of the muscle force from Crowninshield's method and Hase's method generally agrees with the EMG.     

A single-lead ECG enhancement algorithm using a regularized data-driven filter

We presented a novel way of deriving a subspace filter for enhancing a noisy electrocardiogram (ECG) signal contaminated by electromyogram (EMG). The new subspace filter was based on a multiple cycle prediction (MCP) modeling of a single-lead ECG. The adoption of an MCP model resulted in a data matrix more suitable for separating noise and signal subspaces than the linear prediction (LP) model that is implicitly assumed in many existing subspace filters. Alignment of ECG cycles of different length is required for MCP modeling and was handled by a dynamic time warping (DTW) algorithm. A run-time procedure was designed for automatically determining the signal space dimension adaptively. To validate the new filter in a quantitative way, 12 clean realistic ECG segments with different degrees of heart rate variability generated using the ECGSyn program were mixed with different realizations of EMG noise in the MIT-BIH Noise Stress Test Database and locally acquired EMG at a typical 10-dB signal-to-noise ratio. The performance of the proposed method was compared to three existing ECG enhancement algorithms and achieved encouraging results. In addition, various ECG recordings from MIT-Arrythmia database were also mixed with EMG noise and subjected to the same four filters resulting in a qualitative comparison of them.     

A Supervised Feature-Projection-Based Real-Time EMG Pattern Recognition for Multifunction Myoelectric Hand Control

Electromyographic (EMG) pattern recognition is essential for the control of a multifunction myoelectric hand. The main goal of this study was to develop an efficient feature- projection method for EMG pattern recognition. To this end, a linear supervised feature projection is proposed that utilizes a linear discriminant analysis (LDA). First, a wavelet packet transform (WPT) is performed to extract a feature vector from four-channel EMG signals. To dimensionally reduce and cluster the WPT features, an LDA, then, incorporates class information into the learning procedure, and identifies a linear matrix to maximize the class separability for the projected features. Finally, a multilayer perceptron classifies the LDA-reduced features into nine hand motions. To evaluate the performance of the LDA for WPT features, the LDA is compared with three other feature-projection methods. From a visualization and quantitative comparison, it is shown that the LDA produces a better performance for the class separability, plus the LDA-projected features improve the classification accuracy with a short processing time. A real-time pattern-recognition system is then implemented for a multifunction myoelectric hand. Experiments show that the proposed method achieves a 97.4% recognition accuracy, and all processes, including the generation of control commands for the myoelectric hand, are completed within 97 ms. Consequently, these results confirm that the proposed method is applicable to real-time EMG pattern recognition for multifunction myoelectric hand control.     

Single site electromyograph amplitude estimation

Previous investigators have experimentally demonstrated and/or analytically predicted that temporal whitening of the surface electromyograph (EMG) waveform prior to demodulation improves the EMG amplitude estimate. However, no systematic study of the influence of various whitening filters upon amplitude estimate performance has been reported. The authors describe a phenomenological mathematical model of a single site of the surface EMG waveform and reports on experimental studies which examined the performance of several temporal whitening filters. Surface EMG waveforms were sampled during nonfatiguing, constant-force, isometric contractions of the biceps or triceps muscles, over the range of 10-75% maximum voluntary contraction. A signal-to-noise ratio (SNR) was computed from each amplitude estimate (deviations about the mean value of the estimate were considered as noise). A moving average root mean square estimator (245 ms window) provided an average±standard deviation (A±SD) SNR of 10.7±3.3 for the individual recordings. Temporal whitening with one fourth-order whitening filter designed per site improved the A±SD SNR to 17.6±6.0     

Blind separation of linear instantaneous mixtures of nonstationary surface myoelectric signals

