Estimation Of Ground Reaction Forces Research Articles

Design of a low-cost force insoles to estimate ground reaction forces during human gait

This paper proposes a low-cost electronic system for estimating ground reaction forces (GRF) during human gait. The device consists of one master node and two slave nodes. The master node sends instructions to slave nodes that sample and store data from two force insoles located at the participant’s feet. These insoles are equipped with 14 piezo-resistive FlexiForce A301 sensors (FSR). The slave nodes are attached to the ankles and feet of each participant. Subsequently, the start command is transmitted through the master node, which is connected to the USB port of a personal computer (PC). Once the walking session is completed, the information obtained by the slave nodes can be downloaded by accessing the access point generated by these devices through Wi-Fi communication. The GRF estimation system was validated with force platforms (BTS Bioengineering P6000, Italy), giving on average a fit measure equal to 68.71%±4.80% in dynamic situations. Future versions of this device are expected to increase this fit by using machine learning models.

Distributed Wearable Ultrasound Sensors Predict Isometric Ground Reaction Force.

Rehabilitation from musculoskeletal injuries focuses on reestablishing and monitoring muscle activation patterns to accurately produce force. The aim of this study is to explore the use of a novel low-powered wearable distributed Simultaneous Musculoskeletal Assessment with Real-Time Ultrasound (SMART-US) device to predict force during an isometric squat task. Participants (N = 5) performed maximum isometric squats under two medical imaging techniques; clinical musculoskeletal motion mode (m-mode) ultrasound on the dominant vastus lateralis and SMART-US sensors placed on the rectus femoris, vastus lateralis, medial hamstring, and vastus medialis. Ultrasound features were extracted, and a linear ridge regression model was used to predict ground reaction force. The performance of ultrasound features to predict measured force was tested using either the Clinical M-mode, SMART-US sensors on the vastus lateralis (SMART-US: VL), rectus femoris (SMART-US: RF), medial hamstring (SMART-US: MH), and vastus medialis (SMART-US: VMO) or utilized all four SMART-US sensors (Distributed SMART-US). Model training showed that the Clinical M-mode and the Distributed SMART-US model were both significantly different from the SMART-US: VL, SMART-US: MH, SMART-US: RF, and SMART-US: VMO models (p < 0.05). Model validation showed that the Distributed SMART-US model had an R2 of 0.80 ± 0.04 and was significantly different from SMART-US: VL but not from the Clinical M-mode model. In conclusion, a novel wearable distributed SMART-US system can predict ground reaction force using machine learning, demonstrating the feasibility of wearable ultrasound imaging for ground reaction force estimation.

Self-Supervised Learning Improves Accuracy and Data Efficiency for IMU-Based Ground Reaction Force Estimation.

Recent deep learning techniques hold promise to enable IMU-driven kinetic assessment; however, they require large extents of ground reaction force (GRF) data to serve as labels for supervised model training. We thus propose using existing self-supervised learning (SSL) techniques to leverage large IMU datasets to pre-train deep learning models, which can improve the accuracy and data efficiency of IMU-based GRF estimation. We performed SSL by masking a random portion of the input IMU data and training a transformer model to reconstruct the masked portion. We systematically compared a series of masking ratios across three pre-training datasets that included real IMU data, synthetic IMU data, or a combination of the two. Finally, we built models that used pre-training and labeled data to estimate GRF during three prediction tasks: overground walking, treadmill walking, and drop landing. When using the same amount of labeled data, SSL pre-training significantly improved the accuracy of 3-axis GRF estimation during walking compared to baseline models trained by conventional supervised learning. Fine-tuning SSL model with 1-10% of walking data yielded comparable accuracy to training baseline model with 100% of walking data. The optimal masking ratio for SSL is 6.25-12.5%. SSL leveraged large real and synthetic IMU datasets to increase the accuracy and data efficiency of deep-learning-based GRF estimation, reducing the need for labeled data. This work, with its open-source code and models, may unlock broader use cases of IMU-driven kinetic assessment by mitigating the scarcity of GRF measurements in practical applications.

Effect of sensor number and location on accelerometry-based vertical ground reaction force estimation during walking.

