sEMG Data Research Articles

Surface electromyography for testing motor dysfunction in amyotrophic lateral sclerosis

ObjectivesTo investigate the use of a set of dynamical features, extracted from surface electromyography, to study upper motor neuron (UMN) degeneration in amyotrophic lateral sclerosis (ALS). MethodsWe acquired surface EMG signals from the upper limb muscles of 13 ALS patients and 20 control subjects and classified them according to a novel set of muscle activity features, describing the temporal and frequency dynamic behavior of the signals, as well as measures of its complexity. Using a battery of classification approaches, we searched for the most discriminating combination of those features, as well as a suitable strategy to identify ALS. ResultsWe observed significant differences between ALS patients and controls, in particular when considering features highlighting differences between forearm and hand recordings, for which classification accuracies of up to 94% were achieved. The most robust discriminations were achieved using features based on detrended fluctuation analysis and peak frequency, and classifiers such as decision trees, random forest and Adaboost. ConclusionThe current work shows that it is possible to achieve good identification of UMN changes in ALS by taking into consideration the dynamical behavior of surface electromyographic (sEMG) data.

Feasibility assessment of transfer functions describing biomechanics of the human lower limb during the gait cycle

Background and objectiveTransfer functions could model biomechanical parameters to conveniently analyze the system dynamics of the lower limb during daily activities, e.g., walking. The current study evaluates the feasibility of transfer functions as a means of predicting surface electromyography (sEMG) of lower limb muscles based on axial tibial (ATA) and femoral (FA) accelerations. Since the transfer functions are comparable in accuracy to commonly used long short-term memory models (LSTM), the transfer function-based methodology can assist in the design of novel prostheses while being competitive with mainstream models. MethodsData were collected from eight participants with no medical history that would alter their gait cycles. The data included the sEMG of four primary muscle groups, ATA and FA, and heel-strike triggering signal, while the subjects walked at 5.28 km/h on a treadmill. A fast Fourier transform (FFT) was performed on the filtered ensemble averages of the ATA, FA and normalized average rectified sEMG signals, which were used to approximate the time domain Fourier series. Transfer functions relating ATA-to-sEMG and FA-to-sEMG were derived from the Fourier series equations to generate time-domain intra-subject predictions of sEMG signals. With respect to predictive accuracy, these transfer functions were compared with sEMG-to-sEMG LSTM models trained on sEMG data from some of the participants. It was hypothesized that the transfer functions would be at least comparable to the LSTM models. ResultsThe results indicate that the predicted sEMG signals were able to capture the temporal characteristics of the measured sEMG signal based on either ATA or FA. The muscle activities and acceleration were in good agreement with the walking gait cycle events. Only 13 Fourier terms were needed to effectively predict the sEMG from the acceleration signals, indicating the computational efficiency of the investigated analysis framework. For the ATA-to-sEMG transfer functions, the mean square difference (MSD) between the predicted and measured sEMG signals was low and comparable to the LSTM models. ConclusionsOverall, the feasibility and competitiveness of the transfer functions with LSTM models was confirmed with respect to predictive accuracy. Potential applications of the transfer functions include the control of powered prosthetics and the detection of gait pathologies.

A Novel Technique to Reject Artifact Components for Surface EMG Signals Recorded During Walking With Transcutaneous Spinal Cord Stimulation: A Pilot Study.

Transcutaneous spinal cord electrical stimulation (tSCS) is an emerging technology that targets to restore functionally integrated neuromuscular control of gait. The purpose of this study was to demonstrate a novel filtering method, Artifact Component Specific Rejection (ACSR), for removing artifacts induced by tSCS from surface electromyogram (sEMG) data for investigation of muscle response during walking when applying spinal stimulation. Both simulated and real tSCS contaminated sEMG data from six stroke survivors were processed using ACSR and notch filtering, respectively. The performance of the filters was evaluated with data collected in various conditions (e.g., simulated artifacts contaminating sEMG in multiple degrees, various tSCS intensities in five lower-limb muscles of six participants). In the simulation test, after applying the ACSR filter, the contaminated-signal was well matched with the original signal, showing a high correlation (r = 0.959) and low amplitude difference (normalized root means square error = 0.266) between them. In the real tSCS contaminated data, the ACSR filter showed superior performance on reducing the artifacts (96% decrease) over the notch filter (25% decrease). These results indicate that ACSR filtering is capable of eliminating artifacts from sEMG collected during tSCS application, improving the precision of quantitative analysis of muscle activity.

