Modeling nonlinear errors in surface electromyography due to baseline noise: A new methodology

The surface electromyographic (EMG) signal is often contaminated by some degree of baseline noise. It is customary for scientists to subtract baseline noise from the measured EMG signal prior to further analyses based on the assumption that baseline noise adds linearly to the observed EMG signal. The stochastic nature of both the baseline and EMG signal, however, may invalidate this assumption. Alternately, “true” EMG signals may be either minimally or nonlinearly affected by baseline noise. This information is particularly relevant at low contraction intensities when signal-to-noise ratios (SNR) may be lowest. Thus, the purpose of this simulation study was to investigate the influence of varying levels of baseline noise (approximately 2–40% maximum EMG amplitude) on mean EMG burst amplitude and to assess the best means to account for signal noise. The simulations indicated baseline noise had minimal effects on mean EMG activity for maximum contractions, but increased nonlinearly with increasing noise levels and decreasing signal amplitudes. Thus, the simple baseline noise subtraction resulted in substantial error when estimating mean activity during low intensity EMG bursts. Conversely, correcting EMG signal as a nonlinear function of both baseline and measured signal amplitude provided highly accurate estimates of EMG amplitude. This novel nonlinear error modeling approach has potential implications for EMG signal processing, particularly when assessing co-activation of antagonist muscles or small amplitude contractions where the SNR can be low.     

Estimation of monopolar signals from sphincter muscles and removal of common mode interference

The detection of surface electromyogram (EMG) by multi-electrode systems is applied in many research studies. The signal is usually recorded by means of spatial filters (linear combination of the potential under at least two electrodes) with vanishing sum of weights. Nevertheless, more information could be extracted from monopolar signals measured with respect to a reference electrode away from the muscle. Under certain conditions, surface EMG signal along a curve parallel to the fibre path has zero mean (property approximately satisfied when EMG is sampled by an array of electrodes that covers the entire support of the signal in space). This property allows estimating monopolar from single differential (SD) signals by pseudoinversion of the matrix relating monopolar to SD signals. The method applies to EMG signals from the external anal sphincter muscle, recorded using a specific cylindrical probe with an array of electrodes located along the circular path of the fibres. The performance of the algorithm for the estimation of monopolar from SD signals is tested on simulated signals. The estimation error of monopolar signals decreases by increasing the number of channels. Using at least 12 electrodes, the estimation error is negligible. The method applies to single fibre action potentials, single motor unit action potentials, and interference signals.The same method can also be applied to reduce common mode interference from SD signals from muscles with rectilinear fibres. In this case, the last SD channel defined as the difference between the potentials of the last and the first electrodes must be recorded, so that the sum of all the SD signals vanishes. The SD signals estimated from the double differential signals by pseudoinvertion are free of common mode.     

On a recurrent neural network producing oscillations

A recurrent two-node neural network producing oscillations is analyzed. The network has no true inputs and the outputs from the network exhibit a circular phase portrait. The weight configuration of the network is investigated, resulting in analytical weight expressions, which are compared with numerical weight estimates obtained by training the network on the desired trajectories. The values predicted by the analytical expressions agree well with the findings from the numerical study, and can also explain the asymptotic properties of the networks studied.     

Neuromuscular adjustments that constrain submaximal EMG amplitude at task failure of sustained isometric contractions

The amplitude of the surface EMG does not reach the level achieved during a maximal voluntary contraction force at the end of a sustained, submaximal contraction, despite near-maximal levels of voluntary effort. The depression of EMG amplitude may be explained by several neural and muscular adjustments during fatiguing contractions, including decreased net neural drive to the muscle, changes in the shape of the motor unit action potentials, and EMG amplitude cancellation. The changes in these parameters for the entire motor unit pool, however, cannot be measured experimentally. The present study used a computational model to simulate the adjustments during sustained isometric contractions and thereby determine the relative importance of these factors in explaining the submaximal levels of EMG amplitude at task failure. The simulation results indicated that the amount of amplitude cancellation in the simulated EMG (～ 40%) exhibited a negligible change during the fatiguing contractions. Instead, the main determinant of the submaximal EMG amplitude at task failure was a decrease in muscle activation (number of muscle fiber action potentials), due to a reduction in the net synaptic input to motor neurons, with a lesser contribution from changes in the shape of the motor unit action potentials. Despite the association between the submaximal EMG amplitude and reduced muscle activation, the deficit in EMG amplitude at task failure was not consistently associated with the decrease in neural drive (number of motor unit action potentials) to the muscle. This indicates that the EMG amplitude cannot be used as an index of neural drive.     

