Control Of Artificial Limbs Research Articles

Deep Learning for Enhanced Prosthetic Control: Real-Time Motor Intent Decoding for Simultaneous Control of Artificial Limbs.

The development of advanced prosthetic devices that can be seamlessly used during an individual's daily life remains a significant challenge in the field of rehabilitation engineering. This study compares the performance of deep learning architectures to shallow networks in decoding motor intent for prosthetic control using electromyography (EMG) signals. Four neural network architectures, including a feedforward neural network with one hidden layer, a feedforward neural network with multiple hidden layers, a temporal convolutional network, and a convolutional neural network with squeeze-and-excitation operations were evaluated in real-time, human-in-the-loop experiments with able-bodied participants and an individual with an amputation. Our results demonstrate that deep learning architectures outperform shallow networks in decoding motor intent, with representation learning effectively extracting underlying motor control information from EMG signals. Furthermore, the observed performance improvements by using deep neural networks were consistent across both able-bodied and amputee participants. By employing deep neural networks instead of a shallow network, more reliable and precise control of a prosthesis can be achieved, which has the potential to significantly enhance prosthetic functionality and improve the quality of life for individuals with amputations.

Advancing task recognition towards artificial limbs control with ReliefF-based deep neural network extreme learning

In the rapidly advancing field of biomedical engineering, effective real-time control of artificial limbs is a pressing research concern. Addressing this, the current study introduces a pioneering method for augmenting task recognition in prosthetic control systems, combining a ReliefF-based Deep Neural Networks (DNNs) approach. This paper has leveraged the MILimbEEG dataset, a comprehensive rich source collection of EEG signals, to calculate statistical features of Arithmetic Mean (AM), Standard Deviation (SD), and Skewness (S) across various motor activities. Supreme Feature Selection (SFS), of the adopted time-domain features, was performed using the ReliefF algorithm. The highest scored DNN-ReliefF developed model demonstrated remarkable performance, achieving accuracy, precision, and recall rates of 97.4 %, 97.3 %, and 97.4 %, respectively. In contrast, a traditional DNN model yielded accuracy, precision, and recall rates of 50.8 %, 51.1 %, and 50.8 %, highlighting the significant improvements made possible by incorporating SFS. This stark contrast underscores the transformative potential of incorporating ReliefF, situating the DNN-ReliefF model as a robust platform for forthcoming advancements in real-time prosthetic control systems.

Motor Imagery EEG Decoding Based on Multi-scale Hybrid Networks and Feature Enhancement.

Motor Imagery (MI) based on Electroencephalography (EEG), a typical Brain-Computer Interface (BCI) paradigm, can communicate with external devices according to the brain's intentions. Convolutional Neural Networks (CNN) are gradually used for EEG classification tasks and have achieved satisfactory performance. However, most CNN-based methods employ a single convolution mode and a convolution kernel size, which cannot extract multi-scale advanced temporal and spatial features efficiently. What's more, they hinder the further improvement of the classification accuracy of MI-EEG signals. This paper proposes a novel Multi-Scale Hybrid Convolutional Neural Network (MSHCNN) for MI-EEG signal decoding to improve classification performance. The two-dimensional convolution is used to extract temporal and spatial features of EEG signals and the one-dimensional convolution is used to extract advanced temporal features of EEG signals. In addition, a channel coding method is proposed to improve the expression capacity of the spatiotemporal characteristics of EEG signals. We evaluate the performance of the proposed method on the dataset collected in the laboratory and BCI competition IV 2b, 2a, and the average accuracy is at 96.87%, 85.25%, and 84.86%, respectively. Compared with other advanced methods, our proposed method achieves higher classification accuracy. Then we use the proposed method for an online experiment and design an intelligent artificial limb control system. The proposed method effectively extracts EEG signals' advanced temporal and spatial features. Additionally, we design an online recognition system, which contributes to the further development of the BCI system.

Finger Gesture Recognition Using Sensing and Classification of Surface Electromyography Signals With High-Precision Wireless Surface Electromyography Sensors.