Semi-supervised style transfer mapping-based framework for sEMG-based pattern recognition with 1- or 2-DoF forearm motions

BackgroundThe development of body-worn sensors has made it easier to measure surface electromyograms (sEMGs) from individuals and to employ sEMG-based interfaces for user-centric health monitoring, rehabilitation, human augmentation, and amusement. However, it remains difficult to measure large amounts of sEMG data from a user (target). Thus, the development of a subject-to-subject transfer framework that uses the information that is available from other people (source) is challenging work. ObjectivesIn this study, we propose a subject-to-subject transfer framework, which includes the following four steps: (i) the construction of individual support vector machines (SVMs) from the source subjects; (ii) the selection of effective classifiers for the target based on the individual SVM classification results; (iii) the linear projection of the target data into the source subject data using the semi-supervised style transfer mapping algorithm; and (iv) an ensemble strategy of class probabilities for the selected classifiers. MethodsTo evaluate the performance of proposed framework, we collected 8-class, 1-DoF and 14-class, 2-DoF sEMG datasets that were acquired from the same 25 subjects using an eight-channel wearable device. The classifiers were trained with time-domain and autoregressive features. ResultsOur proposed method exhibited superior performance on both datasets compared to a conventional transfer framework using covariate shift adaptation. The sEMG datasets and sample codes used in this study are publicly available at https://github.com/Suguru55/SS-STM_for_MyoDatasets.

SEMG-based consecutive estimation of human lower limb movement by using multi-branch neural network

Surface electromyography (sEMG) has the potential for human lower limb movement analysis, including gait phases recognition and joint angle estimation, which can provide a great level of human interaction with the exoskeleton orthotic devices. In this paper, a method based on deep learning is proposed, which maps the multichannel sEMG signals to human lower limb movement, including 4 different gait phases and 3 flexion/extension joint angles. First, five time-domain features and spectrogram data as frequency domain features are extracted from the sEMG data from 8 muscles of right legs. Then, a multi-branch neural network (MBNN) with convolutional neural layers and recurrent neural layers is constructed, which uses both the extracted features and raw data as input to analyze human movement. Experimental results show that the mean accuracy of classification of our proposed methods can reach high level (90.92 ± 3.58% for speed dependent and 85.04 ± 5.14% for speed independent). Meanwhile, average of the root mean square error between estimated and real joint angles is (3.75 ± 1.52 degree for speed dependent and 6.12 ± 2.54 degree for speed independent). These results indicate that the proposed method can be used to facilitate adoption of exoskeleton orthotic device in real-life applications, with gait phases determining impedance characteristic of devices and angles estimating joint movement.

Alterations in Surface Electromyography of the Upper Leg Muscles at Specified Respiratory Exchange Ratio Thresholds Ranges During a Maximal Exercise Test

Introduction: Assessing muscle electromyography (EMG) in conjunction with physiological alterations to exercise may be valuable to determine a more holistic approach to exercise-induced fatigue. Methods: Thirteen, recreationally trained individuals (n = 7 female, n = 6 males) underwent a maximal exercise test. Throughout the test, physiological variables were measured in addition to surface electromyography (sEMG) of the upper legs. Physiological and sEMG data was then grouped in to four category thresholds based on respiratory exchange ratios (RER) greater than 0.95. Results: There was a main effect of group (p < 0.001) as an increase in exercise intensity assessed by RER threshold ranges resulted in a subsequent reductions in sEMG frequencies with the exception of the sEMG frequencies recorded at VO2peak (p < 0.055). Conclusions: Upper leg sEMG frequencies decrease with increases in high intensity exercise, with the exception of near maximal loads.