Knee osteoarthritis is a major cause of global disability and is a major cost for the healthcare system. Lower extremity loading is a determinant of knee osteoarthritis onset and progression; however, technology that assists rehabilitative clinicians in optimizing key metrics of lower extremity loading is significantly limited. The peak vertical component of the ground reaction force (vGRF) in the first 50% of stance is highly associated with biological and patient-reported outcomes linked to knee osteoarthritis symptoms. Monitoring and maintaining typical vGRF profiles may support healthy gait biomechanics and joint tissue loading to prevent the onset and progression of knee osteoarthritis. Yet, the optimal number of sensors and sensor placements for predicting accurate vGRF from accelerometry remains unknown. Our goals were to: 1) determine how many sensors and what sensor locations yielded the most accurate vGRF loading peak estimates during walking; and 2) characterize how prescribing different loading conditions affected vGRF loading peak estimates. We asked 20 young adult participants to wear 5 accelerometers on their waist, shanks, and feet and walk on a force-instrumented treadmill during control and targeted biofeedback conditions prompting 5% underloading and overloading vGRFs. We trained and tested machine learning models to estimate vGRF from the various sensor accelerometer inputs and identified which combinations were most accurate. We found that a neural network using one accelerometer at the waist yielded the most accurate loading peak vGRF estimates during walking, with average errors of 4.4% body weight. The waist-only configuration was able to distinguish between control and overloading conditions prescribed using biofeedback, matching measured vGRF outcomes. Including foot or shank acceleration signals in the model reduced accuracy, particularly for the overloading condition. Our results suggest that a system designed to monitor changes in walking vGRF or to deploy targeted biofeedback may only need a single accelerometer located at the waist for healthy participants.

Estimation of Ground Reaction Forces during Sports Movements by Sensor Fusion from Inertial Measurement Units with 3D Forward Dynamics Model.

Rotational jumps are crucial techniques in sports competitions. Estimating ground reaction forces (GRFs), a constituting component of jumps, through a biomechanical model-based approach allows for analysis, even in environments where force plates or machine learning training data would be impossible. In this study, rotational jump movements involving twists on land were measured using inertial measurement units (IMUs), and GRFs and body loads were estimated using a 3D forward dynamics model. Our forward dynamics and optimization calculation-based estimation method generated and optimized body movements using cost functions defined by motion measurements and internal body loads. To reduce the influence of dynamic acceleration in the optimization calculation, we estimated the 3D orientation using sensor fusion, comprising acceleration and angular velocity data from IMUs and an extended Kalman filter. As a result, by generating cost function-based movements, we could calculate biomechanically valid GRFs while following the measured movements, even if not all joints were covered by IMUs. The estimation approach we developed in this study allows for measurement condition- or training data-independent 3D motion analysis.

AI-Driven Virtual Sensors for Real-Time Dynamic Analysis of Mechanisms: A Feasibility Study

The measurement of the ground forces on a real structure or mechanism in operation can be time-consuming and expensive, particularly when production cannot be halted to install sensors. In cases in which disassembling the parts of the system to accommodate sensor installation is neither feasible nor desirable, observing the structure or mechanism in operation and quickly deducing its force trends would facilitate monitoring activities in industrial processes. This opportunity is gradually becoming a reality thanks to the coupling of artificial intelligence (AI) with design techniques such as the finite element and multi-body methods. Properly trained inferential models could make it possible to study the dynamic behavior of real systems and mechanisms in operation simply by observing them in real time through a camera, and they could become valuable tools for investigation during the operation of machinery and devices without the use of additional sensors, which are difficult to use and install. In this paper, the idea presented is developed and applied to a simple mechanism for which the reaction forces during operating conditions are to be determined. This paper explores the implementation of an innovative vision-based virtual sensor that, through data-driven training, is able to emulate traditional sensing solutions for the estimation of reaction forces. The virtual sensor and relative inferential model is validated in a scenario as close to the real world as possible, taking into account interfering inputs that add to the measurement uncertainty, as in a real-world measurement scenario. The results indicate that the proposed model has great robustness and accuracy, as evidenced by the low RMSE values in predicting the reaction forces. This demonstrates the model’s effectiveness in reproducing real-world scenarios, highlighting its potential in the real-time estimation of ground reaction forces in industrial settings. The success of this vision-based virtual sensor model opens new avenues for more robust, accurate, and cost-effective solutions for force estimation, addressing the challenges of uncertainty and the limitations of physical sensor deployment.

