No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
Estimation of Limbs Angles Amplitudes During the Use of the Five Minute Shaper Device Using Artificial Neural Networks
Abstract
Biomechanical studies are essential in health research areas, such as rehabilitation, kinesiology, orthopedics, and sports. For example, they provide information to elaborate on patients’ diagnostics or improve athletes’ performance. In recent years, deep learning and other computational methods have started to be used to quantify new biomechanical parameters or perform deeper data analysis. Motion capture is one of the methods commonly used in biomechanical studies. For this method, video-based and marker-based systems are the gold standards; nevertheless, those systems are typically quite expensive. Moreover, experimental errors in data capture are frequently related to the occlusion of the markers during motion capture. Data missed is solved by increasing the number of cameras to cover more angles or by using predetermined interpolation algorithms. However, the last method could fail to predict all the marker data missed, and both options increase the cost of the data analysis. For solving those kinds of problems, novel computational methods could be used. This study aims to implement an artificial neural network (ANN) to estimate the limb angle amplitude during the execution of a movement from a single axis (X-axis). For training and validating the ANN model, the data and features from the Five-Minute Shaper machine (a physical conditioning device) are used. The obtained results include RMSE values smaller than 3.2 (Minimum RMSE of 0.96) and CC values close to 0.99. The predicted values are very close to the experimental amplitude angles, and, according to the Two-sample Kolmogorov-Smirnov test, the experimental and the estimated amplitude angles follow the same continuous distribution (p-value > 0.05). It is shown that these methods could help researchers in biomechanics to perform accurate analysis, reducing the number of needed cameras and avoid problems due to occlusion by only needing information from a specific axis.
In biomechanics, motion capture systems based on video and markers are the most widely used method to estimate kinematic parameters. However, from a technical standpoint, experimental errors in data capture are often related to the masking of markers during motion capture. This phenomenon generates data loss that can affect the analysis of the results. The lack of data is solved by increasing the number of cameras or using additional devices such as inertial sensors. However, those additions increase the experimental cost of this method. Nowadays, new computational methods can be used to solve such problems less expensively. This study implemented two computational methods based on Artificial Neural Networks (ANNs) and Support Vector Regression (SVR) to estimate the amplitude of limb angles during the execution of a movement on a single axis (i.e., the z-axis). The characteristics of the squats were used to train and validate the models. The results obtained include RMSE values lower than 14 (minimum RMSE of 5.35) and CC values close to 0.98. The estimated values are very close to the experimental amplitude angles, and the statistical analyses showed no significant differences between the distributions and means of the estimated amplitude values and their actual counterparts (p-value>0.05). The results show that these methods could help biomechanics researchers perform accurate analyses, decrease the number of cameras needed, reduce uncertainty, and avoid data loss problems.
The purpose of this research was to continuous knee joint angle estimation from sEMG during squat using artificial neural networks. sEMG signals of vastus medialis, rectus femoris, biceps femoris and 3D kinematics of lower extremity joints for four participants during squat were captured at 1500 Hz and 100 Hz, respectively. sEMG signals were preprocessed and RMS and variance were extracted as input features. The processed input data was given to a three-layer feed forward neural network with one hidden layer. The proposed network was trained by the Levenberg-Marquardt algorithm. The root mean square error (RMSE) and correlation coefficient (CC) were used to evaluate the accuracy of estimation. The results showed that this network is able to continuously estimate the knee joint angle with global RMSE of 5.0041° ± 0.9963° and CC of 0.9898 ± 0.0039. It concludes that a multilayer neural network with a simple structure has the ability to continuously estimate the joint angle from sEMG data while performing an athletic movement under real loading situation.
