Correcting Temporal Inaccuracies in Labeled Training Data for Electromyographic Control Algorithms.

Background: Effective muscular recruitment during cycling is crucial to optimising performance.  As such, much research has examined when muscles are active during the pedal revolution and how muscle co-activation may influence pedalling technique but little evidence associated with the effect of fatigue on lower limb joint angles and its subsequent effect on power output and muscle recruitment.  Purpose: Are changes in muscular recruitment patterns and changes in lower limb kinematics detrimental to performance?  Methods: A preliminary case study of a well-trained veteran cyclist (V I‡O2peak  49.9 ml . kg-1 . min-1 and PPO of 400W) conducted a 16.1km simulated cycling time trial on a Wattbike Pro at 80% PPO with a cadence of 98 rpm.  Bilateral surface EMG were employed using pairs of Ag/AgCl electrodes adhered to both right and left vastus medialis (VM), biceps femoris (BF), tibialis anterior (TA) and gastrocnemius (GM).  Intervals of sixty seconds in duration, of 3D kinematic and electromyography (EMG) data were synchronised using a Vicon motion capture system throughout the time trial.  Ten consecutive pedal revolutions per 60 s interval were processed for subsequent analysis.    Kinematic and EMG data signals were imported into Visual 3D v.5 and Matlab for processing.  Results: There were statistically significant changes (p<0.05) in all kinematic data across nine time points.  Fundamental changes occurred between minute 10-12 with a decrease in flexion angle for the left hip (z = -9.56, p<0.05, r = 0.55), and the right hip, (z = -7.67, p<0.05, r = 0.44).  The left ankle starts to have a greater planter flexion (z = -9.69, p<0.05, r = 0.56) between minutes 10-12 and this continues to have a significant change between time points for the remainder of the trial.  The magnitude of EMG data revealed that there were statistically significant changes (p<0.05) in all muscles across seven time points.  Decreases in the magnitude of activation were observed between minutes 8-10 for both the left (z = -9.17, p<0.05, r = 0.53) and right (z = -8.21, p<0.0005, r = 0.48) TA.  The frequency domain EMG analysis indicated a trend towards lower frequencies being recorded for the left and right TA and GM with a trend towards higher frequencies being recorded for the left and right VM.  Analysis of the Wattbike data indicated that there were statistically significant differences from minutes 1 to 20 between left and right angle to peak force applied to the crank (t(19) = 20.89, p<0.05) with no statistical significance between left and right limb percentage use (95% CI 2.81, 3.31%).  Indications showed that there was a higher magnitude of activation for the left leg across all muscles tested and the Wattbike data indicates greater power generation on the left side of the pedal revolution.  Discussion: Based on the results from this study there were decreases in the ankles range of motion (ROM) with increases in hip flexion from minute 10-12 with changes in neuromuscular activity between minutes 8-10 then there were smaller effect changes for the remainder of the time trial.  The alterations in the EMG median frequency indicated a trend towards muscular fatigue for TA and GM with increases in neuromuscular activity in both VM’s possibly counteracting fatigue.  The ankle’s decrease in ROM may be a product of fatigue associated with the TA and GM but by becoming stiffer it may optimise the foot/pedal interface.  Even with these changes in kinematic and EMG data the power output remained constant.  Conclusion: As this was a case study further work in this area needs to be considered in establishing variability in joint angles and how they may be beneficial/detrimental in generating the same level of power output when fatigued.  An association of muscular fatigue in cycling and changes to lower limb joint angle variability requires further investigation examining joint coupling patterns, independent pedal power analysis and cross reference to physiological parameters for fatigue to establish strategies that may be beneficial in improving cycling performance.