Interpretation of EMG integral or RMS and estimates of “neuromuscular efficiency” can be misleading in fatiguing contraction

In occupational and sports physiology, reduction of neuromuscular efficiency (NME) and elevation of amplitude characteristics, such as root mean square (RMS) or integral of surface electromyographic (EMG) signals detected during fatiguing submaximal contraction are often related to changes in neural drive. However, there is data showing changes in the EMG integral (I_EMG) and RMS due to peripheral factors. Causes for these changes are not fully understood. On the basis of computer simulation, we demonstrate that lengthening of intracellular action potential (IAP) profile typical for fatiguing contraction could affect EMG amplitude characteristics stronger than alteration in neural drive (central factors) defined by number of active motor units (MUs) and their firing rates. Thus, relation of these EMG amplitude characteristics only to central mechanisms can be misleading. It was also found that to discriminate between changes in RMS or I_EMG due to alterations in neural drive from changes due to alterations in peripheral factors it is better to normalize RMS of EMG signals to the RMS of M-wave. In massive muscles, such normalization is more appropriate than normalization to either peak-to-peak amplitude or area of M-wave proposed in literature.     

Neural muscle activation detection: A deep learning approach using surface electromyography

The timing of muscles activation which is a key parameter in determining plenty of medical conditions can be greatly assessed by the surface EMG signal which inherently carries an immense amount of information. Many techniques for measuring muscle activity detection exist in the literature. However, due to the complex nature of the EMG signal as well as the interference from other muscles that is observed during the measurement of the EMG signal, the accuracy of these techniques is compromised. In this paper, we introduce the neural muscle activation detection (NMAD) framework that detects the muscle activation based on deep learning. The main motivation behind using deep learning is to allow the neural network to detect based on the appropriate signal features instead of depending on certain assumptions. Not only the presented approach significantly improves the accuracy of timing detection, but because of the training nature, it can adapt to operate under different levels of interference and signal-to-noise ratio.     

Identification of connectivity in neural networks.

Analytical and experimental methods are provided for estimating synaptic connectivities from simultaneous recordings of multiple neurons. The results are based on detailed, yet flexible neuron models in which spike trains are modeled as general doubly stochastic point processes. The expressions derived can be used with nonstationary or stationary records, and can be readily extended from pairwise to multineuron estimates. Furthermore, we show analytically how the estimates are improved as more neurons are sampled, and derive the appropriate normalizations to eliminate stimulus-related correlations. Finally, we illustrate the use and interpretation of the analytical expressions on simulated spike trains and neural networks, and give explicit confidence measures on the estimates.     

Comparison of algorithms for estimation of EMG variables during voluntary isometric contractions.

Many algorithms have been described in the literature for estimating amplitude, frequency variables and conduction velocity of the surface EMG signal detected during voluntary contractions. They have been used in different application areas for the non invasive assessment of muscle functions. Although many studies have focused on the comparison of different methods for information extraction from surface EMG signals, they have been carried out under different conditions and a complete comparison is not available. It is the purpose of this paper to briefly review the most frequently used algorithms for EMG variable estimation, compare them using computer generated as well as real signals and outline the advantages and drawbacks of each. In particular the paper focuses on the issue of EMG amplitude estimation with and without pre-whitening of the signal, mean and median frequency estimation with periodogram and autoregressive based algorithms both in stationary and non-stationary conditions, delay estimation for the calculation of muscle fiber conduction velocity.     