There is a need within human movement sciences for a markerless motion capture system, which is easy to use and sufficiently accurate to evaluate motor performance. This study aims to develop a 3D markerless motion capture technique, using OpenPose with multiple synchronized video cameras, and examine its accuracy in comparison with optical marker-based motion capture. Participants performed three motor tasks (walking, countermovement jumping, and ball throwing), and these movements measured using both marker-based optical motion capture and OpenPose-based markerless motion capture. The differences in corresponding joint positions, estimated from the two different methods throughout the analysis, were presented as a mean absolute error (MAE). The results demonstrated that, qualitatively, 3D pose estimation using markerless motion capture could correctly reproduce the movements of participants. Quantitatively, of all the mean absolute errors calculated, approximately 47% were <20 mm, and 80% were <30 mm. However, 10% were >40 mm. The primary reason for mean absolute errors exceeding 40 mm was that OpenPose failed to track the participant's pose in 2D images owing to failures, such as recognition of an object as a human body segment or replacing one segment with another depending on the image of each frame. In conclusion, this study demonstrates that, if an algorithm that corrects all apparently wrong tracking can be incorporated into the system, OpenPose-based markerless motion capture can be used for human movement science with an accuracy of 30 mm or less.
Biomechanical analyses provide an extensive source of data that are deeply explored by physicians, engineers and trainers from the mechanical and physiological point of view. This data includes kinetic and kinematic parameters that are quite useful to study human locomotion. However, most of these analyses stay on a very superficial level. Recently data and computational science expanded their coverage to new areas and new analysis tools are available. These analyses include the use of machine learning tools for data mining processes. All of these new tools open a total new level of data analysis, thus newer and deeper questions are proposed in order to provide more accurate prediction results with strict decision support. On the other hand, Squat is an exercise widely used for physical conditioning since it puts into operation various muscles at the same time of the lower and upper train. However bad squatting could drive to injuries at the back and knee level. These injuries are especially common in patients without physical conditioning. In this study, squat data is analyzed using Self-Organizing Maps (SOM) to identify possible relevant parameters from the subjects that could affect the movement performance especially at the knee joint.
With the current employment of visualization tools, different techniques may be used to understand multivariable information. This document presents an alternative method for the analysis of biomechanical phenomena, presenting the knee joint execution of free squat exercise. For this, the MaxTRAQ software and the open-source and cross-platform software Mokka were employed. A system of cameras allowed to digitize the movements. Then, with the use of Self-Organizing Maps tasks of visualization and clustering were developed for multivariate data. Results showed that this tool allows finding nonlinear relations between the analyzed input variables and outputs, when the data treatment demands an extra-work because of its dimensionality.
Health care organizations are leveraging machine-learning techniques, such as artificial neural networks (ANN), to improve delivery of care at a reduced cost. Applications of ANN to diagnosis are well-known; however, ANN are increasingly used to inform health care management decisions. We provide a seminal review of the applications of ANN to health care organizational decision-making. We screened 3,397 articles from six databases with coverage of Health Administration, Computer Science and Business Administration. We extracted study characteristics, aim, methodology and context (including level of analysis) from 80 articles meeting inclusion criteria. Articles were published from 1997–2018 and originated from 24 countries, with a plurality of papers (26 articles) published by authors from the United States. Types of ANN used included ANN (36 articles), feed-forward networks (25 articles), or hybrid models (23 articles); reported accuracy varied from 50% to 100%. The majority of ANN informed decision-making at the micro level (61 articles), between patients and health care providers. Fewer ANN were deployed for intra-organizational (meso- level, 29 articles) and system, policy or inter-organizational (macro- level, 10 articles) decision-making. Our review identifies key characteristics and drivers for market uptake of ANN for health care organizational decision-making to guide further adoption of this technique.
Epilepsy, defined as recurrent epileptic seizures that are unprovoked by any immediate identified cause, affects up to 1% of the population worldwide and comes second to stroke as a serious neurological disorder. The total burden due to epilepsy is around 0.5% of the global burden of disease, with approximately 1% of the total days lost due to disease. The signs and symptoms of epilepsy vary due to the involvement of several areas of the cortex and underlying brain systems. One of the most devastating features of epilepsy is the apparently unpredictable nature of the seizures. However, recent literature supports the notion that seizures begin before the onset of clinical signs and symptoms; thus early and accurate detection is important to improve the quality of life of epileptic sufferers. Therefore, the real challenge is in the development of highly reliable methods for seizure onset detection prior to clinical syndromes. Amplitude-integrated electroencephalography (aEEG) is one technique that is useful to monitor brain activity; however, the distinction of seizure activity of aEEG requires a specialist and may not have high specificity and sensitivity. As the difficulty in interpreting aEEG in clinical practice relates mainly to the distinction between artifact and activity, several mathematical algorithms were incorporated for its analysis. The current chapter will emphasize developing algorithmic methods for analyzing epileptic data and extracting specific quantifiers that can be used for its prediction.