A systematic integrated data collection and analysis of kinematic, kinetic, and electromyography (EMG) data allow for the comparison of differences in wheelchair propulsion between able-bodied individuals and persons with paraplegia. Kinematic data from a motion analysis system, kinetic data from force-sensing push rims, and electromyography data from four upper-limb muscles were collected for ten push strokes. Results are as follows: Individuals with paraplegia use a greater percentage of their posterior deltoids, biceps, and triceps in relation to maximal voluntary contraction. These persons also reached peak anterior deltoid firing nearly 10 deg earlier on the push rim, while reaching peak posterior deltoid nearly 10 deg later on the push rim. Able-bodied individuals had no triceps activity in the initial stages of propulsion while their paraplegic groups had activity throughout. Able-bodied participants also had, on average, peak resultant, tangential, and radial forces occurring later on the push rim (in degrees). There are two main conclusions that can be drawn from this integrative investigation: (1) A greater "muscle energy," as measured by the area under the curve of the percentage of EMG throughout propulsion, results in a greater resultant joint force in the shoulder and elbow, thus potentially resulting in shoulder pathology. (2) Similarly, a greater muscle energy may result in fatigue and play a factor in the development of shoulder pain and pathology over time; fatigue may compromise an effective propulsive stroke placing undue stresses on the joint capsule. Muscle activity differences may be responsible for the observed kinematic and kinetic differences between the two groups. The high incidence of shoulder pain in manual wheelchair users as compared to the general population may be the result of such differences, although the results from this biomedical investigation should be examined with caution. Future research into joint forces may shed light on this. Further investigation needs to focus on whether the pattern of kinematics, kinetics, and muscle activity during wheelchair propulsion is compensatory or evolutionary by tracking individuals longitudinally.

In order to mimic the natural appearance, motion, and perception of the human hand, a bio-mechatronic approach to design an anthropomorphic prosthetic hand — HIT/DLR Prosthetic Hand has been presented. It reproduces human hand in its fundamental structure such as appearance, weight, and dimensions. Its thumb can move along a cone surface in 3D space. Similar with that of human's, it combines with abduction and adduction from palmar position to lateral position. Actuated by only one motor, the middle finger, ring finger, and little finger can envelop complex-shaped objects. In addition, the bio-mechatronic integration and cosmetic designation make it much more like a genuine human hand. HIT/DLR Prosthetic Hand can be controlled through voice control strategy, Electromyography (EMG) control strategy, EMG, and electrocutaneous sensory feedback (ESF) close loop control strategy. In EMG control system, 10 types of hand posture are recognized through six electrodes on the basis of support vector machine (SVM). The last control strategy can help an amputee recover the grasp perception, further improve the efficiency of EMG control, and reduce the hand mis-manipulation and force delivery mistakes.

Background: In this study, different intent prediction strategies were explored with the objective of determining the best approach to predicting continuous multi-axial user motion based solely on surface EMG (electromyography) data. These strategies were explored as the first step to better facilitating control of a multi-axis transtibial powered prosthesis.Methods: Based on data acquired from gait experiments, different data sets, prediction approaches and classification algorithms were explored. The effect of varying EMG electrode positioning was also tested. EMG data measured from three lower leg muscles was the sole data type used for making intent predictions. The motions to be predicted were along both the sagittal plane (foot dorsiflexion and plantarflexion) and the frontal plane (foot eversion and inversion).Results: The deviation of EMG data from its optimal pattern led to a decrease in prediction accuracy of up to 23%. However, using features that were calculated based on a participant's specific walking pattern limited this loss of prediction accuracy as a result of EMG electrode placement. A decoupled data set, one wherein the terrain type was accounted for beforehand, yielded the highest intent prediction accuracy of 77.2%.Conclusions: The results of this study highlighted the challenges faced when using very limited EMG data to predict multi-axial ankle motion. They also indicated that approaches that are more user-centric by design could led to more accurate motion predictions, possibly enabling more intuitive control.