Jumping towards field-based ground reaction force estimation and assessment with OpenCap

Low-cost and field-viable methods that can simultaneously assess external kinetics and kinematics are necessary to enhance field-based biomechanical monitoring. The aim of this study was to determine the accuracy and usability of ground reaction force (GRF) profiles estimated from segmental kinematics, measured with OpenCap (a low-cost markerless motion-capture system), during common jumping movements. Full-body segmental kinematics were recorded for fifteen recreational athletes performing countermovement, squat, bilateral drop, and unilateral drop jumps, and used to estimate vertical GRFs with a mechanics-based method. Eleven distinct performance-, fatigue-, or injury-related GRF variables were then validated against a gold-standard force platform. Across jumping movements, a total of six and three GRF variables were estimated with a bias or limits of agreement <5 % respectively. Bias and limits of agreement were between 5 and 15 % for seventeen and nineteen variables respectively. Moreover, we show that estimated force variables with a bias <15 % can adequately assess the within-athlete changes in GRF variables between jumping conditions (arm swing or leg dominance). These findings indicate that using a low-cost and field-viable markerless motion capture system (OpenCap) to estimate and assess GRF profiles during common jumping movements is approaching acceptable limits of accuracy. The presented method can be used to monitor force variables of interest and examine underlying segmental kinematics. This application is a jump towards researchers and sports practitioners performing biomechanical monitoring of jumping efficiently, regularly, and extensively in field settings.

Impact of PCA Pre-Normalization Methods on Ground Reaction Force Estimation Accuracy.

Ground reaction force (GRF) components can be estimated using insole pressure sensors. Principal component analysis in conjunction with machine learning (PCA-ML) methods are widely used for this task. PCA reduces dimensionality and requires pre-normalization. In this paper, we evaluated the impact of twelve pre-normalization methods using three PCA-ML methods on the accuracy of GRF component estimation. Accuracy was assessed using laboratory data from gold-standard force plate measurements. Data were collected from nine subjects during slow- and normal-speed walking activities. We tested the ANN (artificial neural network) and LS (least square) methods while also exploring support vector regression (SVR), a method not previously examined in the literature, to the best of our knowledge. In the context of our work, our results suggest that the same normalization method can produce the worst or the best accuracy results, depending on the ML method. For example, the body weight normalization method yields good results for PCA-ANN but the worst performance for PCA-SVR. For PCA-ANN and PCA-LS, the vector standardization normalization method is recommended. For PCA-SVR, the mean method is recommended. The final message is not to define a normalization method a priori independently of the ML method.

An IMU-Based Ground Reaction Force Estimation Method and Its Application in Walking Balance Assessment.

Walking is one of the most common daily movements of the human body. Therefore, quantitative evaluation of human walking has been commonly used to assist doctors in grasping the disease degree and rehabilitation process of patients in the clinic. Compared with the kinematic characteristics, the ground reaction force (GRF) during walking can directly reflect the dynamic characteristics of human walking. It can further help doctors understand the degree of muscle recovery and joint coordination of patients. This paper proposes a GRF estimation method based on the elastic elements and Newton-Euler equation hybrid driving GRF estimation method. Compared with the existing research, the innovations are as follows. 1) The hardware system consists of only two inertial measurement units (IMUs) placed on shanks. The acquisition of the overall motion characteristics of human walking is realized through the simplified four-link walking model and the thigh prediction method. 2) The method was validated not only on 10 healthy subjects but also on 11 Parkinson's patients and 10 stroke patients with normalized mean absolute errors (NMAEs) of 5.95%±1.32%, 6.09%±2.00%, 5.87%±1.59%. 3) This paper proposes a dynamic balance assessment method based on the acquired motion data and the estimated GRF. It evaluates the overall balance ability and fall risk at four key time points for all subjects recruited. Because of the low-cost system, ease of use, low motion interference and environmental constraints, and high estimation accuracy, the proposed GRF estimation method and walking balance automatic assessment have broad clinical value.

Artificial neural network-based ground reaction force estimation and learning for dynamic-legged robot systems.

