InverseMuscleNET: Alternative Machine Learning Solution to Static Optimization and Inverse Muscle Modeling

Nasr, Ali; Inkol, Keaton A.; Bell, Sydney; McPhee, John

doi:10.3389/fncom.2021.759489

Cited by 14 publications

(12 citation statements)

References 53 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Furthermore, the increase in muscle activity when accounting for GH stability is greater when the same movements are performed with a handheld weight. Interestingly, GH stability had virtually no effect on activations of surface muscles, whose EMG recordings are typically used to validate estimated muscle activity [ 18 , 24 , 27 , 34 , 46 , 47 ]. Indeed our comparisons indicated the RMR estimates with and without GH stability were comparable to CMC with respect to filtered EMG signals, with all methods having excellent agreement (MAE ≤ 0.1, Table 1 ).…”

Section: Discussionmentioning

confidence: 99%

“…Neglecting passive forces leads to simplifications of muscle function [ 21 ], together with poor performances in estimating antagonist muscle activity at the GH joint [ 22 ]. Finally, recent data-driven machine learning methods achieved promising results [ 17 , 18 ], yet they currently disregard musculoskeletal properties altogether, retaining little direct connection with the way the human body actually functions.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Does enforcing glenohumeral joint stability matter? A new rapid muscle redundancy solver highlights the importance of non-superficial shoulder muscles

Belli,

Joshi,

Prendergast

et al. 2023

PLoS ONE

View full text Add to dashboard Cite

The complexity of the human shoulder girdle enables the large mobility of the upper extremity, but also introduces instability of the glenohumeral (GH) joint. Shoulder movements are generated by coordinating large superficial and deeper stabilizing muscles spanning numerous degrees-of-freedom. How shoulder muscles are coordinated to stabilize the movement of the GH joint remains widely unknown. Musculoskeletal simulations are powerful tools to gain insights into the actions of individual muscles and particularly of those that are difficult to measure. In this study, we analyze how enforcement of GH joint stability in a musculoskeletal model affects the estimates of individual muscle activity during shoulder movements. To estimate both muscle activity and GH stability from recorded shoulder movements, we developed a Rapid Muscle Redundancy (RMR) solver to include constraints on joint reaction forces (JRFs) from a musculoskeletal model. The RMR solver yields muscle activations and joint forces by minimizing the weighted sum of squared-activations, while matching experimental motion. We implemented three new features: first, computed muscle forces include active and passive fiber contributions; second, muscle activation rates are enforced to be physiological, and third, JRFs are efficiently formulated as linear functions of activations. Muscle activity from the RMR solver without GH stability was not different from the computed muscle control (CMC) algorithm and electromyography of superficial muscles. The efficiency of the solver enabled us to test over 3600 trials sampled within the uncertainty of the experimental movements to test the differences in muscle activity with and without GH joint stability enforced. We found that enforcing GH stability significantly increases the estimated activity of the rotator cuff muscles but not of most superficial muscles. Therefore, a comparison of shoulder model muscle activity to EMG measurements of superficial muscles alone is insufficient to validate the activity of rotator cuff muscles estimated from musculoskeletal models.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Does enforcing glenohumeral joint stability matter? A new rapid muscle redundancy solver highlights the importance of non-superficial shoulder muscles

Belli,

Joshi,

Prendergast

et al. 2023

PLoS ONE

View full text Add to dashboard Cite

show abstract

“…h, ε, and f are the right-hand side of transformation between the derivative of coordinates and generalized speeds, the error, and the reaction wrench that enforces the kinematic constraint equations. The activation signal, which is estimated via InverseMuscleNET, maps the human activation torque, joint torque, joint angle, joint velocity, and joint acceleration to the activation signals or EMG signals [45].…”

Section: Control System Evaluationmentioning

confidence: 99%

“…The activation signal, which is estimated via InverseMuscleNET, maps the human activation torque, joint torque, joint angle, joint velocity, and joint acceleration to the activation signals or EMG signals [ 45 ].…”

Section: Control System Evaluationmentioning

confidence: 99%

Model-Based Mid-Level Regulation for Assist-As-Needed Hierarchical Control of Wearable Robots: A Computational Study of Human-Robot Adaptation