This article attempts to review papers on power assist rehabilitation robots, human motion intention, control laws, and estimation of power assist rehabilitation robots based on human motion intention in recent years. This paper presents the various ways in which human motion intention in rehabilitation can be estimated. This paper also elaborates on the control laws for the estimation of motion intention of the power assist rehabilitation robot. From the review, it has been found that the motion intention estimation method includes: Artificial Intelligence-based motion intention and Model-based motion intention estimation. The controllers include hybrid force/position control, EMG control, and adaptive control. Furthermore, Artificial Intelligence based motion intention estimation can be subdivided into Electromyography (EMG), Surface Electromyography (SEMG), Extreme Learning Machine (ELM), and Electromyography-based Admittance Control (EAC). Also, Model-based motion intention estimation can be subdivided into Impedance and Admittance control interaction. Having reviewed several papers, EAC and ELM are proposed for efficient motion intention estimation under artificial-based motion intention. In future works, Impedance and Admittance control methods are suggested under model-based motion intention for efficient estimation of motion intention of power assist rehabilitation robot. In addition, hybrid force/position control and adaptive control are suggested for the selection of control laws. The findings of this review paper can be used for developing an efficient power assist rehabilitation robot with motion intention to aid people with lower or upper limb impairment.

BackgroundThe flexion relaxation phenomenon (FRP) represents a well-studied neuromuscular response that occurs in the lumbar and cervical spine. However, the cervical spine FRP has not been investigated extensively, and the speed of movement and loading effects remains to be characterized. The objectives of the present study were to evaluate the influence of load and speed on cervical FRP electromyographic (EMG) and kinematic parameters and to assess the measurement of cervical FRP kinematic and EMG parameter repeatability.MethodsEighteen healthy adults (6 women and 12 men), aged 20 to 39 years, participated in this study. They undertook 2 sessions in which they had to perform a standardized cervical flexion/extension movement in 3 phases: complete cervical flexion; the static period in complete cervical flexion; and extension with return to the initial position. Two different rhythm conditions and 3 different loading conditions were applied to assess load and speed effects. Kinematic and EMG data were collected, and dependent variables included angles corresponding to the onset and cessation of myoelectric silence as well as the root mean square (RMS) values of EMG signals. Repeatability was examined in the first session and between the 2 sessions.ResultsStatistical analyses revealed a significant load effect (P < 0.001). An augmented load led to increased FRP onset and cessation angles. No load × speed interaction effect was detected in the kinematics data. A significant load effect (P < 0.001) was observed on RMS values in all phases of movement, while a significant speed effect (P < 0.001) could be seen only during the extension phase. Load × speed interaction effect was noted in the extension phase, where higher loads and faster rhythm generated significantly greater muscle activation. Intra-session and inter-session repeatability was good for the EMG and kinematic parameters.ConclusionsThe load increase evoked augmented FRP onset and cessation angles as well as heightened muscle activation. Such increments may reflect the need to enhance spinal stability under loading conditions. The kinematic and EMG parameters showed promising repeatability. Further studies are needed to assess kinematic and EMG differences between healthy subjects and patients with neck pain.

Classification of human hand movements based on EMG signals using nonlinear dimensionality reduction and data fusion techniques

