Latent Factors Limiting the Performance of sEMG-Interfaces

Machine learning for hand pose classification from phasic and tonic EMG signals during bimanual activities in virtual reality

Myoelectric prostheses have recently shown significant promise for restoring hand function in individuals with upper limb loss or deficiencies, driven by advances in machine learning and increasingly accessible bioelectrical signal acquisition devices. Here, we first introduce and validate a novel experimental paradigm using a virtual reality headset equipped with hand-tracking capabilities to facilitate the recordings of synchronized EMG signals and hand pose estimation. Using both the phasic and tonic EMG components of data acquired through the proposed paradigm, we compare hand gesture classification pipelines based on standard signal processing features, convolutional neural networks, and covariance matrices with Riemannian geometry computed from raw or xDAWN-filtered EMG signals. We demonstrate the performance of the latter for gesture classification using EMG signals. We further hypothesize that introducing physiological knowledge in machine learning models will enhance their performances, leading to better myoelectric prosthesis control. We demonstrate the potential of this approach by using the neurophysiological integration of the "move command" to better separate the phasic and tonic components of the EMG signals, significantly improving the performance of sustained posture recognition. These results pave the way for the development of new cutting-edge machine learning techniques, likely refined by neurophysiology, that will further improve the decoding of real-time natural gestures and, ultimately, the control of myoelectric prostheses.

The assistance of BAZAR robot promotes improved upper limb motor coordination in workers performing an actual use-case manual material handling

This study aims at evaluating upper limb muscle coordination and activation in workers performing an actual use-case manual material handling (MMH). The study relies on the comparison of the workers' muscular activity while they perform the task, with and without the help of a dual-arm cobot (BAZAR). Eleven participants performed the task and the flexors and extensors muscles of the shoulder, elbow, wrist, and trunk joints were recorded using bipolar electromyography. The results showed that, when the particular MMH was carried out with BAZAR, both upper limb and trunk muscular co-activation and activation were decreased. Therefore, technologies that enable human-robot collaboration (HRC), which share a workspace with employees, relieve employees of external loads and enhance the effectiveness and calibre of task completion. Additionally, these technologies improve the worker's coordination, lessen the physical effort required to interact with the robot, and have a favourable impact on his or her physiological motor strategy. Practitioner summary: Upper limb and trunk muscle co-activation and activation is reduced when a specific manual material handling was performed with a cobot than without it. By improving coordination, reducing physical effort, and changing motor strategy, cobots could be proposed as an ergonomic intervention to lower workers' biomechanical risk in industry.

Cross-subject EMG hand gesture recognition based on dynamic domain generalization

Electromyography (EMG) signal based cross-subject gesture recognition methods reduce the influence of individual differences using transfer learning technology. These methods generally require calibration data collected from new subjects to adapt the pre-trained model to existing subjects. However, collecting calibration data is usually trivial and inconvenient for new subjects. This is currently a major obstacle to the daily use of hand gesture recognition based on EMG signals. To tackle the problem, we propose a novel dynamic domain generalization (DDG) method which is able to achieve accurate recognition on the hand gesture of new subjects without any calibration data. In order to extract more robust and adaptable features, a meta-adjuster is leveraged to generate a series of template coefficients to dynamically adjust dynamic network parameters. Specifically, two different kinds of templates are designed, in which the first one is different kinds of features, such as temporal features, spatial features, and spatial-temporal features, and the second one is different normalization layers. Meanwhile, a mix-style data augmentation method is introduced to make the meta-adjuster's training data more diversified. Experimental results on a public dataset verify that the proposed DDG outperforms the counterpart methods.

Study on the methods of feature extraction based on electromyographic signal classification

Electromyography (EMG) is a form of biological information, which is used in many fields to help people study human muscle movement, especially in the study of bionic hands. EMG signals can be used to explain the activity at a certain moment through the signal changes of human muscles, and it is a very complex signal, so processing it is very important. The process of EMG signals can be divided into acquisition, pre-processing, feature extraction, and classification. Not all signal channels are useful in EMG acquisition, and it is important to select useful signals among them. Therefore, this study proposes a feature extraction method to extract the most representative two-channel signals from the eight-channel signals. In this paper, the traditional principal component analysis method and support vector machine feature elimination are used to extract signal channels. At the same time, a new method, correlation heat map, is proposed to implement feature extraction method by using three methods, and three classification algorithms of K-nearest neighbor, random forest, and support vector machine are used to verify. The results show that the classification accuracy of the proposed method is better than that of the other two traditional methods.