Deep learning for surface electromyography artifact contamination type detection

The quality of surface Electromyography (sEMG) signals could be an issue if highly contaminated by Power Line Interference (PLI), Electrocardiogram signal (ECG), Movement Artifact (MOA) or White Gaussian Noise (WGN), that could lead to unsafe operation of devices that is controlled by sEMG data, such as electro-mechanical prothesis. There are some mitigation methods proposed for some specifics sEMG contaminants and to use these methods in an efficient way is important to identify the contaminant in the sEMG signal. In this work we propose the use of a Recurrent Neural Network (RNN) using Long Short-Term Memory (LSTM) units in the hidden layer with no need of features extraction with the objective to classify the signal directly from sequences of the band-pass filtered data. The method proposed use the NinaPro database with amputee and non-amputee subjects. Only non-amputee subjects are used for parameters selection and then tested on both databases. The results show that 98% of the non-contaminated sEMG data was corrected classified and more than 95% of the contaminants were identified inside the training SNR range. Also, in this work is presented a SNR sensibility control and the contamination analysis in the range from −40 dB to 40 dB in 10 dB steps. The conclusion is that is possible to classify the contamination type in sEMG signals with a RNN-LSTM with a 112.5 ms time window and to predicted with a small error the classification hit rate for each SNR level in some cases.

Surface Electromyography and Electroencephalogram-Based Gait Phase Recognition and Correlations Between Cortical and Locomotor Muscle in the Seven Gait Phases.

The classification of gait phases based on surface electromyography (sEMG) and electroencephalogram (EEG) can be used to the control systems of lower limb exoskeletons for the rehabilitation of patients with lower limb disorders. In this study, the slope sign change (SSC) and mean power frequency (MPF) features of EEG and sEMG were used to recognize the seven gait phases [loading response (LR), mid-stance (MST), terminal stance (TST), pre-swing (PSW), initial swing (ISW), mid-swing (MSW), and terminal swing (TSW)]. Previous researchers have found that the cortex is involved in the regulation of treadmill walking. However, corticomuscular interaction analysis in a high level of gait phase granularity remains lacking in the time–frequency domain, and the feasibility of gait phase recognition based on EEG combined with sEMG is unknown. Therefore, the time–frequency cross mutual information (TFCMI) method was applied to research the theoretical basis of gait control in seven gait phases using beta-band EEG and sEMG data. We firstly found that the feature set comprising SSC of EEG as well as SSC and MPF of sEMG was robust for the recognition of seven gait phases under three different walking speeds. Secondly, the distribution of TFCMI values in eight topographies (eight muscles) was different at PSW and TSW phases. Thirdly, the differences of corticomuscular interaction between LR and MST and between TST and PSW of eight muscles were not significant. These insights enrich previous findings of the authors who have carried out gait phase recognition and provide a theoretical basis for gait recognition based on EEG and sEMG.

Hand Posture and Force Estimation Using Surface Electromyography and an Artificial Neural Network.

The purpose of this study was to develop an approach to predict hand posture (pinch versus grip) and grasp force using forearm surface electromyography (sEMG) and artificial neural networks (ANNs) during tasks that varied repetition rate and duty cycle. Prior studies have used electromyography with machine learning models to predict grip force but relatively few studies have assessed whether both hand posture and force can be predicted, particularly at varying levels of duty cycle and repetition rate. Fourteen individuals participated in this experiment. sEMG data for five forearm muscles and force output data were collected. Calibration data (25, 50, 75, 100% of maximum voluntary contraction (MVC)) were used to train ANN models to predict hand posture (pinch versus grip) and force magnitude while performing tasks that varied load, repetition rate, and duty cycle. Across all participants, overall hand posture prediction accuracy was 79% (0.79 ± .08), whereas overall hand force prediction accuracy was 73% (0.73 ± .09). Accuracy ranged between 0.65 and 0.93 based on varying repetition rate and duty cycle. Hand posture and force prediction were possible using sEMG and ANNs, though there were important differences in the accuracy of predictions based on task characteristics including duty cycle and repetition rate. The results of this study could be applied to the development of a dosimeter used for distal upper extremity biomechanical exposure measurement, risk assessment, job (re)design, and return to work programs.