Kinematic and electromyography (EMG) aspects of running on a firm surface and on soft, dry sand were studied to elucidate mechanisms contributing to the higher energy cost (EC) of sand running. Eight well-trained males (mean VO(2max) 64.3+/-8.6 ml.kg(-1).min(-1)) performed barefoot running trials on a firm surface (wooden floor) and on a soft, dry sand surface (track dimensions 8.8 mx60 cm; depth 13 cm) at 8 and 11 km.h(-1). Kinematic and EMG data were collected simultaneously using an integrated six-camera 50 Hz VICON motion analysis system, an AMTI force-plate and a 10-channel EMG system. Running at 8 km.h(-1) on sand resulted in a greater (P<0.05) stance time (t(s)) compared with the firm surface. At 11 km.h(-1), sand running resulted in a greater stance-to-stride ratio (P<0.005), a shorter stride length (SL) (P<0.05), and a greater cadence (P<0.001) compared with the firm surface values. Hip and knee flexion at initial foot contact (IFC), mid-support (MS) and flexion maximum were greater (P<0.001) running on sand compared with firm surface values at 8 and 11 km.h(-1). Over duration of stride, Hamstring (semimembranosus and biceps femoris) EMG was greater running on sand compared with the firm surface at 8 (P<0.001) and 11 (P<0.05) km.h(-1). During the stance phase in the 8-km.h(-1) trials, EMG in the Hamstrings (P<0.001), Vastii (Vastus lateralis and Vastus Medialis) (P<0.02), Rectus femoris (Rec Fem) (P<0.01) and Tensor Fascia Latae (Tfl) (P<0.0001) were greater than the firm surface measures. During stance in the 11-km.h(-1) trials, Tfl EMG was greater (P<0.02) running on sand compared with the firm surface. At IFC and MS, Hamstrings' EMG was greater on sand at both running speeds (P<0.001). For the Vastii (P<0.02), Rec Fem (P<0.0001) and Tfl (P<0.0001) muscles, the EMG at MS running on sand at both speeds was greater than the firm surface values. The increased EC of running on sand can be attributed in part to the increased EMG activation associated with greater hip and knee range of motion compared with firm surface running.

Reducing the effect of wrist variation on pattern recognition of Myoelectric Hand Prostheses Control through Dynamic Time Warping

Electromyography (EMG) signal classification is vital to diagnose musculoskeletal abnormalities and control devices by motion intention detection. Machine learning assists both areas by classifying conditions or motion intentions. This paper proposes a novel window length insensitive EMG classification method utilizing the Entropy feature. The main goal of this study is to show that entropy can be used as the only feature for fast real-time classification of EMG signals of hand gestures. The main goal of this study is to show that entropy can be used as the only feature for fast real-time classification of EMG signals of hand gestures. Additionally, the entropy feature can classify feature vectors of different sliding window lengths without including them in the training data. Many kinds of entropy feature succeeded in electroencephalography (EEG) and electrocardiography (ECG) classification research. However, to the best of our knowledge, the Entropy Feature proposed by Shannon stays untested for EMG classification to this day. All the machine learning models are tested on datasets NinaPro DB5 and the newly collected SingleMyo. As an initial analysis to test the entropy feature, classic Machine Learning (ML) models are trained on the NinaPro DB5 dataset. This stage showed that except for the K Nearest Neighbor (kNN) with high inference time, Support Vector Machines (SVM) gave the best validation accuracy. Later, SVM models trained with feature vectors created by 1 s (200 samples) sliding windows are tested on feature vectors created by 250 ms (50 samples) to 1500 ms (300 samples) sliding windows. This experiment resulted in slight accuracy differences through changing window length, indicating that the Entropy feature is insensitive to this parameter. Lastly, Locally Parsed Histogram (LPH), typical in standard entropy functions, makes learning hard for ML methods. Globally Parsed Histogram (GPH) was proposed, and classification accuracy increased from 60.35% to 89.06% while window length insensitivity is preserved. This study shows that Shannon’s entropy is a compelling feature with low window length sensitivity for EMG hand gesture classification. The effect of the GPH approach against an easy-to-make mistake LPH is shown. A real-time classification algorithm for the entropy features is tested on the newly created SingleMyo dataset.

Lower extremity injuries have immediate and long-term consequences. Lower extremity movement assessments can assist with identifying individuals at greater injury risk and guide injury prevention interventions. Movement assessments identify similar movement characteristics and evidence suggests large magnitude kinematic relationships exist between movement patterns observed across assessments; however, the magnitude of the relationships for electromyographic (EMG) measures across movement assessments remains largely unknown. This study examined relationships between lower extremity kinematic and EMG measures during jump landings and single leg squats. Lower extremity three-dimensional kinematic and EMG data were sampled from healthy adults (males = 20, females = 20) during the movement assessments. Pearson correlations examined the relationships of the kinematic and EMG measures and paired samples t-tests compared mean kinematic and EMG measures between the assessments. Overall, significant moderate correlations were observed for lower extremity kinematic (ravg = 0.41, rrange = 0.10–0.61) and EMG (ravg = 0.47, rrange = 0.32–0.80) measures across assessments. Kinematic and EMG measures were greater during the jump landings. Jump landings and single leg squats place different demands on the body and necessitate different kinematic and EMG patterns, such that these measures are not highly correlated between assessments. Clinicians should, therefore, use multiple assessments to identify aberrant movement and neuromuscular control patterns so that comprehensive interventions can be implemented.