Neurointerface with oscillator motifs for inhibitory effect over antagonist muscles

The effect of inhibitory management is usually underestimated in artificial control systems, using biological analogy. According to our hypothesis, the muscle hypertonus could be effectively compensated via stimulation by bio-plausible patterns. We proposed an approach for the compensatory stimulation device as implementation of previously presented architecture of the neurointerface, where (1) the neuroport is implemented as a DAC and stimulator, (2) neuroterminal is used for neurosimulation of a set of oscillator motifs on one-board computer. In the set of experiments with five volunteers, we measured the efficacy of motor neuron inhibition via the antagonist muscle or nerve stimulation registering muscle force with and without antagonist stimulation. For the agonist activation, we used both voluntary activity and electrical stimulation. In the case of stimulation of both the agonist and the antagonist muscles and nerves, we experimented with delays between muscle stimulation in the range of 0-20 ms. We registered the subjective discomfort rate. We did not identify any significant difference between the antagonist muscle and nerve stimulation in both voluntary activity and electrical stimulation of cases showing agonist activity. We determined the most effective delay between the stimulation of the agonist and the antagonist muscles and nerves as 10-20 ms.

Feature-Based Information Retrieval of Multimodal Biosignals with a Self-Similarity Matrix: Focus on Automatic Segmentation

Biosignal-based technology has been increasingly available in our daily life, being a critical information source. Wearable biosensors have been widely applied in, among others, biometrics, sports, health care, rehabilitation assistance, and edutainment. Continuous data collection from biodevices provides a valuable volume of information, which needs to be curated and prepared before serving machine learning applications. One of the universal preparation steps is data segmentation and labelling/annotation. This work proposes a practical and manageable way to automatically segment and label single-channel or multimodal biosignal data using a self-similarity matrix (SSM) computed with signals' feature-based representation. Applied to public biosignal datasets and a benchmark for change point detection, the proposed approach delivered lucid visual support in interpreting the biosignals with the SSM while performing accurate automatic segmentation of biosignals with the help of the novelty function and associating the segments grounded on their similarity measures with the similarity profiles. The proposed method performed superior to other algorithms in most cases of a series of automatic biosignal segmentation tasks; of equal appeal is that it provides an intuitive visualization for information retrieval of multimodal biosignals.

Multi-user motion recognition using sEMG via discriminative canonical correlation analysis and adaptive dimensionality reduction

The inability of new users to adapt quickly to the surface electromyography (sEMG) interface has greatly hindered the development of sEMG in the field of rehabilitation. This is due mainly to the large differences in sEMG signals produced by muscles when different people perform the same motion. To address this issue, a multi-user sEMG framework is proposed, using discriminative canonical correlation analysis and adaptive dimensionality reduction (ADR). The interface projects the feature sets for training users and new users into a low-dimensional uniform style space, overcoming the problem of individual differences in sEMG. The ADR method removes the redundant information in sEMG features and improves the accuracy of system motion recognition. The presented framework was validated on eight subjects with intact limbs, with an average recognition accuracy of 92.23% in 12 categories of upper-limb movements. In rehabilitation laboratory experiments, the average recognition rate reached 90.52%. The experimental results suggest that the framework offers a good solution to enable new rehabilitation users to adapt quickly to the sEMG interface.

Optimised EMG pipeline for gesture classification

In the expanding field of robotic prosthetics, surface electromyography (sEMG) signals can be decoded to seamlessly control a robotic prosthesis to perform the desired gesture. It is essential to create a pipeline, which can acquire, process, and accurately classify sEMG signals in order to replicate the desired hand gesture in near real-time and in a reliable manner. In this study, an optimised pipeline is proposed. This pipeline encompasses the main stages of sEMG signal processing and hand gesture classification and implements a sliding window approach, which is the main focus of the optimisation. In this study, a range of different parameters and modelling approaches are evaluated. The main contributions of this work are a robust and extensive analysis of sliding window parameter selection and an optimised pipeline that could be implemented in practice with minimal overheads. The optimum pipeline is efficient and achieves accurate prediction of hand gestures with an uninterrupted processing pipeline.