The purpose of this study was to develop and train a Neural Network (NN) that uses barbell mass and motions to predict hip, knee, and ankle Net Joint Moments (NJM) during a weightlifting exercise. Seven weightlifters performed two cleans at 85% of their competition maximum while ground reaction forces and 3-D motion data were recorded. An inverse dynamics procedure was used to calculate hip, knee, and ankle NJM. Vertical and horizontal barbell motion data were extracted and, along with barbell mass, used as inputs to a NN. The NN was then trained to model the association between the mass and kinematics of the barbell and the calculated NJM for six weightlifters, the data from the remaining weightlifter was then used to test the performance of the NN - this was repeated 7 times with a k-fold cross-validation procedure to assess the NN accuracy. Joint-specific predictions of NJM produced coefficients of determination (r2) that ranged from 0.79 to 0.95, and the percent difference between NN-predicted and inverse dynamics calculated peak NJM ranged between 5% and 16%. The NN was thus able to predict the spatiotemporal patterns and discrete peaks of the three NJM with reasonable accuracy, which suggests that it is feasible to predict lower extremity NJM from the mass and kinematics of the barbell. Future work is needed to determine whether combining a NN model with low cost technology (e.g., digital video and free digitising software) can also be used to predict NJM of weightlifters during field-testing situations, such as practice and competition, with comparable accuracy.
Problem addressed:
Wrist-worn accelerometers are associated with greater compliance. However, validated algorithms for predicting activity type from wrist-worn accelerometer data are lacking. This study compared the activity recognition rates of an activity classifier trained on acceleration signal collected on the wrist and hip.
Methodology:
52 children and adolescents (mean age 13.7 ± 3.1 year) completed 12 activity trials that were categorized into 7 activity classes: lying down, sitting, standing, walking, running, basketball, and dancing. During each trial, participants wore an ActiGraph GT3X+ tri-axial accelerometer on the right hip and the non-dominant wrist. Features were extracted from 10-s windows and inputted into a regularized logistic regression model using R (Glmnet + L1).
Results:
Classification accuracy for the hip and wrist was 91.0% ± 3.1% and 88.4% ± 3.0%, respectively. The hip model exhibited excellent classification accuracy for sitting (91.3%), standing (95.8%), walking (95.8%), and running (96.8%); acceptable classification accuracy for lying down (88.3%) and basketball (81.9%); and modest accuracy for dance (64.1%). The wrist model exhibited excellent classification accuracy for sitting (93.0%), standing (91.7%), and walking (95.8%); acceptable classification accuracy for basketball (86.0%); and modest accuracy for running (78.8%), lying down (74.6%) and dance (69.4%).Potential Impact: Both the hip and wrist algorithms achieved acceptable classification accuracy, allowing researchers to use either placement for activity recognition.
[Purpose] The present study longitudinally investigated injury occurrences and the risk factors for muscle tightness, core stability, and dynamic standing balance among junior high school student baseball players. [Subjects] Thirty-nine male students, belonging to baseball clubs at 2 junior high schools, participated in this study. [Methods] Study measurements were obtained twice, once in the early stage of the baseball season (March) and once at the end of the season (July). All subjects underwent muscle tightness testing, the Star Excursion Balance Test (SEBT), and trunk endurance testing during each measurement session. [Results] Fifteen players experienced episodes of elbow or shoulder pain while throwing. Players in the pain group demonstrated a significant increase in the tightness of their shoulder internal rotators, axis-leg quadriceps, and axis-leg hamstrings. There was no clear evidence of differences of changes in core stability and dynamic standing balance between the groups. [Conclusion] The results of this study suggest that lower extremity muscle tightness early in a season and the subsequent decline in the flexibility of the axis-leg quadriceps and hamstrings during the season may be due to an increased upper extremity load while throwing, thus producing shoulder and elbow pain.