Legged robots have become popular in recent years due to their ability to locomote on rough terrains; these robots are able to walk on narrow stepping-stones, go upstairs, and explore soft ground such as sand. Ground reaction force (GRF) is the force exerted on the body by the ground when they are in contact. This is a key element and is widely used for programming the locomotion of the legged robots. Being capable of estimating the GRF is advantageous over measuring it with the actual sensor system. Estimating allows one to simplify the system, and it is meant to be capable of prediction, and so on. In this article, we present a neural network approach for GRF estimation for the legged robot system. In order to fundamentally study the GRF estimation of the robot leg, we demonstrate our approach for a single-legged robot with a degree of freedom (DoF) of two with hip and knee joints on a flat-surface. The first joint is directly driven from the actuator, and another joint is belt-pulley driven from the second actuator to take advantage of the long range of motion. The neural network is designed to estimate GRF without attaching force sensors such as load cells, and the encoder is the only sensor used for the estimation. We propose a two-staged multi-layer perceptron (MLP) solution based on supervised learning to estimate GRF in the physical-world. The first stage of the MLP model is trained using datasets from the simulation, enabling it to estimate the simulation-staged GRF. The second stage of the MLP model is trained in the physical world using the simulation-staged GRF obtained from the first stage MLP as the input. This approach enables the second stage MLP to bridge the simulation to the physical world. The root mean squared error (RMSE) is 0.9949 N on the validation datasets in the best case. The performance of the trained network is evaluated when the robot follows trajectories that are not used in training the two-stage GRF estimation network.

Convolutional neural network based ground reaction forces and center of pressure estimation during stair walking using multi-level kinematics

Ground reaction forces (GRFs) and center of pressure trajectories (CoPs) are required for a comprehensive biomechanical analysis. They are also important outcome measures in sports sciences or clinical areas. GRFs and CoPs are usually measured by force plates, which are rarely equipped on staircases in laboratories. We present a one-dimensional convolutional neural network for estimating GRFs and CoPs during stair ascent and descent using multi-level kinematics as input. We collected a dataset of 3782 trials from 172 subjects for training and validating this model. The recruited subjects include healthy subjects and individuals with knee osteoarthritis or moderate-to-high risk of cardiovascular diseases. Our model achieves the state-of-the-art estimating performance with nRMSE of 2.755%∼7.633%, Pearson correlation of 0.950∼0.996 on GRFs estimation, and with nRMSE of 5.519%∼14.669%, Pearson correlation of 0.918∼0.991 on CoPs estimation. With our proposed model, GRFs and CoPs during stair walking can be estimated without force-plate-embedded staircases.

LSTM Network-Based Estimation of Ground Reaction Forces During Walking in Stroke Patients Using Markerless Motion Capture System

LSTM Network-Based Estimation of Ground Reaction Forces During Walking in Stroke Patients Using Markerless Motion Capture System

IMU-based ground reaction force estimation using OpenSim Moco

IMU-based ground reaction force estimation using OpenSim Moco

Predicting vertical and shear ground reaction forces during walking and jogging using wearable plantar pressure insoles

BackgroundThe development of plantar pressure insoles has made them a potential replacement for force plates. These wearable devices can measure multiple steps and might be used outside of the lab environment for rehabilitation and evaluation of sport performance. However, they can only measure the vertical force which does not completely represent the vertical ground reaction force. In addition, they are not able to measure shear forces which play an import role in the dynamic performance of individuals. Indirect approaches might be implemented to improve the accuracy of the force estimated by plantar pressure systems. Research questionThe aim of this study was to predict the vertical and shear components of ground reaction force from plantar pressure data using recurrent neural networks. MethodsGround reaction force and plantar pressure data were collected from 16 healthy individuals during 10 trials of walking and five trials of jogging using Bertec force plates at 1000 Hz and FScan plantar pressure insoles at 100 Hz. A long short-term memory neural network was built to consider the time dependency of pressure and force data in predictions. The data were split into three subsets of train, to train the model, evaluate, to optimize the model hyperparameters, and test sets, to assess the accuracy of the model predictions. ResultsThe results of this study showed that our long short-term memory model could accurately predict the shear and vertical force components during walking and jogging. The predictions were more accurate during walking compared to jogging. In addition, the predictions of mediolateral force had higher error and lower correlation compared to vertical and anteroposterior components. SignificanceThe long short-term memory model developed in this study may be an acceptable option for accurate estimation of ground reaction force during outdoor activities which can have significant impacts in rehabilitation, sport performance, and gaming.