Subject-specific calibration of neuromuscular parameters enables neuromusculoskeletal models to estimate physiologically plausible hip joint contact forces in healthy adults

In-vivo hip joint contact forces (HJCF) can be estimated using computational neuromusculoskeletal (NMS) modelling. However, different neural solutions can result in different HJCF estimations. NMS model predictions are also influenced by the selection of neuromuscular parameters, which are either based on cadaveric data or calibrated to the individual. To date, the best combination of neural solution and parameter calibration to obtain plausible estimations of HJCF have not been identified. The aim of this study was to determine the effect of three electromyography (EMG)-informed neural solution modes (EMG-driven, EMG-hybrid, and EMG-assisted) and static optimisation, each using three different parameter calibrations (uncalibrated, minimise joint moments error, and minimise joint moments error and peak HJCF), on the estimation of HJCF in a healthy population (n = 23) during walking. When compared to existing in-vivo data, the EMG-assisted mode and static optimisation produced the most physiologically plausible HJCF when using a NMS model calibrated to minimise joint moments error and peak HJCF. EMG-assisted mode produced first and second peaks of 3.55 times body weight (BW) and 3.97 BW during walking; static optimisation produced 3.75 BW and 4.19 BW, respectively. However, compared to static optimisation, EMG-assisted mode generated muscle excitations closer to recorded EMG signals (average across hip muscles R² = 0.60 ± 0.37 versus R² = 0.12 ± 0.14). Findings suggest that the EMG-assisted mode combined with minimise joint moments error and peak HJCF calibration is preferable for the estimation of HJCF and generation of realistic load distribution across muscles.     

Removal of ECG interference from surface respiratory electromyography

The signal to noise ratio (SNR) of surface respiratory electromyography signal is very low. Indeed EMG signal is contaminated by different types of noise especially the cardiac artefact ECG. This article explores the problem of removing ECG artefact from respiratory EMG signal. The new method uses an adaptive structure with an electrocardyographic ECG reference signal carried out by wavelet decomposition. The proposed algorithm requires only one channel to both estimating the adaptive filter input reference noise and the respiratory EMG signal. This new technique demonstrates how two steps will be combined: the first step decomposes the signal with forward discrete wavelet transform into sub-bands to get the wavelet coefficients. Then, an improved soft thresholding function was applied. And the ECG input reference signal is reconstructed with the transformed coefficients whereas, the second uses an adaptive filter especially the LMS one to remove the ECG signal. After trying statistical as well as mathematical analysis, the complete investigation ensures that all details and steps make proof that our rigorous method is appropriate. Compared to the results obtained using previous techniques, the results achieved using the new algorithm show a significant improvement in the efficiency of the ECG rejection.     

Increased rate of force development and neural drive of human skeletal muscle following resistance training.

The maximal rate of rise in muscle force [rate of force development (RFD)] has important functional consequences as it determines the force that can be generated in the early phase of muscle contraction (0-200 ms). The present study examined the effect of resistance training on contractile RFD and efferent motor outflow ("neural drive") during maximal muscle contraction. Contractile RFD (slope of force-time curve), impulse (time-integrated force), electromyography (EMG) signal amplitude (mean average voltage), and rate of EMG rise (slope of EMG-time curve) were determined (1-kHz sampling rate) during maximal isometric muscle contraction (quadriceps femoris) in 15 male subjects before and after 14 wk of heavy-resistance strength training (38 sessions). Maximal isometric muscle strength [maximal voluntary contraction (MVC)] increased from 291.1 +/- 9.8 to 339.0 +/- 10.2 N. m after training. Contractile RFD determined within time intervals of 30, 50, 100, and 200 ms relative to onset of contraction increased from 1,601 +/- 117 to 2,020 +/- 119 (P < 0.05), 1,802 +/- 121 to 2,201 +/- 106 (P < 0.01), 1,543 +/- 83 to 1,806 +/- 69 (P < 0.01), and 1,141 +/- 45 to 1,363 +/- 44 N. m. s(-1) (P < 0.01), respectively. Corresponding increases were observed in contractile impulse (P < 0.01-0.05). When normalized relative to MVC, contractile RFD increased 15% after training (at zero to one-sixth MVC; P < 0.05). Furthermore, muscle EMG increased (P < 0.01-0.05) 22-143% (mean average voltage) and 41-106% (rate of EMG rise) in the early contraction phase (0-200 ms). In conclusion, increases in explosive muscle strength (contractile RFD and impulse) were observed after heavy-resistance strength training. These findings could be explained by an enhanced neural drive, as evidenced by marked increases in EMG signal amplitude and rate of EMG rise in the early phase of muscle contraction.     