Finger gesture recognition (FGR) plays a crucial role in achieving, for example, artificial limb control and human-computer interaction. Currently, the most common methods of FGR are visual-based, voice-based, and surface electromyography (EMG)-based ones. Among them, surface EMG-based FGR is very popular and successful because surface EMG is a cumulative bioelectric signal from the surface of the skin that can accurately and intuitively represent the force of the fingers. However, existing surface EMG-based methods still cannot fully satisfy the required recognition accuracy for artificial limb control as the lack of high-precision sensor and high-accurate recognition model. To address this issue, this study proposes a novel FGR model that consists of sensing and classification of surface EMG signals (SC-FGR). In the proposed SC-FGR model, wireless sensors with high-precision surface EMG are first developed for acquiring multichannel surface EMG signals from the forearm. Its resolution is 16 Bits, the sampling rate is 2 kHz, the common-mode rejection ratio (CMRR) is less than 70 dB, and the short-circuit noise (SCN) is less than 1.5 μV. In addition, a convolution neural network (CNN)-based classification algorithm is proposed to achieve FGR based on acquired surface EMG signals. The CNN is trained on a spectrum map transformed from the time-domain surface EMG by continuous wavelet transform (CWT). To evaluate the proposed SC-FGR model, we compared it with seven state-of-the-art models. The experimental results demonstrate that SC-FGR achieves 97.5% recognition accuracy on eight kinds of finger gestures with five subjects, which is much higher than that of comparable models.

EMG-based Monitoring Muscle Contraction Force to Determine Most Effective Exercise

—Electromyography (EMG) signals have been widely applied in fields of biomedical, clinical and human-machine interface [1]. And surface EMG (sEMG) technology could provide a non-invasive way to detect the stimulation degree of muscles and enable the control of artificial limbs. In this research, the main focus lies on one of the social problems—sedentariness. With the development of internet and electronic economy, computing time has undergone a steady increase around the globe, which extends the job time in long-term sitting [2]. Together with some non-ergonomic designs of chairs, many officers might experience an abnormal sitting posture and get injured in their muscles unconsciously. To scope the phenomenon in detail, a simulation is applied to reappear the condition of sedentariness of a common office worker. And a series of easily achieved movements were investigated to find out the corresponding effect in muscle exercise to compensate the negative influence caused by long-term sitting.

A Hybrid Deep Learning Framework for Estimation of Elbow Flexion Force via Electromyography

Real-time and accurate estimation is beneficial for intention recognition, muscle rehabilitation evaluation and artificial limb control. However, it is difficult to estimate the elbow flexion force accurately. The aim of our model is to estimate elbow flexion muscle force, which can be used for elbow joint health assessment and prosthetic control studies. This paper proposed an end-to-end deep learning framework by fusing Convolutional Neural Network (CNN) and Bidirectional Long Short-Term Memory (BiLSTM) neural network with attention mechanism, which is more suitable for time series EMG signal to improve the feature extraction ability and achieve a high flexion force estimation accuracy. Experimental results indicated that the proposed method can automatically extract the proper features of elbow motion behaviors without professional knowledge in feature extraction model. Moreover, the experimental result shows that the proposed framework performs well in the accuracy and generalization ability, outperforming the state-of-the-art methods.

GADF/GASF-HOG:feature extraction methods for hand movement classification from surface electromyography

ObjectiveHuman intention gesture recognition is widely used in hand rehabilitation, artificial limb control, teleoperation, human-computer interaction and other fields. It has great application value, however, how to extract human intention gesture accurately has been a research hotspot.ApproachInspired by the image processing technology of machine vision, the surface electromyographic (sEMG) signal was selected as the source signal of motion intention in this work, and the original sEMG signal was converted into Gramian Angular Summation/Difference Field (GASF/GADF) image. Then, Histogram of Oriented Gradient (HOG) features of the corresponding GADF and GASF image were extracted. The extracted features are named as GASF-HOG and GADF-HOG. The Bagging method was used to map the features to six common gestures to realize the classification of intention gestures. Ten volunteers participated in the experiment, and the experimental data were used to verify the proposed method.Main resultsThe experimental results showed that the average accuracies of the proposed methods (GADF-HOG with Bagging, GASF-HOG with Bagging) were as follow: GADF-HOG with Bagging was with 95.73 ± 1.90%, and GASF-HOG with Bagging was with 93.63 ± 1.54%.SignificanceThe method proposed in this paper is inspired by image processing technology of machine vision, which provides a new idea about the human intention gesture recognition by combining the interdisciplinary knowledge.