A Multi-DoF Prosthetic Hand Finger Joint Controller for Wearable sEMG Sensors by Nonlinear Autoregressive Exogenous Model

The loss of mobility function and sensory information from the arm, hand, and fingertips hampers the activities of daily living (ADL) of patients. A modern bionic prosthetic hand can compensate for the lost functions and realize multiple degree of freedom (DoF) movements. However, the commercially available prosthetic hands usually have limited DoFs due to limited sensors and lack of stable classification algorithms. This study aimed to propose a controller for finger joint angle estimation by surface electromyography (sEMG). The sEMG data used for training were gathered with the Myo armband, which is a commercial EMG sensor. Two features in the time domain were extracted and fed into a nonlinear autoregressive model with exogenous inputs (NARX). The NARX model was trained with pre-selected parameters using the Levenberg–Marquardt algorithm. Comparing with the targets, the regression correlation coefficient (R) of the model outputs was more than 0.982 over all test subjects, and the mean square error was less than 10.02 for a signal range in arbitrary units equal to [0, 255]. The study also demonstrated that the proposed model could be used in daily life movements with good accuracy and generalization abilities.

Functional Gait Recovery after a Combination of Conventional Therapy and Overground Robot-Assisted Gait Training Is Not Associated with Significant Changes in Muscle Activation Pattern: An EMG Preliminary Study on Subjects Subacute Post Stroke.

Background: Overground Robot-Assisted Gait Training (o-RAGT) appears to be a promising stroke rehabilitation in terms of clinical outcomes. The literature on surface ElectroMyoGraphy (sEMG) assessment in o-RAGT is limited. This paper aimed to assess muscle activation patterns with sEMG in subjects subacute post stroke after training with o-RAGT and conventional therapy. Methods: An observational preliminary study was carried out with subjects subacute post stroke who received 15 sessions of o-RAGT (5 sessions/week; 60 min) in combination with conventional therapy. The subjects were assessed with both clinical and instrumental evaluations. Gait kinematics and sEMG data were acquired before (T1) and after (T2) the period of treatment (during ecological gait), and during the first session of o-RAGT (o-RAGT1). An eight-channel wireless sEMG device acquired in sEMG signals. Significant differences in sEMG outcomes were found in the BS of TA between T1 and T2. There were no other significant correlations between the sEMG outcomes and the clinical results between T1 and T2. Conclusions: There were significant functional gains in gait after complex intensive clinical rehabilitation with o-RAGT and conventional therapy. In addition, there was a significant increase in bilateral symmetry of the Tibialis Anterior muscles. At this stage of the signals from the tibialis anterior (TA), gastrocnemius medialis (GM), rectus femoris (RF), and biceps femoris caput longus (BF) muscles of each lower extremity. sEMG data processing extracted the Bilateral Symmetry (BS), the Co-Contraction (CC), and the Root Mean Square (RMS) coefficients. Results: Eight of 22 subjects in the subacute stage post stroke agreed to participate in this sEMG study. This subsample demonstrated a significant improvement in the motricity index of the affected lower limb and functional ambulation. The heterogeneity of the subjects’ characteristics and the small number of subjects was associated with high variability research, functional gait recovery was associated with minimal change in muscle activation patterns.

Prediction of motor Unified Parkinson's Disease Rating Scale scores in patients with Parkinson’s disease using surface electromyography

ObjectiveParkinson's disease (PD) is a chronic neurodegenerative disorder with increasing prevalence in the elderly. Especially patients with advanced PD often require complex medication regimens due to fluctuations, that is abrupt transitions from ON to OFF or vice versa. Current gold standard to quantify PD-patients’ motor symptoms is the assessment of the Unified Parkinson's Disease Rating Scale (UPDRS), which, however, is cumbersome and may depend upon investigators. This work aimed at developing a mobile, objective and unobtrusive measurement of motor symptoms in PD. MethodsData from 45 PD-patients was recorded using surface electromyography (sEMG) electrodes attached to a wristband. The motor paradigm consisted of a tapping task performed with and without dopaminergic medication. Our aim was to predict UPDRS scores from the sEMG characteristics with distinct regression models and machine learning techniques. ResultsA random forest regression model outnumbered other regression models resulting in a correlation of 0.739 between true and predicted UPDRS values. ConclusionsPD-patients’ motor affection can be extrapolated from sEMG data during a simple tapping task. In the future, such records could help determine the need for medication changes in telemedicine applications. SignificanceOur findings support the utility of wearables to detect Parkinson's symptoms and could help in developing tailored therapies in the future.