A novel approach for removing ECG interferences from surface EMG signals using a combined ANFIS and wavelet

In recent years, the removal of electrocardiogram (ECG) interferences from electromyogram (EMG) signals has been given large consideration. Where the quality of EMG signal is of interest, it is important to remove ECG interferences from EMG signals. In this paper, an efficient method based on a combination of adaptive neuro-fuzzy inference system (ANFIS) and wavelet transform is proposed to effectively eliminate ECG interferences from surface EMG signals. The proposed approach is compared with other common methods such as high-pass filter, artificial neural network, adaptive noise canceller, wavelet transform, subtraction method and ANFIS. It is found that the performance of the proposed ANFIS–wavelet method is superior to the other methods with the signal to noise ratio and relative error of 14.97 dB and 0.02 respectively and a significantly higher correlation coefficient (p < 0.05).     

Nature of the coupling between neural drive and force-generating capacity in the human quadriceps muscle

The force produced by a muscle depends on both the neural drive it receives and several biomechanical factors. When multiple muscles act on a single joint, the nature of the relationship between the neural drive and force-generating capacity of the synergistic muscles is largely unknown. This study aimed to determine the relationship between the ratio of neural drive and the ratio of muscle force-generating capacity between two synergist muscles (vastus lateralis (VL) and vastus medialis (VM)) in humans. Twenty-one participants performed isometric knee extensions at 20 and 50% of maximal voluntary contractions (MVC). Myoelectric activity (surface electromyography (EMG)) provided an index of neural drive. Physiological cross-sectional area (PCSA) was estimated from measurements of muscle volume (magnetic resonance imaging) and muscle fascicle length (three-dimensional ultrasound imaging) to represent the muscles'' force-generating capacities. Neither PCSA nor neural drive was balanced between VL and VM. There was a large (r = 0.68) and moderate (r = 0.43) correlation between the ratio of VL/VM EMG amplitude and the ratio of VL/VM PCSA at 20 and 50% of MVC, respectively. This study provides evidence that neural drive is biased by muscle force-generating capacity, the greater the force-generating capacity of VL compared with VM, the stronger bias of drive to the VL.     

Noise and signal power and their effects on evoked potential estimation

Signal power, noise power and their ratio (SNR) are important variables underlying estimation of evoked potential signals, yet, they are rarely explicitly considered in the design or analysis of EP experiments. A model is developed which relates the reliability of the average evoked potential (AEP) wave form to signal power, noise power, SNR, and the number of single trials included in the average. Measurements taken from auditory and visual EP experimental in elderly subjects show that noise power is highly reliable across experimental conditions and probably reflects global CNS anatomic or physiologic factors. In contrast, signal power and SNR are variable across conditions and sensory modalities, but are stable across replications. Thus signal power reflects CNS processes specific to the experimental paradigm. These results have importance for EP estimation. The expected reliability of the AEP cannot be adequately predicted from estimates of a subject's noise power, or from SNR estimated under different experimental conditions. These findings suggest the need for on-line estimation of SNR during data acquisition to ensure adequate reliability of AEPs.     

The influence of hip strength on gluteal activity and lower extremity kinematics

The effects of hip muscle strength and activation on anterior cruciate ligament injury biomechanics, particularly knee valgus loading, have been reported in isolation and with equivocal results. However, the combination of these factors influences joint biomechanics. This investigation evaluated the influence of hip strength on gluteal activation and knee valgus motion. Maximal isometric hip abduction (ABD) and external rotation (ER) contractions were used to define High and Low strength groups. Knee kinematics and gluteus maximus (GMax) and medius (GMed) EMG amplitudes obtained during landing were compared between High and Low strength groups after controlling for the potential confounding influence of sex. Knee valgus motion did not differ between the High and Low hip ABD and ER strength groups. However, the Low ABD and ER strength groups displayed greater GMed and GMax EMG amplitudes, respectively, compared to the High strength groups. These findings suggest that weaker individuals compensate for a lack of force production via heightened neural drive. As such, hip muscle strength influences knee valgus motion indirectly by determining neural drive requirements.     

The extraction of neural strategies from the surface EMG.

This brief review examines some of the methods used to infer central control strategies from surface electromyogram (EMG) recordings. Among the many uses of the surface EMG in studying the neural control of movement, the review critically evaluates only some of the applications. The focus is on the relations between global features of the surface EMG and the underlying physiological processes. Because direct measurements of motor unit activation are not available and many factors can influence the signal, these relations are frequently misinterpreted. These errors are compounded by the counterintuitive effects that some system parameters can have on the EMG signal. The phenomenon of crosstalk is used as an example of these problems. The review describes the limitations of techniques used to infer the level of muscle activation, the type of motor unit recruited, the upper limit of motor unit recruitment, the average discharge rate, and the degree of synchronization between motor units. Although the global surface EMG is a useful measure of muscle activation and assessment, there are limits to the information that can be extracted from this signal.     

