Myoelectric Pattern Recognition for Controlling a Robotic Hand: A Feasibility Study in Stroke

Systematic review of AI/ML applications in multi-domain robotic rehabilitation: trends, gaps, and future directions

Robotic technology is expected to transform rehabilitation settings, by providing precise, repetitive, and task-specific interventions, thereby potentially improving patients' clinical outcomes. Artificial intelligence (AI) and machine learning (ML) have been widely applied in different areas to support robotic rehabilitation, from controlling robot movements to real-time patient assessment. To provide an overview of the current landscape and the impact of AI/ML use in robotics rehabilitation, we performed a systematic review focusing on the use of AI and robotics in rehabilitation from a broad perspective, encompassing different pathologies and body districts, and considering both motor and neurocognitive rehabilitation. We searched the Scopus and IEEE Xplore databases, focusing on the studies involving human participants. After article retrieval, a tagging phase was carried out to devise a comprehensive and easily-interpretable taxonomy: its categories include the aim of the AI/ML within the rehabilitation system, the type of algorithms used, and the location of robots and sensors. The 201 selected articles span multiple domains and diverse aims, such as movement classification, trajectory prediction, and patient evaluation, demonstrating the potential of ML to revolutionize personalized therapy and improve patient engagement. ML is reported as highly effective in predicting movement intentions, assessing clinical outcomes, and detecting compensatory movements, providing insights into the future of personalized rehabilitation interventions. Our analysis also reveals pitfalls in the current use of AI/ML in this area, such as potential explainability issues and poor generalization ability when these systems are applied in real-world settings.

Deep Learning Based Post-stroke Myoelectric Gesture Recognition: From Feature Construction to Network Design

Recently, robot-assisted rehabilitation has emerged as a promising solution to increase the training intensity of stroke patients while reducing workload on therapists, whilst surface electromyography (sEMG) is expected to serve as a viable control source. In this paper, we delve into the potential of deep learning (DL) for post-stroke hand gesture recognition by collecting the sEMG signals of eight chronic stroke subjects, focusing on three primary aspects: feature domains of sEMG (time, frequency, and wavelet), data structures (one or two-dimensional images), and neural network architectures (CNN, CNN-LSTM, and CNN-LSTM-Attention). A total of 18 DL models were comprehensively evaluated in both intra-subject testing and inter-subject transfer learning tasks, with two post-processing algorithms (Model Voting and Bayesian Fusion) analysed subsequently. Experiment results infer that for intra-subject testing, the average accuracy of CNN-LSTM using two-dimensional frequency features is the highest, reaching 72.95%. For inter-subject transfer learning, the average accuracy of CNN-LSTM-Attention using one-dimensional frequency features is the highest, reaching 68.38%. Through these two experiments, it was found that frequency features had significant advantages over other features in gesture recognition after stroke. Moreover, the post-processing algorithm can further improve the recognition accuracy, and the recognition effect can be increased by 2.03% through the model voting algorithm.

AI Applications in Adult Stroke Recovery and Rehabilitation: A Scoping Review Using AI

Stroke is a leading cause of long-term disability worldwide. With the advancements in sensor technologies and data availability, artificial intelligence (AI) holds the promise of improving the amount, quality and efficiency of care and enhancing the precision of stroke rehabilitation. We aimed to identify and characterize the existing research on AI applications in stroke recovery and rehabilitation of adults, including categories of application and progression of technologies over time. Data were collected from peer-reviewed articles across various electronic databases up to January 2024. Insights were extracted using AI-enhanced multi-method, data-driven techniques, including clustering of themes and topics. This scoping review summarizes outcomes from 704 studies. Four common themes (impairment, assisted intervention, prediction and imaging, and neuroscience) were identified, in which time-linked patterns emerged. The impairment theme revealed a focus on motor function, gait and mobility, while the assisted intervention theme included applications of robotic and brain-computer interface (BCI) techniques. AI applications progressed over time, starting from conceptualization and then expanding to a broader range of techniques in supervised learning, artificial neural networks (ANN), natural language processing (NLP) and more. Applications focused on upper limb rehabilitation were reviewed in more detail, with machine learning (ML), deep learning techniques and sensors such as inertial measurement units (IMU) used for upper limb and functional movement analysis. AI applications have potential to facilitate tailored therapeutic delivery, thereby contributing to the optimization of rehabilitation outcomes and promoting sustained recovery from rehabilitation to real-world settings.

Feature evaluation for myoelectric pattern recognition of multiple nearby reaching targets

Intention detection of the reaching movement is considerable for myoelectric human and machine collaboration applications. A comprehensive set of handcrafted features was mined from windows of electromyogram (EMG) of the upper-limb muscles while reaching nine nearby targets like activities of daily living. The feature selection-based scoring method, neighborhood component analysis (NCA), selected the relevant feature subset. Finally, the target was recognized by the support vector machine (SVM) model. The classification performance was generalized by a nested cross-validation structure that selected the optimal feature subset in the inner loop. According to the low spatial resolution of the target location on display and following the slight discrimination of signals between targets, the best classification accuracy of 77.11 % was achieved for concatenating the features of two segments with a length of 2 and 0.25 s. Due to the lack of subtle variation in EMG, while reaching different targets, a wide range of features was applied to consider additional aspects of the knowledge contained in EMG signals. Furthermore, since NCA selected features that provided more discriminant power, it became achievable to employ various combinations of features and even concatenated features extracted from different movement parts to improve classification performance.

