A Novel Sliding Mode Differentiator-Based Feature for EMG-Based Hand Gesture Characterization.

To enhance the reproducibility of motor unit number index (MUNIX) for evaluating neurological disease progression, this paper proposes a negative entropy-based fast independent component analysis (FastICA) demixing method to assess MUNIX reproducibility in the presence of inter-channel mixing of electromyography (EMG) signals acquired by high-density electrodes. First, composite surface EMG (sEMG) signals were obtained using high-density surface electrodes. Second, the FastICA algorithm based on negative entropy was employed to determine the orthogonal projection matrix that minimizes the negative entropy of the projected signal and effectively separates mixed sEMG signals. Finally, the proposed experimental approach was validated by introducing an interrelationship criterion to quantify independence between adjacent channel EMG signals, measuring MUNIX repeatability using coefficient of variation (CV), and determining motor unit number and size through MUNIX. Results analysis shows that the inclusion of the full (128) channel sEMG information leads to a reduction in CV value by $1.5 \pm 0.1$ and a linear decline in CV value with an increase in the number of channels. The correlation between adjacent channels in participants decreases by $0.12 \pm 0.05$ as the number of channels gradually increases. The results demonstrate a significant reduction in the number of interrelationships between sEMG signals following negative entropy-based FastICA processing, compared to the mixed sEMG signals. Moreover, this decrease in interrelationships becomes more pronounced with an increasing number of channels. Additionally, the CV of MUNIX gradually decreases with an increase in the number of channels, thereby optimizing the issue of abnormal MUNIX repeatability patterns and further enhancing the reproducibility of MUNIX based on high-density surface EMG signals.