Artificial neural network learning of nonstationary behavior in time series

We refine and complement a previously-proposed artificial neural network method for learning hidden signals forcing nonstationary behavior in time series. The method adds an extra input unit to the network and feeds it with the proposed profile for the unknown perturbing signal. The correct time evolution of this new input parameter is learned simultaneously with the intrinsic stationary dynamics underlying the series, which is accomplished by minimizing a suitably-defined error function for the training process. We incorporate here the use of validation data, held out from the training set, to accurately determine the optimal value of a hyperparameter required by the method. Furthermore, we evaluate this algorithm in a controlled situation and show that it outperforms other existing methods in the literature. Finally, we discuss a preliminary application to the real-world sunspot time series and link the obtained hidden perturbing signal to the secular evolution of the solar magnetic field.     

Reflex facilitation during the stretch-shortening cycle.

Maximal torque during the concentric phase of a movement has been shown to be enhanced by prior eccentric muscle actions, a movement strategy referred to as the stretch-shortening cycle. Although the mechanical basis for this enhancement is well established, the neural component is not. We hypothesized that brief high-frequency bursts of spindle afferent discharge during the eccentric phase of the stretch-shortening cycle could be one mechanism for facilitating the volitional drive. To test this hypothesis, three sets of experiments were done. In the first (N=15), we demonstrated that both the peak and mean EMG of the soleus (S) and lateral gastrocnemius (LG) muscles were considerably greater during a reciprocal hopping (RHOP) task than for maximum isometric contractions (MIVCs). In the second experiment, we tested whether the dynamic nature of the RHOP or the eccentric phase of the RHOP contributed to the EMG potentiation. Peak and mean EMG produced with a concentric hop (CHOP), in which the lengthening phase of the hop was eliminated, were compared with that produced with the RHOP and MIVCs conditions (N=7). The RHOP produced greater peak EMG than either the CHOP or the MIVCs while the mean EMG for both hopping conditions was considerably more than the MIVCs. In the final experiment, we attempted to mimic the brief high-frequency burst of spindle afferent activity during the lengthening phase of the stretch-shortening cycle in the absence of muscle length changes. High-frequency (100 Hz) afferent stimulation (HFS) was delivered during MIVCs. At rest, the HFS produced negligible EMG activity but when superimposed over MIVCs produced a marked potentiation of the S EMG over values obtained during MIVCs alone. Evidence that HFS synchronizes the EMG associated with volitional activation is also provided. We conclude that a substantial but brief facilitation and possible synchronization of the neural drive is provided by the spindle afferents during the eccentric phase of the stretch-shortening cycle.     

Comparison of methods for estimating motor unit firing rate time series from firing times

The central nervous system regulates recruitment and firing of motor units to modulate muscle tension. Estimation of the firing rate time series is typically performed by decomposing the electromyogram (EMG) into its constituent firing times, then lowpass filtering a constituent train of impulses. Little research has examined the performance of different estimation methods, particularly in the inevitable presence of decomposition errors. The study of electrocardiogram (ECG) and electroneurogram (ENG) firing rate time series presents a similar problem, and has applied novel simulation models and firing rate estimators. Herein, we adapted an ENG/ECG simulation model to generate realistic EMG firing times derived from known rates, and assessed various firing rate time series estimation methods. ENG/ECG-inspired rate estimation worked exceptionally well when EMG decomposition errors were absent, but degraded unacceptably with decomposition error rates of ⩾1%. Typical EMG decomposition error rates—even after expert manual review—are 3–5%. At realistic decomposition error rates, more traditional EMG smoothing approaches performed best, when optimal smoothing window durations were selected. This optimal window was often longer than the 400 ms duration that is commonly used in the literature. The optimal duration decreased as the modulation frequency of firing rate increased, average firing rate increased and decomposition errors decreased. Examples of these rate estimation methods on physiologic data are also provided, demonstrating their influence on measures computed from the firing rate estimate.