FPGA-based Implementation of Random Forest Classifier for sEMG Signal Classification

This paper proposes a method that combines a hardware description language and Random Forests theory, to develop a motion recognition embedded system. To validate this system, the Ninapro database DB2 was used as training and test data, resulting in an accuracy of 83.1% for the test base in movement prediction.

A Comparison of Myoelectric Control Modes for an Assistive Robotic Virtual Platform

In this paper, we propose a daily living situation where objects in a kitchen can be grasped and stored in specific containers using a virtual robot arm operated by different myoelectric control modes. The main goal of this study is to prove the feasibility of providing virtual environments controlled through surface electromyography that can be used for the future training of people using prosthetics or with upper limb motor impairments. We propose that simple control algorithms can be a more natural and robust way to interact with prostheses and assistive robotics in general than complex multipurpose machine learning approaches. Additionally, we discuss the advantages and disadvantages of adding intelligence to the setup to automatically assist grasping activities. The results show very good performance across all participants who share similar opinions regarding the execution of each of the proposed control modes.

Enhanced Deep Transfer Learning Model based on Spatial-Temporal driven Scalograms for Precise Decoding of Motor Intent in Stroke Survivors

Intelligent rehabilitation robotics (RR) have been proposed in recent years to aid post-stroke survivors recover their lost limb functions. However, a large proportion of these robotic systems operate in a passive mode that restricts users to predefined trajectories that rarely align with their intended limb movements, precluding full functional recovery. To address this issue, an efficient Transfer Learning based Convolutional Neural Network (TL-CNN) model is proposed to decode post-stroke patients' motion intentions toward realizing dexterously active robotic training during rehabilitation. For the first time, we use Spatial-Temporal Descriptor based Continuous Wavelet Transform (STD-CWT) as input to TL-CNN to optimally decode limb movement intent patterns. We evaluated the STD-CWT method on three distinct wavelets including the Morse, Amor, and Bump, and compared their decoding outcomes with those of the commonly adopted CWT technique under similar experimental conditions. We then validated the method using electromyogram signals of five stroke survivors who performed twenty-one distinct motor tasks. The results showed that the proposed technique recorded a significantly higher (p<0.05) decoding accuracy and faster convergence compared to the common method. Our method equally recorded obvious class separability for individual motor tasks across subjects. The findings suggest that the STD-CWT Scalograms have the potential for robust decoding of motor intention and could facilitate intuitive and active motor training in stroke RR.Clinical Relevance- The study demonstrated the potential of Spatial Temporal based Scalograms in aiding precise and robust decoding of multi-class motor tasks, upon which dexterously active rehabilitation robotic training for full motor function restoration could be realized.

Hand Grasp Motion Intention Recognition Based on High-Density Electromyography in Chronic Stroke Patients

Stroke is a debilitating condition that leads to a loss of motor function, inability to perform daily life activities, and ultimately worsening quality of life. Robot-based rehabilitation is a more effective method than conventional rehabilitation but needs to accurately recognize the patient's intention so that the robot can assist the patient's voluntary motion. This study focuses on recognizing hand grasp motion intention using high-density electromyography (HD-EMG) in patients with chronic stroke. The study was conducted with three chronic stroke patients and involved recording HD-EMG signals from the muscles involved in hand grasp motions. The adaptive onset detection algorithm was used to accurately identify the start of hand grasp motions accurately, and a convolutional neural network (CNN) was trained to classify the HD-EMG signals into one of four grasping motions. The average true positive and false positive rates of the grasp onset detection on three subjects were 91.6% and 9.8%, respectively, and the trained CNN classified the grasping motion with an average accuracy of 76.3%. The results showed that using HD-EMG can provide accurate hand grasp motion intention recognition in chronic stroke patients, highlighting the potential for effective robot-based rehabilitation.

Surface EMG analysis of weakness distribution in upper limb muscles post-stroke

Weakness is a common symptom after a stroke. This study aims to depict weakness distribution among forearm muscles given that joints in the upper limb are usually driven by a group of muscles. Multi-channel electromyography (EMG) was applied to assess the muscle group, and an EMG-based index was proposed to quantify the weakness of individual muscles. By using this method, four weakness distribution patterns were observed in extensor muscles from five out of eight subjects after stroke. Complex weakness distribution patterns were observed in flexor muscles from seven out of the eight subjects when they performed grasp, tripod pinch, and hook grip. The findings can help determine the weak muscles in a clinic and facilitate the development of appropriate interventions in stroke rehabilitation targeting specific muscle weakness.

Similarity Function for One-Shot Learning to Enhance the Flexibility of Myoelectric Interfaces

$\textit {Objective:}$ This study aims to develop a flexible myoelectric pattern recognition (MPR) method based on one-shot learning, which enables convenient switching across different usage scenarios, thereby reducing the re-training burden. $\textit {Methods}$ : First, a one-shot learning model based on a Siamese neural network was constructed to assess the similarity for any given sample pair. In a new scenario involving a new set of gestural categories and/or a new user, just one sample of each category was required to constitute a support set. This enabled the quick deployment of the classifier suitable for the new scenario, which decided for any unknown query sample by selecting the category whose sample in the support set was quantified to be the most like the query sample. The effectiveness of the proposed method was evaluated by experiments conducting MPR across diverse scenarios. Results: The proposed method achieved high recognition accuracy of over 89% under the cross-scenario conditions, and it significantly outperformed other common one-shot learning methods and conventional MPR methods ( ${p} < 0.01$ ). $\textit {Conclusion}$ : This study demonstrates the feasibility of applying one-shot learning to rapidly deploy myoelectric pattern classifiers in response to scenario change. It provides a valuable way of improving the flexibility of myoelectric interfaces toward intelligent gestural control with extensive applications in medical, industrial, and consumer electronics.