Mechanical analysis of movement plays an important role in clinical management of neurological and orthopedic conditions. There has been increasing interest in performing movement analysis in real-time, to provide immediate feedback to both therapist and patient. However, such work to date has been limited to single-joint kinematics and kinetics. Here we present a software system, named human body model (HBM), to compute joint kinematics and kinetics for a full body model with 44 degrees of freedom, in real-time, and to estimate length changes and forces in 300 muscle elements. HBM was used to analyze lower extremity function during gait in 12 able-bodied subjects. Processing speed exceeded 120 samples per second on standard PC hardware. Joint angles and moments were consistent within the group, and consistent with other studies in the literature. Estimated muscle force patterns were consistent among subjects and agreed qualitatively with electromyography, to the extent that can be expected from a biomechanical model. The real-time analysis was integrated into the D-Flow system for development of custom real-time feedback applications and into the gait real-time analysis interactive lab system for gait analysis and gait retraining.
Electronic supplementary material
The online version of this article (doi:10.1007/s11517-013-1076-z) contains supplementary material, which is available to authorized users.
All life forms on earth, including humans, are constantly subjected to the universal force of gravitation, and thus to forces from within and surrounding the body. Through the study of the interaction of these forces and their effects, the form, function and motion of our bodies can be examined and the resulting knowledge applied to promote quality of life. Under gravity and other loads, and controlled by the nervous system, human movement is achieved through a complex and highly coordinated mechanical interaction between bones, muscles, ligaments and joints within the musculoskeletal system. Any injury to, or lesion in, any of the individual elements of the musculoskeletal system will change the mechanical interaction and cause degradation, instability or disability of movement. On the other hand, proper modification, manipulation and control of the mechanical environment can help prevent injury, correct abnormality, and speed healing and rehabilitation. Therefore, understanding the biomechanics and loading of each element during movement using motion analysis is helpful for studying disease etiology, making decisions about treatment, and evaluating treatment effects. In this article, the history and methodology of human movement biomechanics, and the theoretical and experimental methods developed for the study of human movement, are reviewed. Examples of motion analysis of various patient groups, prostheses and orthoses, and sports and exercises, are used to demonstrate the use of biomechanical and stereophotogrammetry-based human motion analysis studies to address clinical issues. It is suggested that further study of the biomechanics of human movement and its clinical applications will benefit from the integration of existing engineering techniques and the continuing development of new technology.
Video analysis is used in sport to derive kinematic variables of interest but often relies on time-consuming tracking operations. The purpose of this study was to determine speed, accuracy and reliability of 2D body landmark digitisation by a neural network (NN), compared with manual digitisation, for the glide phase in swimming. Glide variables including glide factor; instantaneous hip angles, trunk inclines and horizontal velocities were selected as they influence performance and are susceptible to digitisation propagation error. The NN was ‘trained’ on 400 frames of 2D glide video from a sample of eight elite swimmers. Four glide trials of another swimmer were used to test agreement between the NN and a manual operator for body marker position data of the knee, hip and shoulder, and the effect of digitisation on glide variables. The NN digitised body landmarks 233 times faster than the manual operator, with digitising root-mean-square-error of ~4-5mm. High accuracy and reliability was found between body position and glide variable data between the two methods with relative error ≤5.4% and correlation coefficients >0.95 for all variables. NNs could be applied to greatly reduce the time of kinematic analysis in sports and facilitate rapid feedback of performance measures.