BACKGROUND Parkinson’s disease (PD) is a progressive neurodegenerative disorder characterized by motor impairments, including tremors, rigidity, and bradykinesia. These symptoms significantly affect patients' daily activities and quality of life. Current treatment options, such as medication and deep brain stimulation (DBS), have limitations, including side effects and high costs. Therefore, there is a need for an alternative, non-invasive assistive solution to improve motor function in Parkinson’s patients. OBJECTIVE The objective of this study is to develop a brain-body interface that utilizes EEG, EMG, and FES to assist Parkinson’s patients in controlling their motor movements. By synchronizing neural and muscular signals, the system aims to facilitate voluntary movement and reduce tremors without invasive procedures. This research seeks to establish a theoretical model for signal processing and movement generation, forming the foundation for future prototype development and clinical validation. METHODS This section provides a detailed explanation of the methodology used in developing the Parkinson’s assistive device. The approach involves multiple components, each playing a crucial role in capturing, processing, and responding to neurological and muscular signals to facilitate controlled movement. i. EEG Sensor Electroencephalography (EEG) sensors are used to capture brain signals, specifically detecting neural activity associated with movement intention. The EEG data is processed using signal processing algorithms to extract relevant patterns that indicate the user's intent to move a specific muscle group. ii. EMG Sensor Electromyography (EMG) sensors detect electrical activity in muscles. These sensors help in monitoring voluntary and involuntary muscle contractions. By integrating EEG and EMG data, the system enhances the accuracy of movement prediction and stimulation. iii. Functional Electrical Stimulation (FES) FES is used to generate electrical impulses that stimulate specific muscles, facilitating movement in patients experiencing tremors or rigidity. The FES unit receives processed signals from the STM32 microcontroller, ensuring precise and controlled stimulation. iv. STM32 Microcontroller The STM32 microcontroller serves as the central processing unit, responsible for handling signals from EEG and EMG sensors, processing them using machine learning algorithms, and sending appropriate stimulation signals to the FES system. It ensures real-time synchronization between brain activity, muscle response, and electrical stimulation. v. Battery System The device is powered by a rechargeable battery system, providing a stable and efficient energy supply. Power management circuits are implemented to optimize energy consumption, ensuring long-term usability without frequent recharging. vi. Signal Processing and Data Flow  EEG signals are collected and filtered to remove noise.  EMG signals are simultaneously captured to correlate neural activity with muscle activity.  The STM32 microcontroller processes these signals and applies machine learning models to predict intended movement.  The microcontroller sends precise electrical stimulation commands to the FES unit.  The FES unit stimulates the target muscles, enabling controlled movement. vii. Synchronization and Feedback Mechanism To improve accuracy, a feedback loop is implemented where real-time responses from the muscles (EMG) are re-evaluated, and adjustments are made dynamically to the stimulation parameters. This ensures adaptive and efficient motor control. RESULTS In this study, a theoretical model was developed to integrate EEG, EBG, and EMG signals for effective control of Functional Electrical Stimulation (FES) in Parkinson’s patients. The preliminary analysis of signal synchronization and processing suggests that this approach has the potential to facilitate controlled motor movements. 1. Theoretical Validation: The signal processing framework was designed based on existing neurophysiological principles. The expected interactions between EEG and EMG signals indicate that FES can be triggered appropriately to induce movement. 2. Expected Outcomes: The anticipated result of this system is an improvement in motor function for Parkinson’s patients by translating neural intent into physical action. The model predicts that synchronized stimulation can assist in reducing tremors and enhancing voluntary movements. 3. Future Work: The next phase involves developing a prototype for real-world testing. Experimental validation through hardware implementation will be conducted to confirm the effectiveness of the proposed system. CONCLUSIONS This research presents a significant advancement in assistive technology for individuals with Parkinson’s disease. By integrating EEG, EMG, and FES sensors with an STM32 microcontroller, the system effectively interprets neural and muscular signals to generate precise stimulation, aiding in movement control. The study highlights the potential of electrical stimulation and brain-computer interface technology in enhancing motor functions without invasive procedures. The development of this device marks a step forward in neuro-assistive solutions, providing a non-invasive, adaptive, and user-friendly system for patients. The integration of AI-driven signal processing and real-time data adaptation further enhances the system’s efficiency and accuracy. The miniaturization of components and transition to wearable technology will improve usability and accessibility for daily life applications. Future improvements will focus on optimizing the device’s performance, expanding its application to other neurological disorders, and conducting extensive clinical trials to validate its effectiveness. Through continuous innovation and collaboration with healthcare professionals, this technology has the potential to revolutionize treatment approaches for movement impairments, offering a better quality of life for affected individuals. CLINICALTRIAL

The effects of the number of channels and gyroscopic data on the classification performance in EMG data acquired by Myo armband

Assistive devices support disabled people through object traction intention and prediction. To develop a smart assistive device to support paralyzed patients through the flexible nature of gloves which allow the user finger movements which has no strength or nerve-based controls. The kinematic signal variations at the upper limb position are collected through the electromyography (EMG) signals. The interaction between the prosthetic hand and the human-disabled part is achieved through inferred parameters used in the communication. The rehabilitation and assistive technologies enhance the process of user intention to perform the task under various situations. The intended object prediction model involves human interaction with prosthetic signals. The object tracking and hand movement positions are combined in the proposed model to predict the exact activities of the user. The low and high muscle variations derive the pattern and its associated task. The proposed model introduces the deep transradial amputation model (DTAM) to predict the user intention movement based on EMG sensor-based hand gesture recognition. The proposed method analyzes the EMG signals collected from the upper and lower hand muscles. The model also constructs and trains the data to predict the 3D hand gestures and their positions from the features collected through EMG signals. The reinforcement-based recurrent fuzzy neural network (RFNN) is used to derive the pattern by combining various positions of the hand gesture. The maximum reward value used to obtain the accurate prediction is a performance metric. The correlation mapping and its coefficient values provide sufficient evidence to analyze the muscle variation data to predict user-intended activities. The 3D prosthetic hand values and finger positions of the complex object task acceleration can be predicted with the mean performance of [Formula: see text] and NRMSE value = 0.101. The maximum reward count to 50 under the various iteration processes to analyze the movements. The proposed transradial amputation manages to predict the user’s intention within the time period of 124[Formula: see text]ms. Through the results, the model enhances the task intention prediction and movement position quickly compared to the other models.