Objective: Electric hand prostheses are typically controlled using electromyographic (EMG) signals recorded from the residual muscles. However, non-stationarities that are characteristic for EMG interfaces impair the reliability of machine-learning-based approaches during daily life activities-based approaches (e.g., the limb position effect). Including additional EMG-independent information in the classification algorithm may mitigate this problem. Methods: In this study, we systematically investigated an electrical impedance myography (EIM) interface for its possible utility as an additional source of information to EMG. To this goal, six different hand-wrist motions in three arm positions were recorded from ten able-bodied volunteers and three prosthetic hand users. EIM and EMG data were evaluated in terms of information content and classified using linear discriminant analysis (LDA). Results: EIM contained less information and was more strongly influenced by changing limb positions than EMG, but a combination of EIM and EMG outperformed EMG alone. Training with pooled data from multiple arm positions was necessary to mitigate the limb position effect. Conclusion: EIM can be valuable for myoelectric control as it contains complementary information to EMG, but it is also strongly influenced by changing arm positions. Significance: This paper provides fundamental insights required for advancing the application of EIM in the context of modern prosthesis control.

The intrinsic variability of the shoulder joint motion is a critical factor in the characterisation of the shoulder joint. However, traditional computational approaches struggle to account for it. On the other hand, the stochastic approach allows to identify a set of plausible solutions. In this study, the Myobolica toolbox, which yielded promising results in the lower limb, was employed to simulate a shoulder abduction, with twofold aims: to assess its generalisability to other joints, and to evaluate an electromyography (EMG)-informed version of Myobolica. Publicly available kinematics, EMG, and glenohumeral (GH) joint force data measured by an instrumented implant on a 64-year-old man executing three weighted shoulder abductions were used. A previously developed shoulder musculoskeletal model was employed to compute stochastic simulations informed and not with EMG data, sampling 1×105 solutions every 30 timeframes. The predicted GH joint force were compared to the experimental data, and the variance in the solutions across simulations was computed. Overall, the correlation between the GH joint force predicted by Myobolica and the experimental values increased when the EMG-based constraint was applied (from approximately R2=0.06, RMSE=1.82 BW to R2=0.6, RMSE=0.73 BW when all available EMG data were employed). Using EMG led to a reduction (from 2.3 to 0.65 BW) in the solution bandwidths. Providing EMG data to inform the simulations helped improve their accuracy. However, the results obtained otherwise remain promising. Additional work is required to minimize the computational cost of the Myobolica approach. A consistency gap between experimental data and the model is reported.

Internal consistencies of the delayed trunk muscle reaction times following a treadmill-induced slip perturbation while holding and not holding a tray

The purpose of this study was to explore the use of Independent Component Analysis (ICA) on surface Electromyography (EMG) data to distinguish between individual muscle activations due to its capabilities for signal separation. EMG data was gathered on seven participants using the Delsys Trigno Wireless EMG system. Participants performed specific movements which targeted the calves muscle group of the lower leg. EMG sensors were attached according to SENIAM recommendations and extra sensors were attached in non-recommended locations to achieve crosstalk. Signals were acquired using proprietary Delsys software and processed using the ICA algorithm in Matlab to explore crosstalk. Integrated EMG was calculated for all results using custom Matlab code. The results showed moderate levels of agreement between the mixed signals and the original signals (p < 0.01). However, further work is needed to determine the usefulness of the independent components.