Control strategies used in lower limb exoskeletons for gait rehabilitation after brain injury: a systematic review and analysis of clinical effectiveness

BACKGROUND

In the past decade, there has been substantial progress in the development of robotic controllers that specify how lower-limb exoskeletons should interact with brain-injured patients. However, it is still an open question which exoskeleton control strategies can more effectively stimulate motor function recovery. In this review, we aim to complement previous literature surveys on the topic of exoskeleton control for gait rehabilitation by: (1) providing an updated structured framework of current control strategies, (2) analyzing the methodology of clinical validations used in the robotic interventions, and (3) reporting the potential relation between control strategies and clinical outcomes.

METHODS

Four databases were searched using database-specific search terms from January 2000 to September 2020. We identified 1648 articles, of which 159 were included and evaluated in full-text. We included studies that clinically evaluated the effectiveness of the exoskeleton on impaired participants, and which clearly explained or referenced the implemented control strategy.

RESULTS

(1) We found that assistive control (100% of exoskeletons) that followed rule-based algorithms (72%) based on ground reaction force thresholds (63%) in conjunction with trajectory-tracking control (97%) were the most implemented control strategies. Only 14% of the exoskeletons implemented adaptive control strategies. (2) Regarding the clinical validations used in the robotic interventions, we found high variability on the experimental protocols and outcome metrics selected. (3) With high grade of evidence and a moderate number of participants (N = 19), assistive control strategies that implemented a combination of trajectory-tracking and compliant control showed the highest clinical effectiveness for acute stroke. However, they also required the longest training time. With high grade of evidence and low number of participants (N = 8), assistive control strategies that followed a threshold-based algorithm with EMG as gait detection metric and control signal provided the highest improvements with the lowest training intensities for subacute stroke. Finally, with high grade of evidence and a moderate number of participants (N = 19), assistive control strategies that implemented adaptive oscillator algorithms together with trajectory-tracking control resulted in the highest improvements with reduced training intensities for individuals with chronic stroke.

CONCLUSIONS

Despite the efforts to develop novel and more effective controllers for exoskeleton-based gait neurorehabilitation, the current level of evidence on the effectiveness of the different control strategies on clinical outcomes is still low. There is a clear lack of standardization in the experimental protocols leading to high levels of heterogeneity. Standardized comparisons among control strategies analyzing the relation between control parameters and biomechanical metrics will fill this gap to better guide future technical developments. It is still an open question whether controllers that provide an on-line adaptation of the control parameters based on key biomechanical descriptors associated to the patients' specific pathology outperform current control strategies.

Interaction with a Hand Rehabilitation Exoskeleton in EMG-Driven Bilateral Therapy: Influence of Visual Biofeedback on the Users' Performance

The effectiveness of EMG biofeedback with neurorehabilitation robotic platforms has not been previously addressed. The present work evaluates the influence of an EMG-based visual biofeedback on the user performance when performing EMG-driven bilateral exercises with a robotic hand exoskeleton. Eighteen healthy subjects were asked to perform 1-min randomly generated sequences of hand gestures (rest, open and close) in four different conditions resulting from the combination of using or not (1) EMG-based visual biofeedback and (2) kinesthetic feedback from the exoskeleton movement. The user performance in each test was measured by computing similarity between the target gestures and the recognized user gestures using the L2 distance. Statistically significant differences in the subject performance were found in the type of provided feedback (p-value 0.0124). Pairwise comparisons showed that the L2 distance was statistically significantly lower when only EMG-based visual feedback was present (2.89 ± 0.71) than with the presence of the kinesthetic feedback alone (3.43 ± 0.75, p-value = 0.0412) or the combination of both (3.39 ± 0.70, p-value = 0.0497). Hence, EMG-based visual feedback enables subjects to increase their control over the movement of the robotic platform by assessing their muscle activation in real time. This type of feedback could benefit patients in learning more quickly how to activate robot functions, increasing their motivation towards rehabilitation.

Reducing the number of EMG electrodes during online hand gesture classification with changing wrist positions

Background

Myoelectric control based on hand gesture classification can be used for effective, contactless human–machine interfacing in general applications (e.g., consumer market) as well as in the clinical context. However, the accuracy of hand gesture classification can be impacted by several factors including changing wrist position. The present study aimed at investigating how channel configuration (number and placement of electrode pads) affects performance in hand gesture recognition across wrist positions, with the overall goal of reducing the number of channels without the loss of performance with respect to the benchmark (all channels).

Methods

Matrix electrodes (256 channels) were used to record high-density EMG from the forearm of 13 healthy subjects performing a set of 8 gestures in 3 wrist positions and 2 force levels (low and moderate). A reduced set of channels was chosen by applying sequential forward selection (SFS) and simple circumferential placement (CIRC) and used for gesture classification with linear discriminant analysis. The classification success rate and task completion rate were the main outcome measures for offline analysis across the different number of channels and online control using 8 selected channels, respectively.

Results

The offline analysis demonstrated that good accuracy (> 90%) can be achieved with only a few channels. However, using data from all wrist positions required more channels to reach the same performance. Despite the targeted placement (SFS) performing similarly to CIRC in the offline analysis, the task completion rate [median (lower–upper quartile)] in the online control was significantly higher for SFS [71.4% (64.8–76.2%)] compared to CIRC [57.1% (51.8–64.8%), p < 0.01], especially for low contraction levels [76.2% (66.7–84.5%) for SFS vs. 57.1% (47.6–60.7%) for CIRC, p < 0.01]. For the reduced number of electrodes, the performance with SFS was comparable to that obtained when using the full matrix, while the selected electrodes were highly subject-specific.