Classification of hand movements using variational mode decomposition and composite permutation entropy index with surface electromyogram signals

Research of human hand movements recognition can be applied to artificial limb control, motion recognition of wearable exoskeleton, human–computer interaction in virtual reality and so on. Surface Electromyogram (sEMG) signal is the preferred source. There are many researches on how to extract information from sEMG signal and apply it to human motion recognition. However, how to extract the feature signal from sEMG signal is a difficult problem in the research of human hand movement recognition based on sEMG signal. In this paper, a method based on Variational Mode Decomposition (VMD) and composite permutation entropy index (CPEI) method is proposed for hand motion classification. Previously, the VMD method had not been used in human hand motion recognition studies. The method proposed in this work applies the VMD method to decompose the original sEMG signal into multiple Variational Mode Functions (VMFs) and calculate the corresponding CPEI of each signal component. Three feature selection methods (Infinite Latent Feature Selection (ILFS), ReliefF, and Laplacian Score) were applied to rank the features and remove the unimportant features. Three classifiers (Naive Bayes, K-NN, and Bagging) were used to recognize the hand actions. Ten volunteers participated in the experiment, and the experimental data were used to verify the proposed method. The average accuracy was 94.28 ± 1.26% for the proposed method with Laplacian Score for feature sorting and selection, and Bagging as classifier. Besides, 600 randomly selected hand movements are predicted (CPU is i5-8250U, ram is 8g, processing software is Spyder, python 3.7), and the corresponding execution time of proposed method is 0.56 s.

Feasibility of Tracking Multiple Implanted Magnets With a Myokinetic Control Interface: Simulation and Experimental Evidence Based on the Point Dipole Model.

The quest for an intuitive and physiologically appropriate human-machine interface for the control of dexterous prostheses is far from being completed. To control a hand prosthesis, a possible approach could consist in using information related to the displacement of forearm muscles of an amputee during contraction. We recently proposed that muscle displacement could be monitored by implanting passive magnetic markers (MMs- i.e., permanent magnets) in them. We dubbed this the myokinetic interface. However, besides the system feasibility, how much its accuracy, precision and computation time are affected by the number and distribution of both the MMs and the sensors used to record the MF was not quantified. Here we investigated, through simulations validated with a physical system, the performance of a system capable to track position and orientation of up to 9 MMs using information from up to 112 sensors in a volume resembling the dimensions of the human forearm. The system was able to track up to 7 MMs in 450 ms, demonstrating position/orientation accuracies in the range of 1mm/5°. The comparison with the experimental recordings demonstrated a median difference with the simulations in the order of 0.45mm. We were able to formulate general guidelines for the implementation of magnetic tracking systems. Our results pave the way towards the development of new human-machine interfaces for the control of artificial limbs, but they are also interesting for the whole range of biomedical engineering applications exploiting magnetic tracking.

Online mapping of EMG signals into kinematics by autoencoding

BackgroundIn this paper, we propose a nonlinear minimally supervised method based on autoencoding (AEN) of EMG for myocontrol. The proposed method was tested against the state-of-the-art (SOA) control scheme using a Fitts’ law approach.MethodsSeven able-bodied subjects performed a series of target acquisition myoelectric control tasks using the AEN and SOA algorithms for controlling two degrees-of-freedom (radial/ulnar deviation and flexion/extension of the wrist), and their online performance was characterized by six metrics.ResultsBoth methods allowed a completion rate close to 100%, however AEN outperformed SOA for all other performance metrics, e.g. it allowed to perform the tasks on average in half the time with respect to SOA. Moreover, the amount of information transferred by the proposed method in bit/s was nearly twice the throughput of SOA.ConclusionsThese results show that autoencoders can map EMG signals into kinematics with the potential of providing intuitive and dexterous control of artificial limbs for amputees.

Low-Dimensional Synergistic Representation of Bilateral Reaching Movements.

Kinematic and neuromuscular synergies have been found in numerous aspects of human motion. This study aims to determine how effectively kinematic synergies in bilateral upper arm movements can be used to replicate complex activities of daily living (ADL) tasks using a sparse optimization algorithm. Ten right-handed subjects executed 18 rapid and 11 natural-paced ADL tasks requiring bimanual coordination while sitting at a table. A position tracking system was used to track the subjects’ arms in space, and angular velocities over time for shoulder abduction, shoulder flexion, shoulder internal rotation, and elbow flexion for each arm were computed. Principal component analysis (PCA) was used to generate kinematic synergies from the rapid-paced task set for each subject. The first three synergies accounted for 80.3 ± 3.8% of variance, while the first eight accounted for 94.8 ± 0.85%. The first and second synergies appeared to encode symmetric reaching motions which were highly correlated across subjects. The first three synergies were correlated between left and right arms within subjects, whereas synergies four through eight were not, indicating asymmetries between left and right arms in only the higher order synergies. The synergies were then used to reconstruct each natural-paced task using the l1-norm minimization algorithm. Temporal dilations of the synergies were introduced in order to model the temporal scaling of movement patterns achieved by the cerebellum and basal ganglia as reported previously in the literature. Reconstruction error was reduced by introducing synergy dilations, and cumulative recruitment of several synergies was significantly reduced in the first 10% of training task time by introducing temporal dilations. The outcomes of this work could open new scenarios for the applications of postural synergies to the control of robotic systems, with potential applications in rehabilitation. These synergies not only help in providing near-natural control but also provide simplified strategies for design and control of artificial limbs. Potential applications of these bilateral synergies were discussed and future directions were proposed.