A Deep Learning Model for 3D Ground Reaction Force Estimation Using Shoes with Three Uniaxial Load Cells

Ground reaction force (GRF) is essential for estimating muscle strength and joint torque in inverse dynamic analysis. Typically, it is measured using a force plate. However, force plates have spatial limitations, and studies of gaits involve numerous steps and thus require a large number of force plates, which is disadvantageous. To overcome these challenges, we developed a deep learning model for estimating three-axis GRF utilizing shoes with three uniaxial load cells. GRF data were collected from 81 people as they walked on two force plates while wearing shoes with three load cells. The three-axis GRF was calculated using a seq2seq approach based on long short-term memory (LSTM). To conduct the learning, validation, and testing, random selection was performed based on the subjects. The 60 selected participants were divided as follows: 37 were in the training set, 12 were in the validation set, and 11 were in the test set. The estimated GRF matched the force plate-measured GRF with correlation coefficients of 0.97, 0.96, and 0.90 and root mean square errors of 65.12 N, 15.50 N, and 9.83 N for the vertical, anterior-posterior, and medial-lateral directions, respectively, and there was a mid-stance timing error of 5.61% in the test dataset. A Bland-Altman analysis showed good agreement for the maximum vertical GRF. The proposed shoe with three uniaxial load cells and seq2seq LSTM can be utilized for estimating the 3D GRF in an outdoor environment with level ground and/or for gait research in which the subject takes several steps at their preferred walking speed, and hence can supply crucial data for a basic inverse dynamic analysis.

Effect of IMU location on estimation of vertical ground reaction force during jumping.

Introduction: Several investigations have examined utilizing inertial measurement units (IMU) to estimate ground reaction force (GRF) during exercise. The purpose of this investigation was to determine the effect of inertial measurement units location on the estimation of ground reaction force during vertical jumping. Methods: Eight male subjects completed a series of ten countermovement jumps on a force plate (FP). The subjects had an inertial measurement units attached to the sacrum, back and chest. Ground reaction force was estimated from data from the individual inertial measurement units and by using a two-segment model and combined sensor approach. Results: The peak ground reaction force values for the sacrum, back, chest and combined inertial measurement units were 1,792 ± 278N, 1,850 ± 341N, 2,054 ± 346N and 1,812 ± 323N, respectively. The sacral inertial measurement units achieved the smallest differences for ground reaction force estimates providing a root mean square error (RMSE) between 88N and 360N. The inertial measurement units on the sacrum also showed significant correlations in peak ground reaction force (p < 0.001) and average ground reaction force (p < 0.001) using the Bland-Altman 95% Limits of Agreement (LOA) when in comparison to the force plate. Discussion: Based on assessment of bias, Limits of Agreement, and RMSE, the inertial measurement units located on the sacrum appears to be the best placement to estimate both peak and average ground reaction force during jumping.

Estimation of gait events and kinetic waveforms with wearable sensors and machine learning when running in an unconstrained environment

Wearable sensors and machine learning algorithms are becoming a viable alternative for biomechanical analysis outside of the laboratory. The purpose of this work was to estimate gait events from inertial measurement units (IMUs) and utilize machine learning for the estimation of ground reaction force (GRF) waveforms. Sixteen healthy runners were recruited for this study, with varied running experience. Force sensing insoles were used to measure normal foot-shoe forces, providing a proxy for vertical GRF and a standard for the identification of gait events. Three IMUs were mounted on each participant, two bilaterally on the dorsal aspect of each foot and one clipped to the back of each participant’s waistband, approximating their sacrum. Participants also wore a GPS watch to record elevation and velocity. A Bidirectional Long Short Term Memory Network (BD-LSTM) was used to estimate GRF waveforms from inertial waveforms. Gait event estimation from both IMU data and machine learning algorithms led to accurate estimations of contact time. The GRF magnitudes were generally underestimated by the machine learning algorithm when presented with data from a novel participant, especially at faster running speeds. This work demonstrated that estimation of GRF waveforms is feasible across a range of running velocities and at different grades in an uncontrolled environment.