Conclusions

The present study demonstrated that the number of channels required for gesture classification with changing wrist positions could be decreased substantially without loss of performance, if those channels are placed strategically along the forearm and individually for each subject. The results also emphasize the importance of online assessment and motivate the development of configurable matrix electrodes with integrated channel selection.

EEG-EMG hybrid real-time classification of hand grasp and release movements intention in chronic stroke patients

Rehabilitation of the hand motor function is essential for stroke patients to resume activities of daily living. Recent studies have shown that wearable robot systems, like a multi degree-of-freedom soft glove, have the potential to improve hand motor impairment. The rehabilitation system, which is intuitively controlled according to the user's intention, is expected to induce active participation of the user and further promote brain plasticity. However, due to the patient-specific nature of stroke patients, extracting the intention from stroke patients is still challenging. In this study, we implemented a classifier that combines EEG and EMG to detect chronic stroke patients' four types of intention: rest, grasp, hold, and release. Three chronic stroke patients participated in the experiment and performed rest, grasp, hold, and release actions. The rest vs. grasp binary classifier and release vs. hold binary classifier showed 76.9% and 86.6% classification accuracy in real-time, respectively. In addition, patient-specific accuracy comparisons showed that the hybrid approach was robust to upper limb impairment level compared to other approaches. We believe that these results could pave the way for the development of BCI-based robotic hand rehabilitation therapy.

Individuals with moderate to severe hand impairments may struggle to use EMG control for assistive devices

Neurological trauma, such as stroke, traumatic brain injury (TBI), spinal cord injury, and cerebral palsy can cause mild to severe upper limb impairments. Hand impairment makes it difficult for individuals to complete activities of daily living, especially bimanual tasks. A robotic hand orthosis or hand exoskeleton can be used to restore partial function of an intact but impaired hand. It is common for upper extremity prostheses and orthoses to use electromyography (EMG) sensing as a method for the user to control their device. However some individuals with an intact but impaired hand may struggle to use a myoelectrically controlled device due to potentially confounding muscle activity. This study was conducted to evaluate the application of conventional EMG control techniques as a robotic orthosis/exoskeleton user input method for individuals with mild to severe hand impairments. Nine impaired subjects and ten healthy subjects were asked to perform repeated contractions of muscles in their forearm and then onset analysis and feature classification were used to determine the accuracy of the employed EMG techniques. The average accuracy for contraction identification across employed EMG techniques was 95.4% ± 4.9 for the healthy subjects and 73.9% ± 13.1 for the impaired subjects with a range of 47.0% ± 19.1 - 91.6% ± 8.5. These preliminary results suggest that the conventional EMG control technologies employed in this paper may be difficult for some impaired individuals to use due to their unreliable muscle control.

Finger Movement Recognition via High-Density Electromyography of Intrinsic and Extrinsic Hand Muscles

Surface electromyography (sEMG) is commonly used to observe the motor neuronal activity within muscle fibers. However, decoding dexterous body movements from sEMG signals is still quite challenging. In this paper, we present a high-density sEMG (HD-sEMG) signal database that comprises simultaneously recorded sEMG signals of intrinsic and extrinsic hand muscles. Specifically, twenty able-bodied participants performed 12 finger movements under two paces and three arm postures. HD-sEMG signals were recorded with a 64-channel high-density grid placed on the back of hand and an 8-channel armband around the forearm. Also, a data-glove was used to record the finger joint angles. Synchronisation and reproducibility of the data collection from the HD-sEMG and glove sensors were ensured. The collected data samples were further employed for automated recognition of dexterous finger movements. The introduced dataset offers a new perspective to study the synergy between the intrinsic and extrinsic hand muscles during dynamic finger movements. As this dataset was collected from multiple participants, it also provides a resource for exploring generalized models for finger movement decoding.

Exploring Biomimetic Stiffness Modulation and Wearable Finger Haptics for Improving Myoelectric Control of Virtual Hand

The embodiment of virtual hand (VH) by the user is generally deemed to be important for virtual reality (VR) based hand rehabilitation applications, which may help to engage the user and promote motor skill relearning. In particular, it requires that the VH should produce task-dependent interaction behaviors from rigid to soft. While such a capability is inherent to humans via hand stiffness regulation and haptic interactions, yet it have not been successfully imitated by VH in existing studies. In this paper, we present a work which integrates biomimetic stiffness regulation and wearable finger force feedback in VR scenarios involving myoelectric control of VH. On one hand, the biomimetic stiffness modulation intuitively enables VH to imitate the stiffness profile of the user's hand in real time. On the other hand, the wearable finger force-feedback device elicits a natural and realistic sensation of external force on the fingertip, which provides the user a proper understanding of the environment for enhancing his/her stiffness regulation. The benefits of the proposed integrated system were evaluated with eight healthy subjects that performed two tasks with opposite stiffness requirements. The achieved performance is compared with reduced versions of the integrated system, where either biomimetic impedance control or wearable force feedback is excluded. The results suggest that the proposed integrated system enables the stiffness of VH to be adaptively regulated by the user through the perception of interaction torques and vision, resulting in task-dependent behaviors from rigid to soft for VH.