Spinal posture is a crucial input in biomechanical models and an essential factor in ergonomics investigations to evaluate risk of low back injury. In vivo measurement of spinal posture through the common motion capture techniques is limited to equipped laboratories and thus impractical for workplace applications. Posture prediction models are therefore considered indispensable tools. This study aims to investigate the capability of artificial neural networks (ANNs) in predicting the three-dimensional posture of the spine (S1, T12 and T1 orientations) in various activities. Two ANNs were trained and tested using measurements from spinal postures of 40 male subjects by an inertial tracking device in various static reaching and lifting (of 5 kg) activities. Inputs of each ANN were position of the hand load and body height, while outputs were rotations of the three foregoing segments relative to their initial orientation in the neutral upright posture. Effect of posture prediction errors on the estimated spinal loads in symmetric reaching activities was also investigated using a biomechanical model. Results indicated that both trained ANNs could generate outputs (three-dimensional orientations of the segments) from novel sets of inputs that were not included in the training processes (root-mean-squared-error (RMSE) < 11 degrees and coefficient-of-determination (R-2) > 0.95). A graphic user interface was designed and made available to facilitate use of the ANNs. The difference between the mean of each measured angle in a reaching task and the corresponding angle in a lifting task remained smaller than 8 degrees. Spinal loads estimated by the biomechanical model based on the predicted postures were on average different by < 12% from those estimated based on the exact measured postures (RMSE=173 and 35 N for the L5-S1 compression and shear loads, respectively).
This article reviews developments in the use of Artificial Intelligence (AI) in sports biomechanics over the last decade. It outlines possible uses of Expert Systems as diagnostic tools for evaluating faults in sports movements ('techniques') and presents some example knowledge rules for such an expert system. It then compares the analysis of sports techniques, in which Expert Systems have found little place to date, with gait analysis, in which they are routinely used. Consideration is then given to the use of Artificial Neural Networks (ANNs) in sports biomechanics, focusing on Kohonen self-organizing maps, which have been the most widely used in technique analysis, and multi-layer networks, which have been far more widely used in biomechanics in general. Examples of the use of ANNs in sports biomechanics are presented for javelin and discus throwing, shot putting and football kicking. I also present an example of the use of Evolutionary Computation in movement optimization in the soccer throw in, which predicted an optimal technique close to that in the coaching literature. After briefly overviewing the use of AI in both sports science and biomechanics in general, the article concludes with some speculations about future uses of AI in sports biomechanics. Key PointsExpert Systems remain almost unused in sports biomechanics, unlike in the similar discipline of gait analysis.Artificial Neural Networks, particularly Kohonen Maps, have been used, although their full value remains unclear.Other AI applications, including Evolutionary Computation, have received little attention.
This paper describes a novel approach towards recognizing of Indian Sign Language (ISL) gestures for Humanoid Robot Interaction (HRI). An extensive approach is being introduced for classification of ISL gesture which imparts an elegant way of interaction between humanoid robot HOAP-2 and human being. ISL gestures are being considered as a communicating agent for humanoid robot which is being used in this context explicitly. It involves different image processing techniques followed by a generic algorithm for feature extraction process. The classification technique deals with the Euclidean distance metric. The concrete HRI system has been established for initiation based learning mechanism. The Real time robotics simulation software, WEBOTS has been adopted to simulate the classified ISL gestures on HOAP-2 robot. The JAVA based software has been developed to deal with the entire HRI process.
Roentgen stereophotogrammetric analysis (RSA) measures micromotion of an orthopaedic implant with respect to its surrounding bone. A problem in RSA is that the markers are sometimes overprojected by the implant itself. This study describes the so-called Marker Configuration Model-based RSA (MCM-based RSA) that is able to measure the pose of a rigid body in situations where less than three markers could be detected in both images of an RSA radiograph. MCM-based RSA is based on fitting a Marker Configuration model (MC-model) to the projection lines from the marker projection positions in the image to their corresponding Roentgen foci. An MC-model describes the positions of markers relative to each other and is obtained using conventional RSA. We used data from 15 double examinations of a clinical study of total knee prostheses and removed projections of the three tibial component markers, simulating occlusion of markers. The migration of the tibial component with respect to the bone, which should be zero, for the double examination is a measure of the accuracy of algorithm. With the new algorithm, it is possible to estimate the pose of a rigid body of which one or two markers are occluded in one of the images of the RSA radiograph with high accuracy as long as a proper MC-model of the markers in the rigid body is available. The new algorithm makes RSA more robust for occlusion of markers. This improves the results of clinical RSA studies because the number of lost RSA follow-up moments is reduced.
Kinematic study of video gait analysis
- T S Ashok
Ashok, T.S., et al.: Kinematic study of video gait analysis. In: 2015 International
Conference on Industrial Instrumentation and Control (ICIC), pp. 1208-1213.
IEEE (2015)