This research examines how advanced machine learning algorithms can be used to classify multichannel electromyographic (EMG) signals with a high level of accuracy to assist in recognizing hand gestures. The goal is to create a robust and scalable system for gesture-based virtual control using EMG signals with potential applications in assistive technologies, rehabilitation, and human-computer interaction. Data were gathered using a MYO Thalmic bracelet containing eight EMG sensors on thirty-six subjects, and a Random Forest classifier was trained to identify seven distinct types of hand gestures (rest, fist clench, wrist flexion/extension, and radial/ulnar deviations). The machine learning pipeline included extensive preprocessing (i.e., EMG signal normalization and signal feature extraction; root mean square, waveform length, and zero crossing rate) and several hyperparameter tuning procedures to improve model performance. The Random Forest model (100 decision trees) achieved an overall classification accuracy of 98.68%, with a range of accuracies for each class (e.g., 95.2% wrist flexion and 91.8% ulnar deviation) when evaluated using cross-validation (i.e., average F1-score = 0.92, precision = 0.94, recall = .91). Overall, the study provides strong evidence for the effectiveness of ensemble learning methods at analyzing complex, multidimensional EMG signals. The high classification accuracy reported, in particular, demonstrates that the system could function for real-time recognition of hand gestures in a virtual environment. Ultimately, the initial work sets the stage for future exploration of a model that may be integrated with actuation models to control prosthetic limbs, virtual actors/avatars, and robotic devices. By demonstrating a scalable and efficient method of gesture recognition using EMG signals, these early findings enable future pathways and possibilities to design innovative, assistive solutions for digital systems that increase accessibility and interaction for users who are motor impaired or have a limited range of motion.

Currently, research on gesture recognition systems has been on the rise due to the capabilities these systems provide to the field of human–machine interaction, however, gesture recognition in prosthesis and orthesis has been carried out through the use of an extensive amount of channels and electrodes to acquire the EMG (Electromyography) signals, increasing the cost and complexity of these systems. The scientific literature shows different approaches related to gesture recognition based on the analysis of EMG signals using deep learning models, highlighting the recurrent neural networks with deep learning structures. This paper presents the implementation of a Recurrent Neural Network (RNN) model using Long-short Term Memory (LSTM) units and dense layers to develop a gesture classifier for hand prosthesis control, aiming to decrease the number of EMG channels and the overall model complexity, in order to increase its scalability for embedded systems. The proposed model requires the use of only four EMG channels to recognize five hand gestures, greatly reducing the number of electrodes compared to other approaches found in the literature. The proposed model was trained using a dataset for each gesture EMG signals, which were recorded for 20 s using a custom EMG armband. The model reached an accuracy of to 99% for the training and validation stages, and an accuracy of 87 ± 7% during real-time testing. The results obtained by the proposed model establish a general methodology for the reduction of complexity in the recognition of gestures intended for human.machine interaction for different computational devices.

In the field of lower limb exoskeletons, besides its electromechanical system design and control, attention has been paid to realizing the linkage of exoskeleton robots to humans via electroencephalography (EEG) and electromyography (EMG). However, even the state of the art performance of lower limb voluntary movement intention decoding still faces many obstacles. In the following work, focusing on the perspective of the inner mechanism, a homology characteristic of EEG and EMG for lower limb voluntary movement intention was conducted. A mathematical model of EEG and EMG was built based on its mechanism, which consists of a neural mass model (NMM), neuromuscular junction model, EMG generation model, decoding model, and musculoskeletal biomechanical model. The mechanism analysis and simulation results demonstrated that EEG and EMG signals were both excited by the same movement intention with a response time difference. To assess the efficiency of the proposed model, a synchronous acquisition system for EEG and EMG was constructed to analyze the homology and response time difference from EEG and EMG signals in the limb movement intention. An effective method of wavelet coherence was used to analyze the internal correlation between EEG and EMG signals in the same limb movement intention. To further prove the effectiveness of the hypothesis in this paper, six subjects were involved in the experiments. The experimental results demonstrated that there was a strong EEG-EMG coherence at 1 Hz around movement onset, and the phase of EEG was leading the EMG. Both the simulation and experimental results revealed that EEG and EMG are homologous, and the response time of the EEG signals are earlier than EMG signals during the limb movement intention. This work can provide a theoretical basis for the feasibility of EEG-based pre-perception and fusion perception of EEG and EMG in human movement detection.