Electromyographic (EMG) recordings detected over the skin may be mixtures of signals generated by different active muscles due to the phenomena related to volume conduction. Separation of the sources is necessary when single muscle activity has to be detected. Signals generated by different muscles may be considered uncorrelated but in general overlap in time and frequency. Under certain assumptions, mixtures of surface EMG signals can be considered as linear instantaneous but no a priori information about the mixing matrix is available when different muscles are active. In this study, we applied blind source separation (BSS) methods to separate the signals generated by two active muscles during a force-varying task. As the signals are non stationary, an algorithm based on spatial time-frequency distributions was applied on simulated and experimental EMG signals. The experimental signals were collected from the flexor carpi radialis and the pronator teres muscles which could be activated selectively for wrist flexion and rotation, respectively. From the simulations, correlation coefficients between the reference and reconstructed sources were higher than 0.85 for signals largely overlapping both in time and frequency and for signal-to-noise ratios as low as 5 dB. The Choi-Williams and Bessel kernels, in this case, performed better than the Wigner-Ville one. Moreover, the selection of time-frequency points for the procedure of joint diagonalization used in the BSS algorithm significantly influenced the results. For the experimental signals, the interference of the other source in each reconstructed source was significantly attenuated by the application of the BSS method. The ratio between root-mean-square values of the signals from the two sources detected over one of the muscles increased from (mean +/- standard deviation) 2.33 +/- 1.04 to 4.51 +/- 1.37 and from 1.55 +/- 0.46 to 2.72 +/- 0.65 for wrist flexion and rotation, respectively. This increment was statistically significant. It was concluded that the BSS approach applied is promising for the separation of surface EMG signals, with applications ranging from muscle assessment to detection of muscle activation intervals, and to the control of myoelectric prostheses.     

Adaptive filtering of the electromyographic signal for prostheticcontrol and force estimation

An adaptive time constant filter is derived for electromyographic (EMG) signal processing in prosthetic control applications. The analysis indicates that the mean-squared estimation error can be reduced by varying the time constant of the filter as a function of the signal and its derivative. Results of several experiments indicated this filter provides faster response and smaller estimation error than several previously available filters     

Analysis and comparison of reflex times and electromyograms of cervical muscles under impact loading using surface and fine-wire electroces

Myoelectric signals [electromyograms (EMGs)] can be collected using either surface or fine-wire electrodes. Application of the latter results in higher-frequency contents of EMG. In the field of impact biomechanics, surface electrodes are more often utilized than fine-wire ones. However, the removal of motion artefacts from EMG recorded under transient loads requires application of high-pass filters with relatively high cutoff frequencies, which may eliminate a significant part of the surface EMG power spectra. Therefore, in the current study, both surface and fine-wire electrodes were utilized to record the EMG of cervical muscles under conditions simulating a rear-end car collision at low speed. The results indicated that application of high-pass filtering at 50 Hz can be necessary to remove motion artefacts from the EMG collected under such conditions. Such filtering resulted in a higher decrease in amplitude of the surface EMG than that of the fine-wire one. However, the reflex times obtained here were not significantly affected by the type of the electrodes utilized to collect EMG.     

Behavior of the partial correlation coefficients of a least squareslattice filter in the presence of a nonstationary chirp input

This paper studies the performance of the a posteriori recursive least squares lattice filter in the presence of a nonstationary chirp signal. The forward and backward partial correlation (PARCOR) coefficients for a Wiener-Hopf optimal filter are shown to be complex conjugates for the general case of a nonstationary input with constant power. Such an optimal filter is compared to a minimum mean square error based least squares lattice adaptive filter. Expressions are found for the behavior of the first stage of the adaptive filter based on the least squares algorithm. For the general nth stage, the PARCOR coefficients of the previous stages are assumed to have attained Wiener-Hopf optimal steady state. The PARCOR coefficients of such a least squares adaptive filter are compared with the optimal coefficients for such a nonstationary input. The optimal lattice fitter is seen to track a chirp input without any error, and the tracking lag in such an adaptive filter is due to the least squares update procedure. The expression for the least squares based PARCOR coefficients are found to contain two terms: a decaying convergence term due to the weighted estimation procedure and a tracking component that asymptotically approaches the optimal coefficient value. The rate of convergence is seen to depend inversely on the forgetting factor. The tracking lag of the filter is derived as a function of the rate of nonstationarity and the forgetting factor. It is shown that for a given chirp rate there is a threshold adaptation constant below which the total tracking error is negligible. For forgetting factors above this threshold, the error increases nonlinearly. Further, this threshold forgetting factor decreases with increasing chirp rate. Simulations are presented to validate the analysis