2022

Self Cite

View full text Add to dashboard Cite

The closed-loop human–robot system requires developing an effective robotic controller that considers models of both the human and the robot, as well as human adaptation to the robot. This paper develops a mid-level controller providing assist-as-needed (AAN) policies in a hierarchical control setting using two novel methods: model-based and fuzzy logic rule. The goal of AAN is to provide the required extra torque because of the robot’s dynamics and external load compared to the human limb free movement. The human–robot adaptation is simulated using a nonlinear model predictive controller (NMPC) as the human central nervous system (CNS) for three conditions of initial (the initial session of wearing the robot, without any previous experience), short-term (the entire first session, e.g., 45 min), and long-term experiences. The results showed that the two methods (model-based and fuzzy logic) outperform the traditional proportional method in providing AAN by considering distinctive human and robot models. Additionally, the CNS actuator model has difficulty in the initial experience and activates both antagonist and agonist muscles to reduce movement oscillations. In the long-term experience, the simulation shows no oscillation when the CNS NMPC learns the robot model and modifies its weights to simulate realistic human behavior. We found that the desired strength of the robot should be increased gradually to ignore unexpected human–robot interactions (e.g., robot vibration, human spasticity). The proposed mid-level controllers can be used for wearable assistive devices, exoskeletons, and rehabilitation robots.

show abstract

“…(c) The absolute value of the signal amplitude or rectifying the signal was used for making the signal positive. (d) A second-order lowpass digital Butterworth filter with a normalized cutoff frequency of 7 Hz [51] was used to smooth the signal as evaluated and analyzed by Nasr et al [53].…”

Section: Emg Data Filteringmentioning

confidence: 99%

MuscleNET: mapping electromyography to kinematic and dynamic biomechanical variables by machine learning

Nasr

Bell

et al. 2021

J. Neural Eng.

View full text Add to dashboard Cite

Objective. This paper proposes machine learning models for mapping surface electromyography (sEMG) signals to regression of joint angle, joint velocity, joint acceleration, joint torque, and activation torque. Approach. The regression models, collectively known as MuscleNET, take one of four forms: ANN (Forward Artificial Neural Network), RNN (Recurrent Neural Network), CNN (Convolutional Neural Network), and RCNN (Recurrent Convolutional Neural Network). Inspired by conventional biomechanical muscle models, delayed kinematic signals were used along with sEMG signals as the machine learning model's input; specifically, the CNN and RCNN were modeled with novel configurations for these input conditions. The models' inputs contain either raw or filtered sEMG signals, which allowed evaluation of the filtering capabilities of the models. The models were trained using human experimental data and evaluated with different individual data. Main results. Results were compared in terms of regression error (using the root-mean-square) and model computation delay. The results indicate that the RNN (with filtered sEMG signals) and RCNN (with raw sEMG signals) models, both with delayed kinematic data, can extract underlying motor control information (such as joint activation torque or joint angle) from sEMG signals in pick-and-place tasks. The CNNs and RCNNs were able to filter raw sEMG signals. Significance. All forms of MuscleNET were found to map sEMG signals within 2 ms, fast enough for real-time applications such as the control of exoskeletons or active prostheses. The RNN model with filtered sEMG and delayed kinematic signals is particularly appropriate for applications in musculoskeletal simulation and biomechatronic device control.

show abstract

InverseMuscleNET: Alternative Machine Learning Solution to Static Optimization and Inverse Muscle Modeling

Cited by 14 publications

References 53 publications

Does enforcing glenohumeral joint stability matter? A new rapid muscle redundancy solver highlights the importance of non-superficial shoulder muscles

Does enforcing glenohumeral joint stability matter? A new rapid muscle redundancy solver highlights the importance of non-superficial shoulder muscles

Model-Based Mid-Level Regulation for Assist-As-Needed Hierarchical Control of Wearable Robots: A Computational Study of Human-Robot Adaptation

MuscleNET: mapping electromyography to kinematic and dynamic biomechanical variables by machine learning

Contact Info

Product

Resources

About