An upper limb movement estimation from electromyography by using BP neural network

Several researchers had worked on collecting electromyography (EMG) signal from amputees and come out with dataset that could be utilized for study in EMG signal processing and classification for decoding of amputee movement intention. This paper presents the work on classification of EMG signal based on the residual limb of amputees with intuitive hand movement based on interactive exercises. Dataset is obtained from NINAPRO public database website where 11 amputee subjects performed intuitive exercise of 17 hand gestures and EMG signal is acquired from the residual arm. Eight feature extraction methods are performed to obtain the EMG feature which are Mean, Minimum, Median, Skewness, Kurtosis, Approximate Entropy, Fuzzy Entropy and Kolmogorov Complexity. Two classifiers are used for EMG classification which are k-Nearest Neighbour and Ensemble classifier. Results shows average accuracy of 87.65% with Ensemble classifier for classification of movement exercise with all features of EMG is used as input to classifier.KeywordsElectromyographyMachine learningAmputee

Hand gesture recognition based on motor unit spike trains decoded from high-density electromyography

Title: Mimicking EMG features of amputated limbs by restricting unaffected limbs Authors: M. B. Kristoffersen, A. W. Franzke, A. Murgia, C. K. van der Sluis, R. M. Bongers Presenter: Morten B. Kristoffersen Affiliation: University of Groningen, University Medical Center Groningen, Netherlands E-mail: m.b.kristoffersen@umcg.nl Abstract Machine learning techniques have been proposed for the control of upper-limb prosthetics. Electromyography (EMG) signals from able-bodied participants are often used to test new algorithms and techniques. Restricting the unaffected hand has been suggested to best mimic the EMG features of the affected limb. It remains unclear whether this results in more comparable EMG features between the two limbs. In this study we measured EMG from both the affected and unaffected limbs of 11 participants who had an amputation at the trans-radial level, while they performed seven different symmetric bi-manual movements. This was done in two conditions, namely with and without restricting the unaffected limb. We hypothesised that the EMG features of the unrestricted unaffected limb differ more from the affected limb than the EMG features of the restricted unaffected limb. Hudgins’ features (1) of the EMG signals as well as offline accuracy of the movements were calculated. Preliminary results show a small-to-medium, but close to significant, interaction effect (p=.071, ηG² = .06) of hand*restriction on wavelength suggesting that wavelength has a tendency to be higher for the unaffected limb in the unrestricted condition, while this would not be the case in the restricted condition. No effects were found for the remaining features and offline accuracy. Further analysis will need to be performed to confirm the robustness of this finding. Based on the current analysis it is suggested that in experiments with able-bodied participants, the hand should be restricted to best mimic the EMG features of people with an amputation.