Design and Characterization of a Rolling-Contact Involute Joint and Its Applications in Finger Exoskeletons

The hand exoskeleton has been widely studied in the fields of hand rehabilitation and grasping assistance tasks. Current hand exoskeletons face challenges in combining a user-friendly design with a lightweight structure and accurate modeling of hand motion. In this study, we developed a finger exoskeleton with a rolling contact involute joint. Specific implementation methods were investigated, including an analysis of the mechanical characteristics of the involute joint model, the formula derivation of the joint parameter optimization algorithm, and the design process for a finger exoskeleton with an involute joint. Experiments were conducted using a finger exoskeleton prototype to evaluate the output trajectory and grasping force of the finger exoskeleton. An EMG-controlled hand exoskeleton was developed to verify the wearability and functionality of the glove. The experimental results show that the proposed involute joint can provide sufficient fingertip force (10N) while forming a lightweight exoskeleton to assist users with functional hand rehabilitation and grasping activities.

Intention Detection Strategies for Robotic Upper-Limb Orthoses: A Scoping Review Considering Usability, Daily Life Application, and User Evaluation

Wearable robotic upper limb orthoses (ULO) are promising tools to assist or enhance the upper-limb function of their users. While the functionality of these devices has continuously increased, the robust and reliable detection of the user's intention to control the available degrees of freedom remains a major challenge and a barrier for acceptance. As the information interface between device and user, the intention detection strategy (IDS) has a crucial impact on the usability of the overall device. Yet, this aspect and the impact it has on the device usability is only rarely evaluated with respect to the context of use of ULO. A scoping literature review was conducted to identify non-invasive IDS applied to ULO that have been evaluated with human participants, with a specific focus on evaluation methods and findings related to functionality and usability and their appropriateness for specific contexts of use in daily life. A total of 93 studies were identified, describing 29 different IDS that are summarized and classified according to a four-level classification scheme. The predominant user input signal associated with the described IDS was electromyography (35.6%), followed by manual triggers such as buttons, touchscreens or joysticks (16.7%), as well as isometric force generated by residual movement in upper-limb segments (15.1%). We identify and discuss the strengths and weaknesses of IDS with respect to specific contexts of use and highlight a trade-off between performance and complexity in selecting an optimal IDS. Investigating evaluation practices to study the usability of IDS, the included studies revealed that, primarily, objective and quantitative usability attributes related to effectiveness or efficiency were assessed. Further, it underlined the lack of a systematic way to determine whether the usability of an IDS is sufficiently high to be appropriate for use in daily life applications. This work highlights the importance of a user- and application-specific selection and evaluation of non-invasive IDS for ULO. For technology developers in the field, it further provides recommendations on the selection process of IDS as well as to the design of corresponding evaluation protocols.

Does the Score on the MRC Strength Scale Reflect Instrumented Measures of Maximal Torque and Muscle Activity in Post-Stroke Survivors?

It remains unknown whether variation of scores on the Medical Research Council (MRC) scale for muscle strength is associated with operator-independent techniques: dynamometry and surface electromyography (sEMG). This study aimed to evaluate whether the scores of the MRC strength scale are associated with instrumented measures of torque and muscle activity in post-stroke survivors with severe hemiparesis both before and after an intervention. Patients affected by a first ischemic or hemorrhagic stroke within 6 months before enrollment and with complete paresis were included in the study. The pre- and post-treatment assessments included the MRC strength scale, sEMG, and dynamometry assessment of the triceps brachii (TB) and biceps brachii (BB) as measures of maximal elbow extension and flexion torque, respectively. Proprioceptive-based training was used as a treatment model, which consisted of multidirectional exercises with verbal feedback. Each treatment session lasted 1 h/day, 5 days a week for a total 15 sessions. Nineteen individuals with stroke participated in the study. A significant correlation between outcome measures for the BB (MRC and sEMG p = 0.0177, ρ = 0.601; MRC and torque p = 0.0001, ρ = 0.867) and TB (MRC and sEMG p = 0.0026, ρ = 0.717; MRC and torque p = 0.0001, ρ = 0.873) were observed post intervention. Regression models revealed a relationship between the MRC score and sEMG and torque measures for both the TB and BB. The results confirmed that variation on the MRC strength scale is associated with variation in sEMG and torque measures, especially post intervention. The regression model showed a causal relationship between MRC scale scores, sEMG, and torque assessments.

sEMG-Based Hand Posture Recognition Considering Electrode Shift, Feature Vectors, and Posture Groups

Surface electromyography (sEMG)-based gesture recognition systems provide the intuitive and accurate recognition of various gestures in human-computer interaction. In this study, an sEMG-based hand posture recognition algorithm was developed, considering three main problems: electrode shift, feature vectors, and posture groups. The sEMG signal was measured using an armband sensor with the electrode shift. An artificial neural network classifier was trained using 21 feature vectors for seven different posture groups. The inter-session and inter-feature Pearson correlation coefficients (PCCs) were calculated. The results indicate that the classification performance improved with the number of training sessions of the electrode shift. The number of sessions necessary for efficient training was four, and the feature vectors with a high inter-session PCC (r > 0.7) exhibited high classification accuracy. Similarities between postures in a posture group decreased the classification accuracy. Our results indicate that the classification accuracy could be improved with the addition of more electrode shift training sessions and that the PCC is useful for selecting the feature vector. Furthermore, hand posture selection was as important as feature vector selection. These findings will help in optimizing the sEMG-based pattern recognition algorithm more easily and quickly.