All-in-Memory Brain-Inspired Computing Using FeFET Synapses

The separation of computing units and memory in the computer architecture mandates energy-intensive data transfers creating the von Neumann bottleneck. This bottleneck is exposed at the application level by the steady growth of IoT and data-centric deep learning algorithms demanding extraordinary throughput. On the hardware level, analog Processing-in-Memory (PiM) schemes are used to build platforms that eliminate the compute-memory gap to overcome the von Neumann bottleneck. PiM can be efficiently implemented with ferroelectric transistors (FeFET), an emerging non-volatile memory technology. However, PiM and FeFET are heavily impacted by process variation, especially in sub 14 nm technology nodes, reducing the reliability and thus inducing errors. Brain-inspired Hyperdimensional Computing (HDC) is robust against such errors. Further, it is able to learn from very little data cutting energy-intensive transfers. Hence, HDC, in combination with PiM, tackles the von Neumann bottleneck at both levels. Nevertheless, the analog nature of PiM schemes necessitates the conversion of results to digital, which is often not considered. Yet, the conversion introduces large overheads and diminishes the PiM efficiency. In this paper, we propose an all-in-memory scheme performing computation and conversion at once, utilizing programmable FeFET synapses to build the comparator used for the conversion. Our experimental setup is first calibrated against Intel 14 nm FinFET technology for both transistor electrical characteristics and variability. Then, a physics-based model of ferroelectric is included to realize the Fe-FinFETs. Using this setup, we analyze the circuit’s susceptibility to process variation, derive a comprehensive error probability model, and inject it into the inference algorithm of HDC. The robustness of HDC against noise and errors is able to withstand the high error probabilities with a loss of merely 0.3% inference accuracy.

A Novel EMG-Based Hand Gesture Recognition Framework Based on Multivariate Variational Mode Decomposition

Surface electromyography (sEMG) is a kind of biological signal that records muscle activity noninvasively, which is of great significance in advanced human-computer interaction, prosthetic control, clinical therapy, and biomechanics. However, the number of hand gestures that can be recognized is limited and the recognition accuracy needs to be further improved. These factors lead to the fact that sEMG products are not widely used in practice. The main contributions of this paper are as follows. Firstly, considering the increasing number of gestures to be recognized and the complexity of gestures, an extensible two-stage machine learning lightweight framework was innovatively proposed for multi-gesture task recognition. Secondly, the multivariate variational mode decomposition (MVMD) is applied to extract the spatial–temporal features from the multiple channels to the EMG signals, and the separable convolutional neural network is used for modelling. In this work, the experimental results for 52 hand gestures recognition task show that the average accuracy on each stage is about 90%. The potential movement information is mainly contained in the low-frequency oscillator of the sEMG signal, and the model performs better with the low-frequency oscillation from the MVMD algorithm on the second stage classification than that of other decomposition methods.

Bipolar versus high-density surface electromyography for evaluating risk in fatiguing frequency-dependent lifting activities

Workers often develop low back pain due to manually lifting heavy loads. Instrumental-based assessment tools are used to quantitatively assess the biomechanical risk in lifting activities. This study aims to verify the hypothesis that high-density surface electromyography (HDsEMG) allows an optimized discrimination of risk levels associated with different fatiguing lifting conditions compared to traditional bipolar sEMG. 15 participants performed three lifting tasks with a progressively increasing lifting index (LI) each lasting 15 min. Erector spinae (ES) activity was recorded using both bipolar and HDsEMG systems. The amplitude of both bipolar and HDsEMG can significantly discriminate each pair of LI. HDsEMG data could discriminate across the different LIs starting from the fourth minute of the task while bipolar sEMG could only do so towards the end. The higher discriminative power of HDsEMG data across the lifting tasks makes such methodology a valuable tool to be used to monitor fatigue while lifting and could extend the possibilities offered by currently available instrumental-based tools.