Ground Reaction Force and Moment Estimation through EMG Sensing Using Long Short-Term Memory Network during Posture Coordination

Motion prediction based on kinematic information such as body segment displacement and joint angle has been widely studied. Because motions originate from forces, it is beneficial to estimate dynamic information, such as the ground reaction force (GRF), in addition to kinematic information for advanced motion prediction. In this study, we proposed a method to estimate GRF and ground reaction moment (GRM) from electromyography (EMG) in combination with and without an inertial measurement unit (IMU) sensor using a machine learning technique. A long short-term memory network, which is suitable for processing long time-span data, was constructed with EMG and IMU as input data to estimate GRF during posture control and stepping motion. The results demonstrate that the proposed method can provide the GRF estimation with a root mean square error (RMSE) of 8.22 ± 0.97% (mean ± SE) for the posture control motion and 11.17 ± 2.16% (mean ± SE) for the stepping motion. We could confirm that EMG input is essential especially when we need to predict both GRF and GRM with limited numbers of sensors attached under knees. In addition, we developed a GRF visualization system integrated with ongoing motion in a Unity environment. This system enabled the visualization of the GRF vector in 3-dimensional space and provides predictive motion direction based on the estimated GRF, which can be useful for human motion prediction with portable sensors.

Estimating Ground Reaction Forces from Two-Dimensional Pose Data: A Biomechanics-Based Comparison of AlphaPose, BlazePose, and OpenPose.

The adoption of computer vision pose estimation approaches, used to identify keypoint locations which are intended to reflect the necessary anatomical landmarks relied upon by biomechanists for musculoskeletal modelling, has gained increasing traction in recent years. This uptake has been further accelerated by keypoint use as inputs into machine learning models used to estimate biomechanical parameters such as ground reaction forces (GRFs) in the absence of instrumentation required for direct measurement. This study first aimed to investigate the keypoint detection rate of three open-source pose estimation models (AlphaPose, BlazePose, and OpenPose) across varying movements, camera views, and trial lengths. Second, this study aimed to assess the suitability and interchangeability of keypoints detected by each pose estimation model when used as inputs into machine learning models for the estimation of GRFs. The keypoint detection rate of BlazePose was distinctly lower than that of AlphaPose and OpenPose. All pose estimation models achieved a high keypoint detection rate at the centre of an image frame and a lower detection rate in the true sagittal plane camera field of view, compared with slightly anteriorly or posteriorly located quasi-sagittal plane camera views. The three-dimensional ground reaction force, instantaneous loading rate, and peak force for running could be estimated using the keypoints of all three pose estimation models. However, only AlphaPose and OpenPose keypoints could be used interchangeably with a machine learning model trained to estimate GRFs based on AlphaPose keypoints resulting in a high estimation accuracy when OpenPose keypoints were used as inputs and vice versa. The findings of this study highlight the need for further evaluation of computer vision-based pose estimation models for application in biomechanical human modelling, and the limitations of machine learning-based GRF estimation models that rely on 2D keypoints. This is of particular relevance given that machine learning models informing athlete monitoring guidelines are being developed for application related to athlete well-being.

UnderPressure: Deep Learning for Foot Contact Detection, Ground Reaction Force Estimation and Footskate Cleanup

AbstractHuman motion synthesis and editing are essential to many applications like video games, virtual reality, and film post‐production. However, they often introduce artefacts in motion capture data, which can be detrimental to the perceived realism. In particular, footskating is a frequent and disturbing artefact, which requires knowledge of foot contacts to be cleaned up. Current approaches to obtain foot contact labels rely either on unreliable threshold‐based heuristics or on tedious manual annotation. In this article, we address automatic foot contact label detection from motion capture data with a deep learning based method. To this end, we first publicly release Under Pressure, a novel motion capture database labelled with pressure insoles data serving as reliable knowledge of foot contact with the ground. Then, we design and train a deep neural network to estimate ground reaction forces exerted on the feet from motion data and then derive accurate foot contact labels. The evaluation of our model shows that we significantly outperform heuristic approaches based on height and velocity thresholds and that our approach is much more robust when applied on motion sequences suffering from perturbations like noise or footskate. We further propose a fully automatic workflow for footskate cleanup: foot contact labels are first derived from estimated ground reaction forces. Then, footskate is removed by solving foot constraints through an optimisation‐based inverse kinematics (IK) approach that ensures consistency with the estimated ground reaction forces. Beyond footskate cleanup, both the database and the method we propose could help to improve many approaches based on foot contact labels or ground reaction forces, including inverse dynamics problems like motion reconstruction and learning of deep motion models in motion synthesis or character animation. Our implementation, pre‐trained model as well as links to database can be found at github.com/InterDigitalInc/UnderPressure.