Actor-Critic Reinforcement Learning Based Algorithm for Contaminant Type Identification in Surface Electromyography Data. ANNUAL INTERNATIONAL CONFERENCE OF THE IEEE ENGINEERING IN MEDICINE AND BIOLOGY SOCIETY. IEEE ENGINEERING IN MEDICINE AND BIOLOGY SOCIETY. ANNUAL INTERNATIONAL CONFERENCE 2021;2021:186-189. [PMID: 34891268 DOI: 10.1109/embc46164.2021.9629967] [Citation(s) in RCA: 0] [Impact Index Per Article: 0]

This paper aims to present an innovative approach based on Reinforcement Learning (RL) concept to detect contaminants' type and minimize their effect on surface electromyography signal (sEMG). An agent-environment model was created based on the following elements: environment (muscle electrical activity), state (set of six features extracted from the signal), actions (application of filters/procedures to reduce the impact of each interference), and agent (controller, which will identify the type of contamination and take the appropriate action). The learning was conducted with Actor-Critic method. An average accuracy of 92.96% was achieved in an off-line experiment when detecting four contaminant types (electrocardiography (ECG) interference, movement artifact, power line interference, and additive white Gaussian noise).

Optimizing Input for Gesture Recognition using Convolutional Networks on HD-sEMG Instantaneous Images

Hand gesture recognition using high-density surface electromyography (HD-sEMG) has gained increasing attention recently due its advantages of high spatio-temporal resolution. Convolutional neural networks (CNN) have also recently been implemented to learn the spatio-temporal features from the instantaneous samples of HD-sEMG signals. While the CNN itself learns the features from the input signal it has not been considered whether certain pre-processing techniques can further improve the classification accuracies established by previous studies. Therefore, common pre-processing techniques were applied to a benchmark HD-sEMG dataset (CapgMyo DB-a) and their validation accuracies were compared. Monopolar, bipolar, rectified, common-average referenced, and Laplacian spatial filtered configurations of the HD-sEMG signals were evaluated. Results showed that the baseline monopolar HD-sEMG signals maintained higher prediction accuracies versus the other signal configurations. The results of this study discourage the use of extra pre-processing steps when using convolutional networks to classify the instantaneous samples of HD-sEMG for gesture recognition.

Pattern recognition of EMG signals for low level grip force classification

Grasping of the objects is the most frequent activity performed by the human upper limb. The amputations of the upper limb results in the need for prosthetic devices. The myoelectric prosthetic devices use muscle signals and apply control techniques for identification of different levels of hand gesture and force levels. In this study; a different level force contraction experiment was performed in which Electromyography (EMGs) signals and fingertip force signals were acquired. Using this experimental data; a two-step feature selection process is applied for the designing of a pattern recognition algorithm for the classification of different force levels. The two step feature selection process consist of generalized feature ranking using ReliefF, followed by personalized feature selection using Neighborhood Component Analysis (NCA) from the shortlisted features by earlier technique. The classification algorithms applied in this study were Support Vector Machines (SVM) and Random Forest (RF). Besides feature selection; optimization of the number of muscles during classification of force levels was also performed using designed algorithm. Based on this algorithm; the maximum classification accuracy using SVM classifier and two muscle set was achieved as high as 99%. The optimal feature set consisted features such as Auto Regressive coefficients, Willison Amplitude and Slope Sign Change. The mean classification accuracy for different subjects, achieved using SVM and RF was 94.5% and 91.7% respectively.

English Feature Recognition Based on GA-BP Neural Network Algorithm and Data Mining

With the development of society and the promotion of science and technology, English, as the largest universal language in the world, is used by more and more people. In the life around us, there is information in English all the time. However, because the process of manual recognition of English letters is very labor-intensive and inefficient, the demand for computer recognition of English letters is increasing. This paper studies the influence of the parameters of BP neural network and genetic algorithm on the whole network, including the input, output, and number of hidden layer nodes. Finally, it improves and determines the settings and values of the relevant parameters. On this basis, it shows the rationality of the selected parameters through experiments. The results show that only GA-BP neural network and feature data mining algorithm can complete feature extraction and become the main function of feature classification at the same time. After enough initial data sample analysis training, the GA-BP neural network was found to have good data fault tolerance and feature recognition. The experimental results show that the genetic algorithm can find the best weights and thresholds and the weights and thresholds are given to the BP neural network. After training, the recognition of handwritten letters can be realized. Finally, the convergence of the two algorithms is compared through experiments, which shows that the overall performance of the BP neural network algorithm is improved after genetic algorithm optimization. It can be seen that the genetic algorithm has a good effect in improving the BP neural network and this method has a broad prospect in English feature recognition.

Prediction of Myoelectric Biomarkers in Post-Stroke Gait

Electromyography (EMG) is sensitive to neuromuscular changes resulting from ischemic stroke and is considered a potential predictive tool of post-stroke gait and rehabilitation management. This study aimed to evaluate the potential myoelectric biomarkers for the classification of stroke-impaired muscular activity of the stroke patient group and the muscular activity of the control healthy adult group. We also proposed an EMG-based gait monitoring system consisting of a portable EMG device, cloud-based data processing, data analytics, and a health advisor service. This system was investigated with 48 stroke patients (mean age 70.6 years, 65% male) admitted into the emergency unit of a hospital and 75 healthy elderly volunteers (mean age 76.3 years, 32% male). EMG was recorded during walking using the portable device at two muscle positions: the bicep femoris muscle and the lateral gastrocnemius muscle of both lower limbs. The statistical result showed that the mean power frequency (MNF), median power frequency (MDF), peak power frequency (PKF), and mean power (MNP) of the stroke group differed significantly from those of the healthy control group. In the machine learning analysis, the neural network model showed the highest classification performance (precision: 88%, specificity: 89%, accuracy: 80%) using the training dataset and highest classification performance (precision: 72%, specificity: 74%, accuracy: 65%) using the testing dataset. This study will be helpful to understand stroke-impaired gait changes and decide post-stroke rehabilitation.