Motor Function Assessment of Upper Limb in Stroke Patients.

Background Quantitative assessment of motor function is extremely important for poststroke patients as it can be used to develop personalized treatment strategies. This study aimed to propose an evaluation method for upper limb motor function in stroke patients. Methods Thirty-four stroke survivors and twenty-five age-matched healthy volunteers as the control group were recruited for this study. Inertial sensor data and surface electromyography (sEMG) signals were collected from the upper limb during voluntary upward reaching. Five features included max shoulder joint angle, peak and average speeds, torso balance calculated from inertial sensor data, and muscle synergy similarity extracted from sEMG data by the nonnegative matrix factorization algorithm. Meanwhile, the Fugl–Meyer score of each patient was graded by professional rehabilitation therapist. Results Statistically significant differences were observed among severe, mild-to-moderate, and control group of five features (p ≤ 0.001). The features varied as the level of upper limb motor function changes since these features significantly correlated with the Fugl–Meyer assessment scale (p ≤ 0.001). Moreover, the Bland–Altman method was conducted and showed high consistency between the evaluation method of five features and Fugl–Meyer scale. Therefore, the five features proposed in this paper can quantitatively evaluate the motor function of stroke patients which is very useful in the rehabilitation process.

Decoding of Ankle Joint Movements in Stroke Patients Using Surface Electromyography.

Stroke is a cerebrovascular disease (CVD), which results in hemiplegia, paralysis, or death. Conventionally, a stroke patient requires prolonged sessions with physical therapists for the recovery of motor function. Various home-based rehabilitative devices are also available for upper limbs and require minimal or no assistance from a physiotherapist. However, there is no clinically proven device available for functional recovery of a lower limb. In this study, we explored the potential use of surface electromyography (sEMG) as a controlling mechanism for the development of a home-based lower limb rehabilitative device for stroke patients. In this experiment, three channels of sEMG were used to record data from 11 stroke patients while performing ankle joint movements. The movements were then decoded from the sEMG data and their correlation with the level of motor impairment was investigated. The impairment level was quantified using the Fugl-Meyer Assessment (FMA) scale. During the analysis, Hudgins time-domain features were extracted and classified using linear discriminant analysis (LDA) and artificial neural network (ANN). On average, 63.86% ± 4.3% and 67.1% ± 7.9% of the movements were accurately classified in an offline analysis by LDA and ANN, respectively. We found that in both classifiers, some motions outperformed others (p < 0.001 for LDA and p = 0.014 for ANN). The Spearman correlation (ρ) was calculated between the FMA scores and classification accuracies. The results indicate that there is a moderately positive correlation (ρ = 0.75 for LDA and ρ = 0.55 for ANN) between the two of them. The findings of this study suggest that a home-based EMG system can be developed to provide customized therapy for the improvement of functional lower limb motion in stroke patients.

Muscle Function and Kinematics during Submaximal Equine Jumping: What Can Objective Outcomes Tell Us about Athletic Performance Indicators?