This paper considers the problem of classifying human hand gestures by using electromyography (EMG) signals that are usually corrupted with noise. Noisy EMG signals result in significant degradation of classification performance and to enhance the performance, a Gaussian Smoothing Filter (GSF) is employed to remove the noise in the sensed EMG signals. The filtered signals, along with various classification schemes, are used to classify several hand gestures. The features of the GSF include: high filtering efficiency, simple implementation, and equal support in frequency and time domains, endowing the GSF with the ability to filter out the noise while partially retaining high frequency components of the original signal. The use of GSF produces smoothed EMG signals that not only enhances the classification accuracy but also reduces the computational time required to develop and test the classifiers. Experiments are conducted on EMG signals, captured from a MYO band, using multiple classification techniques and a significant improvement is observed in the classification performance when using the GSF to filter out the noise in the EMG signals. The classification performance for the EMG signals, for both unfiltered and filtered cases, is compared and the use of GSF is shown to yield significant performance enhancement. Moreover, a significant reduction in the computational time is reported when employing the GSF-based classification, demonstrating the advantages of the GSF for classifying EMG signals. Finally, a comparison is performed for classifying the EMG signals smoothed using a Median Filter (MF) versus the GSF and the superiority of the GSF is shown.

To improve the reliability and safety of myoelectric prosthetic control, many researchers tend to use multi-modal signals. The combination of electromyography (EMG) and forcemyography (FMG) has been proved to be a practical choice. However, an integrative and compact design of this hybrid sensor is lacking. This paper presents a novel modular EMG–FMG sensor; the sensing module has a novel design that consists of floating electrodes, which act as the sensing probe of both the EMG and FMG. This design improves the integration of the sensor. The whole system contains one data acquisition unit and eight identical sensor modules. Experiments were conducted to evaluate the performance of the sensor system. The results show that the EMG and FMG signals have good consistency under standard conditions; the FMG signal shows a better and more robust performance than the EMG. The average accuracy is 99.07% while using both the EMG and FMG signals for recognition of six hand gestures under standard conditions. Even with two layers of gauze isolated between the sensor and the skin, the average accuracy reaches 90.9% while using only the EMG signal; if we use both the EMG and FMG signals for classification, the average accuracy is 99.42%.

There is an increasing demand for accurate prediction of joint moments using wearable sensors for robotic exoskeletons to achieve precise control and for rehabilitation care to remotely monitor users’ condition. In this study, we used electromyography (EMG) signals to first identify muscle synergies, then used them to train of a long short-term memory network to predict knee joint moments during walking. Kinematics, ground reaction forces and EMG from 10 muscles on the right limb were collected from 6 able-bodied subjects during normal gait. Between 4 and 6 muscle synergies were extracted from the EMG signals, generating two outputs - the muscle synergies weight matrix and the time-dependent muscle synergies action signals. The muscle synergies action signals and measured knee joint moments from inverse dynamics were then used as inputs to train the joint moment prediction model using a long short-term memory network. For testing, between 4 and 7 EMG signals were used to estimate the muscle synergies action signals with the extracted muscle synergies weights matrix. The estimated muscle synergies action signals were then used to predict knee joint moments. Knee joint moments were also predicted directly from all 10 EMGs, then from 4-7 EMG signals using another long short-term memory network. Prediction accuracy from the synergies-trained network vs. the EMG-trained network were compared, using the same number of EMG signals in each. Prediction error with respect to moments measured via inverse dynamics was computed for both networks. Knee moments predicted with as few as 4 EMGs was at least as accurate as moments predicted from all 10 EMGs when muscle synergies were exploited. Predicted knee moments from muscle synergies achieved an average of 4.63% root mean square error from 4 EMG signals, which was lower than error when predicted directly from 4 EMG signals (5.63%).

Electromyographic (EMG) signals have been widely used as motion commands of prosthetic arms. Although EMG signals contain meaningful information including the movement intentions of human body, it is difficult to predict the subject’s motion by analyzing EMG signals in real-time due to the difficulties in extracting motion information from the signals including a lot of noises inherently. In this paper, four Ag/AgCl electrodes are placed on the surface of the subject’s major muscles which are in charge of four upper arm movements (wrist flexion, wrist extension, ulnar deviation, finger flexion) to measure EMG signals corresponding to the movements. The measured signals are sampled using DAQ module and clustered sequentially. The Fuzzy C-Means (FCMs) method calculates the center values of the clustered data group. The fuzzy system designed to detect the upper arm movement intention utilizing the center values as input signals shows about 90% success in classifying the movement intentions.