In vivo Characterization of Amorphous Silicon Carbide As a Biomaterial for Chronic Neural Interfaces.

Implantable microelectrode arrays (MEAs) offer clinical promise for prosthetic devices by enabling restoration of communication and control of artificial limbs. While proof-of-concept recordings from MEAs have been promising, work in animal models demonstrates that the obtained signals degrade over time. Both material robustness and tissue response are acknowledged to have a role in device lifetime. Amorphous Silicon carbide (a-SiC), a robust material that is corrosion resistant, has emerged as an alternative encapsulation layer for implantable devices. We systematically examined the impact of a-SiC coating on Si probes by immunohistochemical characterization of key markers implicated in tissue-device response. After implantation, we performed device capture immunohistochemical labeling of neurons, astrocytes, and activated microglia/macrophages after 4 and 8 weeks of implantation. Neuron loss and microglia activation were similar between Si and a-SiC coated probes, while tissue implanted with a-SiC displayed a reduction in astrocytes adjacent to the probe. These results suggest that a-SiC has a similar biocompatibility profile as Si, and may be suitable for implantable MEA applications as a hermetic coating to prevent material degradation.

English

English

Myoelectric control of artificial limb inspired by quantum information processing

Precise and elegant coordination of a prosthesis across many degrees of freedom represents a significant challenge to efficient rehabilitation of people with limb deficiency. Processing the electrical neural signals collected from the surface of the remnant muscles of the stump is a common way to initiate and control the different movements available to the artificial limb. Based on the assumption that there are distinguishable and repeatable signal patterns among different types of muscular activation, the problem of prosthesis control reduces to one of pattern recognition. Widely accepted classical methods for pattern recognition, however, cannot provide simultaneous and proportional control of the artificial limb. Here we show that, in principle, quantum information processing of the neural signals allows us to overcome the above-mentioned difficulties, suggesting a very simple scheme for myoelectric control of artificial limb with advanced functionalities.

Electromyographic assessment on transfemoral amputees to possibly control artificial lower limbs

This study aimed to assess the electromyographic feasibility of different hip muscles on transfemoral unilateral lower limb amputees to a possible control of artificial limbs. The volunteers were split into two groups: eight males, physically active amputees and a control group composed of eight males, healthy, non–amputees individuals. The hip muscles were assessed in accordance to the general consensus by SENIAM project, and some adaptations were made in the amputees when the anatomical references were absents, such as the knee articulation. Thus, agonists and antagonists skeleton muscles were evaluated during the isometric contraction of hips movements of extension and flexion, which were controlled by an isokinect chair. The median frequency (Fmed) values did not represent significant differences and the RMS is the best muscle activity of amputees.

An osseointegrated human-machine gateway for long-term sensory feedback and motor control of artificial limbs.

A major challenge since the invention of implantable devices has been a reliable and long-term stable transcutaneous communication. In the case of prosthetic limbs, existing neuromuscular interfaces have been unable to address this challenge and provide direct and intuitive neural control. Although prosthetic hardware and decoding algorithms are readily available, there is still a lack of appropriate and stable physiological signals for controlling the devices. We developed a percutaneous osseointegrated (bone-anchored) interface that allows for permanent and unlimited bidirectional communication with the human body. With this interface, an artificial limb can be chronically driven by implanted electrodes in the peripheral nerves and muscles of an amputee, outside of controlled environments and during activities of daily living, thus reducing disability and improving quality of life. We demonstrate in one subject, for more than 1 year, that implanted electrodes provide a more precise and reliable control than surface electrodes, regardless of limb position and environmental conditions, and with less effort. Furthermore, long-term stable myoelectric pattern recognition and appropriate sensory feedback elicited via neurostimulation was demonstrated. The opportunity to chronically record and stimulate the neuromuscular system allows for the implementation of intuitive control and naturally perceived sensory feedback, as well as opportunities for the prediction of complex limb motions and better understanding of sensory perception. The permanent bidirectional interface presented here is a critical step toward more natural limb replacement, by combining stable attachment with permanent and reliable human-machine communication.

