An official website of the United States government
Abstract Pathological tremors significantly affect the quality of life for patients worldwide. Rehabilitation exoskeletons serve as one of the solutions to alleviate these pathological tremors, and voluntary motion prediction-based motion planning has been employed to enhance the performance of these devices. This paper presents a method for predicting future voluntary movement in tremor-alleviating rehabilitation exoskeletons that use voluntary motion prediction-based motion planning. In this study, a Convolutional Neural Network and Transformer architecture based neural network work with EMG sensors to predict future voluntary movements. The results show that approach performs well in predicting future voluntary movements, but there is still a limitation to filter out the tremors completely. In summary, we provide a concept for predicting future voluntary movement, which has the potential to improve the effectiveness of rehabilitation exoskeletons in tremor alleviation.
more »
« less
Exploring Deep Learning Models for Pathological Tremors Prediction Using EMG and Kinematic Measurements
Abstract Pathological tremor is a common neuromuscular disorder that significantly affects the quality of life for patients worldwide. With recent developments in robotics, rehabilitation exoskeletons serve as one of the solutions to alleviate these tremors. Accurate predictive modeling of tremor signals can be used to provide alleviation from these tremors via various currently available solutions like adaptive deep brain stimulation, electrical stimulation and rehabilitation orthoses, motivating us to explore better modeling of tremors for long-term predictions and analysis. This study is a preliminary step towards the prediction of tremors using artificial neural networks using EMG signals, leveraging the 20–100 ms of Electromechanical Delay. The kinematics and EMG data of a publicly available Parkinsonian tremor dataset is first analyzed, which confirms that the underlying EMGs have similar frequency composition as the actual tremor. 2 hybrid CNN-LSTM based deep learning architectures are then proposed to predict the tremor kinematics ahead of time using EMG signals and tremor kinematics history, and the results are compared with baseline models. The motivation behind hybrid CNN-LSTM models is to exploit both the temporal and spatial dependencies using CNN and LSTM respectively. This is then further extended by adding constraints-based losses in an attempt to further improve the predictions.
more »
« less
- Award ID(s):
- 2306984
- PAR ID:
- 10621682
- Publisher / Repository:
- American Society of Mechanical Engineers
- Date Published:
- ISBN:
- 978-0-7918-8835-3
- Format(s):
- Medium: X
- Location:
- Washington, DC, USA
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Abstract Estimating muscle forces is crucial for understanding joint dynamics and improving rehabilitation strategies, particularly for patients with neurological disorders who suffer from impaired muscle function. Muscle forces are directly proportional to muscle activations, which can be obtained using electromyography (EMG). EMG-driven modeling estimates muscle forces and joint moments from muscle activations. While surface muscles' activations can be obtained using surface electrodes, deep muscles require invasive methods and are not readily available for real-time applications. This study aims to extend our previously developed method for a single unmeasured muscle to a comprehensive approach for the simultaneous prediction of multiple unmeasured muscle activations in the upper extremity using muscle synergy extrapolation and EMG-driven modeling. By employing non-negative matrix factorization to decompose known EMG data into synergy components, the activations of unmeasured muscles are reconstructed with high accuracy by minimizing differences between joint moments obtained by EMG-driven modeling and inverse dynamics. This methodology is validated through experimentally collected muscle activations, demonstrating over 90% correlation with EMG signals in various scenarios.more » « less
-
Robotic exoskeletons can assist humans with walking by providing supplemental torque in proportion to the user's joint torque. Electromyographic (EMG) control algorithms can estimate a user's joint torque directly using real-time EMG recordings from the muscles that generate the torque. However, EMG signals change as a result of supplemental torque from an exoskeleton, resulting in unreliable estimates of the user's joint torque during active exoskeleton assistance. Here, we present an EMG control framework for robotic exoskeletons that provides consistent joint torque predictions across varying levels of assistance. Experiments with three healthy human participants showed that using diverse training data (from different levels of assistance) enables robust torque predictions, and that a convolutional neural network (CNN), but not a Kalman filter (KF), can capture the non-linear transformations in EMG due to exoskeleton assistance. With diverse training, the CNN could reliably predict joint torque from EMG during zero, low, medium, and high levels of exoskeleton assistance [root mean squared error (RMSE) below 0.096 N-m/kg]. In contrast, without diverse training, RMSE of the CNN ranged from 0.106 to 0.144 N-m/kg. RMSE of the KF ranged from 0.137 to 0.182 N-m/kg without diverse training, and did not improve with diverse training. When participant time is limited, training data should emphasize the highest levels of assistance first and utilize at least 35 full gait cycles for the CNN. The results presented here constitute an important step toward adaptive and robust human augmentation via robotic exoskeletons. This work also highlights the non-linear reorganization of locomotor output when using assistive exoskeletons; significant reductions in EMG activity were observed for the soleus and gastrocnemius, and a significant increase in EMG activity was observed for the erector spinae. Control algorithms that can accommodate spatiotemporal changes in muscle activity have broad implications for exoskeleton-based assistance and rehabilitation following neuromuscular injury.more » « less
-
Predicting the transpiration stream concentration factor (TSCF) and other concentration factors is essential in understanding the plant uptake of organic contaminants. Traditional mechanistic and numerical modeling methods often fail to reliably predict the TSCF. This study developed a hybrid deep model to predict TSCF by integrating convolutional neural networks (CNNs) and long short-term memory (LSTM) networks. This hybrid CNN-LSTM model used eight physicochemical properties of organic contaminants to predict TSCF. The training procedure for this hybrid model was successful. The results indicated the training and test losses for predicting TSCF were both in the same order and close to zero. This study showed that the hybrid CNN-LSTM model can outperform mechanistic models and have higher performances compared to classical machine learning models. Feature importance analysis using extreme gradient boosting highlighted the role and importance of lipophilicity in predicting uptake and translocation of organic contaminants.more » « less
-
Objective: In this study, we aimed to develop a novel electromyography (EMG)-based neural machine interface (NMI), called the Neural Network-Musculoskeletal hybrid Model (N2M2), to decode continuous joint angles. Our approach combines the concepts of machine learning and musculoskeletal modeling. Methods: We compared our novel design with a musculoskeletal model (MM) and 2 continuous EMG decoders based on artificial neural networks (ANNs): multilayer perceptrons (MLPs) and nonlinear autoregressive neural networks with exogenous inputs (NARX networks). EMG and joint kinematics data were collected from 10 non-disabled and 1 transradial amputee subject. The offline performance tested across 3 different conditions (i.e., varied arm postures, shifted electrode locations, and noise-contaminated EMG signals) and online performance for a virtual postural matching task was quantified. Finally, we implemented the N2M2 to operate a prosthetic hand and tested functional task performance. Results: The N2M2 made more accurate predictions than the MLP in all postures and electrode locations (p < 0.003). For estimated MCP joint angles, the N2M2 was less sensitive to noisy EMG signals than the MM or NARX network with respect to error (p < 0.032) as well as the NARX network with respect to correlation (p = 0.007). Additionally, the N2M2 had better online task performance than the NARX network (p ≤ 0.030). Conclusion: Overall, we have found that combining the concepts of machine learning and musculoskeletal modeling has resulted in a more robust joint kinematics decoder than either concept individually. Significance: The outcome of this study may result in a novel, highly reliable controller for powered prosthetic hands.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Conference Paper:
- https://doi.org/10.1115/DETC2024-146284