Development of a Low-Cost, Modular Muscle-Computer Interface for At-Home Telerehabilitation for Chronic Stroke

Stroke is a leading cause of long-term disability in the United States. Recent studies have shown that high doses of repeated task-specific practice can be effective at improving upper-limb function at the chronic stage. Providing at-home telerehabilitation services with therapist supervision may allow higher dose interventions targeted to this population. Additionally, muscle biofeedback to train patients to avoid unwanted simultaneous activation of antagonist muscles (co-contractions) may be incorporated into telerehabilitation technologies to improve motor control. Here, we present the development and feasibility of a low-cost, portable, telerehabilitation biofeedback system called Tele-REINVENT. We describe our modular electromyography acquisition, processing, and feedback algorithms to train differentiated muscle control during at-home therapist-guided sessions. Additionally, we evaluated the performance of low-cost sensors for our training task with two healthy individuals. Finally, we present the results of a case study with a stroke survivor who used the system for 40 sessions over 10 weeks of training. In line with our previous research, our results suggest that using low-cost sensors provides similar results to those using research-grade sensors for low forces during an isometric task. Our preliminary case study data with one patient with stroke also suggest that our system is feasible, safe, and enjoyable to use during 10 weeks of biofeedback training, and that improvements in differentiated muscle activity during volitional movement attempt may be induced during a 10-week period. Our data provide support for using low-cost technology for individuated muscle training to reduce unintended coactivation during supervised and unsupervised home-based telerehabilitation for clinical populations, and suggest this approach is safe and feasible. Future work with larger study populations may expand on the development of meaningful and personalized chronic stroke rehabilitation.

Critical Issues and Imminent Challenges in the Use of sEMG in Return-To-Work Rehabilitation of Patients Affected by Neurological Disorders in the Epoch of Human-Robot Collaborative Technologies

Patients affected by neurological pathologies with motor disorders when they are of working age have to cope with problems related to employability, difficulties in working, and premature work interruption. It has been demonstrated that suitable job accommodation plans play a beneficial role in the overall quality of life of pathological subjects. A well-designed return-to-work program should consider several recent innovations in the clinical and ergonomic fields. One of the instrument-based methods used to monitor the effectiveness of ergonomic interventions is surface electromyography (sEMG), a multi-channel, non-invasive, wireless, wearable tool, which allows in-depth analysis of motor coordination mechanisms. Although the scientific literature in this field is extensive, its use remains significantly underexploited and the state-of-the-art technology lags expectations. This is mainly attributable to technical and methodological (electrode-skin impedance, noise, electrode location, size, configuration and distance, presence of crosstalk signals, comfort issues, selection of appropriate sensor setup, sEMG amplitude normalization, definition of correct sEMG-related outcomes and normative data) and cultural limitations. The technical and methodological problems are being resolved or minimized also thanks to the possibility of using reference books and tutorials. Cultural limitations are identified in the traditional use of qualitative approaches at the expense of quantitative measurement-based monitoring methods to design and assess ergonomic interventions and train operators. To bridge the gap between the return-to-work rehabilitation and other disciplines, several teaching courses, accompanied by further electrodes and instrumentations development, should be designed at all Bachelor, Master and PhD of Science levels to enhance the best skills available among physiotherapists, occupational health and safety technicians and ergonomists.

A Neuromuscular Interface for Robotic Devices Control

A neuromuscular interface (NI) that can be employed to operate external robotic devices (RD), including commercial ones, was proposed. Multichannel electromyographic (EMG) signal is used in the control loop. Control signal can also be supplemented with electroencephalography (EEG), limb kinematics, or other modalities. The multiple electrode approach takes advantage of the massive resources of the human brain for solving nontrivial tasks, such as movement coordination. Multilayer artificial neural network was used for feature classification and further to provide command and/or proportional control of three robotic devices. The possibility of using biofeedback can compensate for control errors and implement a fundamentally important feature that has previously limited the development of intelligent exoskeletons, prostheses, and other medical devices. The control system can be integrated with wearable electronics. Examples of technical devices under control of the neuromuscular interface (NI) are presented.