Toward Hand Pattern Recognition in Assistive and Rehabilitation Robotics Using EMG and Kinematics

Wearable hand robots are becoming an attractive means in the facilitating of assistance with daily living and hand rehabilitation exercises for patients after stroke. Pattern recognition is a crucial step toward the development of wearable hand robots. Electromyography (EMG) is a commonly used biological signal for hand pattern recognition. However, the EMG based pattern recognition performance in assistive and rehabilitation robotics post stroke remains unsatisfactory. Moreover, low cost kinematic sensors such as Leap Motion is recently used for pattern recognition in various applications. This study proposes feature fusion and decision fusion method that combines EMG features and kinematic features for hand pattern recognition toward application in upper limb assistive and rehabilitation robotics. Ten normal subjects and five post stroke patients participating in the experiments were tested with eight hand patterns of daily activities while EMG and kinematics were recorded simultaneously. Results showed that average hand pattern recognition accuracy for post stroke patients was 83% for EMG features only, 84.71% for kinematic features only, 96.43% for feature fusion of EMG and kinematics, 91.18% for decision fusion of EMG and kinematics. The feature fusion and decision fusion was robust as three different levels of noise was given to the classifiers resulting in small decrease of classification accuracy. Different channel combination comparisons showed the fusion classifiers would be robust despite failure of specific EMG channels which means that the system has promising potential in the field of assistive and rehabilitation robotics. Future work will be conducted with real-time pattern classification on stroke survivors.

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.

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, p < 0.0001). (2) The interface performed better than chance in the experiment where participants pricked balloons with a needle tip (779/1000, p < 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%, p < 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.

Myoelectric untethered robotic glove enhances hand function and performance on daily living tasks after stroke

Introduction

Wearable robots controlled using electromyography could motivate greater use of the affected upper extremity after stroke and enable bimanual activities of daily living to be completed independently.

Methods

We have developed a myoelectric untethered robotic glove (My-HERO) that provides five-finger extension and grip assistance.

Results

The myoelectric controller detected the grip and release intents of the 9 participants after stroke with 84.7% accuracy. While using My-HERO, all 9 participants performed better on the Fugl-Meyer Assessment-Hand (8.4 point increase, scale out of 14, p < 0.01) and the Chedoke Arm and Hand Activity Inventory (8.2 point increase, scale out of 91, p < 0.01). Established criteria for clinically meaningful important differences were surpassed for both the hand function and daily living task assessments. The majority of participants provided satisfaction and usability questionnaire scores above 70%. Seven participants desired to use My-HERO in the clinic and at home during their therapy and daily routines.

Conclusions

People with hand impairment after stroke value that myoelectric untethered robotic gloves enhance their motion and bimanual task performance and motivate them to use their muscles during engaging activities of daily living. They desire to use these gloves daily to enable greater independence and investigate the effects on neuromuscular recovery.

Decoding Attempted Hand Movements in Stroke Patients Using Surface Electromyography

Brain- and muscle-triggered exoskeletons have been proposed as a means for motor training after a stroke. With the possibility of performing different movement types with an exoskeleton, it is possible to introduce task variability in training. It is difficult to decode different movement types simultaneously from brain activity, but it may be possible from residual muscle activity that many patients have or quickly regain. This study investigates whether nine different motion classes of the hand and forearm could be decoded from forearm EMG in 15 stroke patients. This study also evaluates the test-retest reliability of a classical, but simple, classifier (linear discriminant analysis) and advanced, but more computationally intensive, classifiers (autoencoders and convolutional neural networks). Moreover, the association between the level of motor impairment and classification accuracy was tested. Three channels of surface EMG were recorded during the following motion classes: Hand Close, Hand Open, Wrist Extension, Wrist Flexion, Supination, Pronation, Lateral Grasp, Pinch Grasp, and Rest. Six repetitions of each motion class were performed on two different days. Hudgins time-domain features were extracted and classified using linear discriminant analysis and autoencoders, and raw EMG was classified with convolutional neural networks. On average, 79 ± 12% and 80 ± 12% (autoencoders) of the movements were correctly classified for days 1 and 2, respectively, with an intraclass correlation coefficient of 0.88. No association was found between the level of motor impairment and classification accuracy (Spearman correlation: 0.24). It was shown that nine motion classes could be decoded from residual EMG, with autoencoders being the best classification approach, and that the results were reliable across days; this may have implications for the development of EMG-controlled exoskeletons for training in the patient’s home.

Characterizing Multiple Patterns of Motor Intent Using Spatial-Temporal Information for Intuitively Active Motor Training in Stroke Survivors

Upper extremity motor function loss severely affects stroke survivors during daily life activities. Different rehabilitation robotic systems have been developed to allow stroke survivors regain their motor function. Meanwhile, most of the robots only operate in a passive mode and restrict the users to navigate predefined trajectories which may not align with their motion intent, thus limiting motor recovery. One way to resolve this issue would be to utilize a decoded movement intent to trigger intuitively active motor training for patients. In this direction, this study proposed and investigated the use of spatial-temporal neuromuscular descriptor (STD) for optimal decoding of multiple patterns of movement intents in patient to provide inputs for active motor training in the rehabilitation robotic systems. The STD performance was validated using High-Density surface electromyogram recordings from five stroke survivors who performed 21 limb movements. Experimental results show that the STD achieved a significant reduction in limb movement classification error (13.36%) even in the presence of the inevitable White Gaussian Noise compared to other methods (p<0.05). The STD also showed obvious class separability for individual movement. Findings from this study suggest that the STD may provide potential inputs for intuitively active motor training in stroke rehabilitation robotic systems.Clinical Relevance- This study showed that spatial-temporal neuromuscular information could aid adequate decoding of movement intents upon which intuitively active motor training could be achieved in stroke rehabilitation robotic systems.