Simple SummaryJumping represents the most popular equestrian discipline. However, traditional selection and training strategies for jumping horses have not been validated using instrumented performance analyses to scientifically inform the optimization of athlete selection, training and competitive performance. We aimed to quantify the role of muscle function during the equine jump, its relationship to athletic performance indicators, and how this objective information can inform equestrian selection and training practices. We used three-dimensional kinematic and surface electromyography data to quantify movement and muscle activation, respectively, from horses executing a submaximal jump. Horses were grouped based on their ability to raise the center of mass during the jump suspension—a fundamental, objective measure of jumping performance. Kinematic data were used to objectively measure equestrian-derived preferences for movement traits related to impulsion, engagement and joint articulation. Horses that raised the center of mass highest during jumping displayed muscle activation and movement strategies that indicate a greater ability to rapidly generate hindlimb muscular force during jump take-off. These findings provide objective support for equestrian preferences related to the generation of engagement, impulsion and hindlimb muscle power when selecting and training jumping horses and justify their prioritization as objective performance indicators for the sport of equine jumping.Selection and training practices for jumping horses have not yet been validated using objective performance analyses. This study aimed to quantify the differences and relationships between movement and muscle activation strategies in horses with varying jump technique to identify objective jumping performance indicators. Surface electromyography (sEMG) and three-dimensional kinematic data were collected from horses executing a submaximal jump. Kinematic variables were calculated based on equestrian-derived performance indicators relating to impulsion, engagement and joint articulation. Horses were grouped using an objective performance indicator—center of mass (CM) elevation during jump suspension (ZCM). Between-group differences in kinematic variables and muscle activation timings, calculated from sEMG data, were analyzed using one-way ANOVA. Statistical parametric mapping (SPM) evaluated between-group differences in time and amplitude-normalized sEMG waveforms. Relationships between movement and muscle activation were evaluated using Pearson correlation coefficients. Horses with the greatest ZCM displayed significantly (p < 0.05) shorter gluteal contractions at take-off, which were significantly correlated (p < 0.05) with a faster approach and more rapid hindlimb shortening and CM vertical displacement and velocity, as well as shorter hindlimb stance duration at take-off. Findings provide objective support for prioritizing equestrian-derived performance indicators related to the generation of engagement, impulsion and hindlimb muscle power when selecting or training jumping horses.

A novel energy-motion model for continuous sEMG decoding: from muscle energy to motor pattern

At present, sEMG-based gesture recognition requires vast amounts of training data; otherwise it is limited to a few gestures. Objective. This paper presents a novel dynamic energy model that decodes continuous hand actions by training small amounts of sEMG data. Approach. The activation of forearm muscles can set the corresponding fingers in motion or state with movement trends. The moving fingers store kinetic energy, and the fingers with movement trends store potential energy. The kinetic energy and potential energy in each finger are dynamically allocated due to the adaptive-coupling mechanism of five-fingers in actual motion. Meanwhile, the sum of the two energies remains constant at a certain muscle activation. We regarded hand movements with the same direction of acceleration for five-finger as the same in energy mode and divided hand movements into ten energy modes. Independent component analysis and machine learning methods were used to model associations between sEMG signals and energy modes and expressed gestures by energy form adaptively. This theory imitates the self-adapting mechanism in actual tasks. Thus, ten healthy subjects were recruited, and three experiments mimicking activities of daily living were designed to evaluate the interface: (1) the expression of untrained gestures, (2) the decoding of the amount of single-finger energy, and (3) real-time control. Main results. (1) Participants completed the untrained hand movements (100/100, < 0.0001). (2) The interface performed better than chance in the experiment where participants pricked balloons with a needle tip (779/1000, < 0.0001). (3) In the experiment where participants punched a hole in the plasticine on the balloon, the success rate was over 95% (97.67 ± 5.04%, < 0.01). Significance. The model can achieve continuous hand actions with speed or force information by training small amounts of sEMG data, which reduces learning task complexity.

Effect of biomechanical footwear on upper and lower leg muscle activity in comparison with knee brace and normal walking

AimTo evaluate the activity of knee stabilizing muscles while using custom-made biomechanical footwear (BF) and to compare it when walking barefoot and with a knee brace (Unloader®). MethodsSeventeen healthy working-aged (mean age: 29 years; standard deviation: 8 years) individuals participated. The knee brace was worn on the right knee and BF in both legs. Surface electromyography (sEMG) data was recorded bilaterally from vastus medialis (VM), semitendinosus (ST), tibialis anterior (TA) and lateral gastrocnemius (LG) muscles during walking, and repeated-measures ANOVA with a post-hoc t-test was used to determine differences between the different walking modalities (barefoot, brace and BF). ResultsAveraged sEMG was significantly higher when walking with BF than barefoot or knee brace in the ST muscles, in the right LG, and left TA muscle. It was significantly lower when walking with the brace compared to barefoot in the right ST and LG muscles, and left TA muscle. Analysis of the ensemble-averaged sEMG profiles showed earlier activation of TA muscles when walking with BF compared to other walking modalities. ConclusionBF produced greater activation in evaluated lower leg muscles compared to barefoot walking. Thus BF may have an exercise effect in rehabilitation and further studies about its effectiveness are warranted.