Novel wearable EMG sensors based on nanowire technology.

Wearable electrodes made of silver nanowires (AgNWs) have demonstrated great potential for sensing a variety of physical and physiological signals. This paper aimed to study the feasibility of AgNWs electrodes for measuring surface electromyographic (sEMG) signals. One human subject was recruited and instructed to perform wrist extension repetitively or to produce no movement in the experiment. sEMG signals were collected from the right extensor digitorum communis of the human subject by an AgNWs electrode and a commercially available Ag/AgCl wet sEMG electrode, separately. The quality of recorded sEMG in time and frequency domains was compared between the two types of electrodes. The results showed that the sEMG signals recorded by the AgNW electrode were comparable with that by the Ag/AgCl electrode. Since the dry AgNWs electrodes are flexible, wearable, and potentially robust for daily use, novel AgNW-based EMG electrodes are promising for many biomedical applications, such as myoelectric control of artificial limbs.

A differential evolution based energy trajectory planner for artificial limb control using motor imagery EEG signal

Current research on neuro-prosthetics is aimed at designing several computational models and techniques to trigger the neuro-motor rehabilitative aids. Researchers are taking keen interest to accurately classify the stimulated electroencephalography (EEG) signals to interpret motor imagery tasks. In this paper we aim to classify the finger-, elbow- and shoulder-classification along with left- and right-hand classification to move a simulated robot arm in 3D space towards a target of known location. The contribution of the paper lies in the design of an energy optimal trajectory planner, based on differential evolution, which would decide the optimal path for the robot arm to move towards the target based on the classifier output. Each different set of movements consists of a trajectory planner which is activated by the classifier output. The energy distribution of wavelet coefficients of the incoming EEG signals is used as features to be used as inputs in a naïve Bayesian classifier to discriminate among the different mental tasks. The average training classification accuracy obtained is 76.88% and the success rate of the simulated robot arm reaching the target is 85%.

Real-Time and Simultaneous Control of Artificial Limbs Based on Pattern Recognition Algorithms

The prediction of simultaneous limb motions is a highly desirable feature for the control of artificial limbs. In this work, we investigate different classification strategies for individual and simultaneous movements based on pattern recognition of myoelectric signals. Our results suggest that any classifier can be potentially employed in the prediction of simultaneous movements if arranged in a distributed topology. On the other hand, classifiers inherently capable of simultaneous predictions, such as the multi-layer perceptron (MLP), were found to be more cost effective, as they can be successfully employed in their simplest form. In the prediction of individual movements, the one-vs-one (OVO) topology was found to improve classification accuracy across different classifiers and it was therefore used to benchmark the benefits of simultaneous control. As opposed to previous work reporting only offline accuracy, the classification performance and the resulting controllability are evaluated in real time using the motion test and target achievement control (TAC) test, respectively. We propose a simultaneous classification strategy based on MLP that outperformed a top classifier for individual movements (LDA-OVO), thus improving the state-of-the-art classification approach. Furthermore, all the presented classification strategies and data collected in this study are freely available in BioPatRec, an open source platform for the development of advanced prosthetic control strategies.

Intuitive, Online, Simultaneous, and Proportional Myoelectric Control Over Two Degrees-of-Freedom in Upper Limb Amputees

We propose an approach for online simultaneous and proportional myoelectric control of two degrees-of-freedom (DoF) of the wrist, using surface electromyographic signals. The method is based on the nonnegative matrix factorization (NMF) of the wrist muscle activation to extract low-dimensional control signals translated by the user into kinematic variables. This procedure does not need a training set of signals for which the kinematics is known (labeled dataset) and is thus unsupervised (although it requires an initial calibration without labeled signals). The estimated control signals using NMF are used to directly control two DoFs of wrist. The method was tested on seven subjects with upper limb deficiency and on seven able-bodied subjects. The subjects performed online control of a virtual object with two DoFs to achieve goal-oriented tasks. The performance of the two subject groups, measured as the task completion rate, task completion time, and execution efficiency, was not statistically different. The approach was compared, and demonstrated to be superior to the online control by the industrial state-of-the-art approach. These results show that this new approach, which has several advantages over the previous myoelectric prosthetic control systems, has the potential of providing intuitive and dexterous control of artificial limbs for amputees.