Determining User Intent of Partly Dynamic Shoulder Tasks in Individuals With Chronic Stroke Using Pattern Recognition

Stroke remains the leading cause of long-term disability in the US. Although therapy can achieve limited improvement of paretic arm use and performance, weakness and abnormal muscle synergies-which cause unintentional elbow, wrist, and finger flexion during shoulder abduction-contribute significantly to limb disuse and compound rehabilitation efforts. Emerging wearable exoskeleton technology could provide powered abduction support for the paretic arm, but requires a clinically feasible, robust control scheme capable of differentiating multiple shoulder degrees-of-freedom. This study examines whether pattern recognition of sensor data can accurately identify user intent for 9 combinations of 1- and 2- degree-of-freedom shoulder tasks. Participants with stroke (n = 12) used their paretic and non-paretic arms, and healthy controls (n = 12) used their dominant arm to complete tasks on a lab-based robot involving combinations of abduction, adduction, and internal and external rotation of the shoulder. We examined the effect of arm (paretic, non-paretic), load level (25% vs 50% maximal voluntary torque), and dataset (electromyography, load cell, or combined) on classifier performance. Results suggest that paretic arm, lower load levels, and using load cell or EMG data alone reduced classifier accuracy. However, this method still shows promise. Further work will examine classifier-user interaction during active control of a robotic device and optimization/minimization of sensors.

Myoelectric Control of a Soft Hand Exoskeleton Using Kinematic Synergies

Soft hand exoskeletons offer a lightweight, low-profile alternative to rigid rehabilitative robotic systems, enabling their use to restore activities of daily living (ADL) in those with hand paresis due to stroke or other conditions. The hand exoskeleton with embedded synergies (HEXOES) is a soft cable-driven hand exoskeleton capable of independently actuating and sensing 10 degrees of freedom (DoF) of the hand. Control of the 10 DoF exoskeleton is dimensionally reduced using three manually defined synergies in software corresponding to thumb, index, and 3-finger flexion and extension. In this paper, five healthy subjects control HEXOES using a neural network which decodes synergy weights from contralateral electromyography (EMG) activity. The three synergies are manipulated in real time to grasp and lift 15 ADL objects of various sizes and weights. The neural network's training and validation mean squared error, object grasp time, and grasp success rate were measured for five healthy subjects. The final training error of the neural network was 4.8 ± 1.8% averaged across subjects and tasks, with 8.3 ± 3.4% validation error. The time to reach, grasp, and lift an object was 11.15 ± 4.35 s on average, with an average success rate of 66.7% across all objects. The complete system demonstrates real time use of biosignals and machine learning to allow subjects to operate kinematic synergies to grasp objects using a wearable hand exoskeleton. Future work and applications are further discussed, including possible design improvements and enrollment of individuals with stroke.

Hands-Free Human-Computer Interface Based on Facial Myoelectric Pattern Recognition

Patients with no or limited hand function usually have difficulty in using conventional input devices such as a mouse or a touch screen. Having the ability of manipulating electronic devices can give patients full access to the digital world, thereby increasing their independence and confidence, and enriching their lives. In this study, a hands-free human-computer interface was developed in order to help patients manipulate computers using facial movements. Five facial movement patterns were detected by four electromyography (EMG) sensors, and classified using myoelectric pattern recognition algorithms. Facial movement patterns were mapped to cursor actions including movements in different directions and click. A typing task and a drawing task were designed in order to assess the interaction performance of the interface in daily use. Ten able-bodied subjects participated in the experiment. In the typing task, the median path efficiency was 80.4%, and the median input rate was 5.9 letters per minute. In the drawing task, the median time to accomplish was 239.9 s. Moreover, all the subjects achieved high classification accuracy (median: 98.0%). The interface driven by facial EMG achieved high performance, and will be assessed on patients with limited hand functions in the future.

Offline and online myoelectric pattern recognition analysis and real-time control of a robotic hand after spinal cord injury

OBJECTIVE

The objective of this study was to investigate the feasibility of applying myoelectric pattern recognition for controlling a robotic hand in individuals with spinal cord injury (SCI).

APPROACH

Surface electromyogram (sEMG) signals of six hand motion patterns were recorded from 12 subjects with SCI. Online and offline classification performance of two classifiers (Gaussian Naive Bayes classifier, GNB, and support vector machine, SVM) were investigated. An exoskeleton hand was then controlled in real-time using the classification results. The control accuracy and its correlation with function assessments were investigated.

MAIN RESULTS

Average offline classification accuracy of all tested SCI subjects was (73.6  ±  14.0)% for GNB and (77.6  ±  11.6)% for SVM, respectively. Average online classification accuracy was significantly lower, (64.3  ±  15.0)% for GNB and (70.2  ±  13.2)% for SVM. Average control accuracy of (81.0  ±  16.3)% was achieved in real-time control of the robotic hand using myoelectric pattern recognition. Correlation between control accuracy and grip/pinch force was observed.

SIGNIFICANCE

The results show that it is feasible to extract hand motion intent from individuals with SCI and control a robotic hand device using myoelectric pattern recognition. The performance of real-time control can be predicted based on functional assessments.