Ergonomic evaluation of the effects of forearm conditions and body postures on trapezius muscle activity during smartphone texting

Neck-shoulder and upper back musculoskeletal symptoms in smartphone users have gained increasing attention. We evaluated trapezius muscle activity and fatigue using an objective method (surface electromyography, sEMG), and discomfort using a subjective method (questionnaire), in smartphone users during the performance of different text-entry tasks. Fifteen participants were recruited to perform six text-entry tasks under different forearm conditions (floating and supported) and body postures (sitting, standing, lying, and walking). We collected upper trapezius (UT) and lower trapezius (LT) sEMG data, and recorded muscle discomfort scores after each task. We found that static postures (especially sitting) during smartphone use predisposed to lower muscle activity with higher fatigue level and discomfort scores than dynamic posture; there was a significant main effect of body posture on average sEMG amplitude (aEMG) and discomfort scores of both muscles (all p < 0.05). Moreover, using a smartphone with the forearm supported can reduce muscle activity, fatigue level, and discomfort scores; there was a significant main effect of the forearm condition on aEMG and discomfort scores of both muscles (all p < 0.05). Our study indicates that smartphone use with the forearm floating for a long time in a static posture should be avoided (especially while sitting).

Human knee abnormality detection from imbalanced sEMG data

The classification of imbalanced datasets, especially in medicine, is a major problem in data mining. Such a problem is evident in analyzing normal and abnormal subjects about knee from data collected during walking. In this work, surface electromyography (sEMG) data were collected during walking from the lower limb of 22 individuals (11 with and 11 without knee abnormality). Subjects with a knee abnormality take longer to complete the walking task than healthy subjects. Therefore, the SEMG signal length of unhealthy subjects is longer than that of healthy subjects, resulting in a problem of imbalance in the collected sEMG signal data. Thus, the development of a classification model for such datasets is challenging due to the bias towards the majority class in the data. The collected sEMG signals are challenging due to the contribution of multiple motor units at a time and their dependency on neuromuscular activity, physiological and anatomical properties of the involved muscles. Hence, automated analysis of such sEMG signals is an arduous task. A multi-step classification scheme is proposed in this research to overcome this limitation. The wavelet denoising (WD) scheme is used to denoise the collected sEMG signals, followed by the extraction of eleven time-domain features. The oversampling techniques are then used to balance the data under analysis by increasing the training minority class. The competency of the proposed scheme was assessed using various computational classifiers with 10 fold cross-validation. It was found that the oversampling techniques improve the performance of all studied classifiers when applied to the studied imbalanced sEMG data.

Inter-Subject Domain Adaptation for CNN-Based Wrist Kinematics Estimation Using sEMG.

Recently, convolutional neural network (CNN) has been widely investigated to decode human intentions using surface Electromyography (sEMG) signals. However, a pre-trained CNN model usually suffers from severe degradation when testing on a new individual, and this is mainly due to domain shift where characteristics of training and testing sEMG data differ substantially. To enhance inter-subject performances of CNN in the wrist kinematics estimation, we propose a novel regression scheme for supervised domain adaptation (SDA), based on which domain shift effects can be effectively reduced. Specifically, a two-stream CNN with shared weights is established to exploit source and target sEMG data simultaneously, such that domain-invariant features can be extracted. To tune CNN weights, both regression losses and a domain discrepancy loss are employed, where the former enable supervised learning and the latter minimizes distribution divergences between two domains. In this study, eight healthy subjects were recruited to perform wrist flexion-extension movements. Experiment results illustrated that the proposed regression SDA outperformed fine-tuning, a state-of-the-art transfer learning method, in both single-single and multiple-single scenarios of kinematics estimation. Unlike fine-tuning which suffers from catastrophic forgetting, regression SDA can maintain much better performances in original domains, which boosts the model reusability among multiple subjects.