Are Action Sport Cameras Accurate Enough for 3D Motion Analysis? A Comparison With a Commercial Motion Capture System

BlazePose-Seq2Seq: Leveraging Regular RGB Cameras for Robust Gait Assessment

Evaluation of human gait through smartphone-based pose estimation algorithms provides an attractive alternative to costly lab-bound instrumented assessment and offers a paradigm shift with real time gait capture for clinical assessment. Systems based on smart phones, such as OpenPose and BlazePose have demonstrated potential for virtual motion assessment but still lack the accuracy and repeatability standards required for clinical viability. Seq2seq architecture offers an alternative solution to conventional deep learning techniques for predicting joint kinematics during gait. This study introduces a novel enhancement to the low-powered BlazePose algorithm by incorporating a Seq2seq autoencoder deep learning model. To ensure data accuracy and reliability, synchronized motion capture involving an RGB camera and ten Vicon cameras were employed across three distinct self-selected walking speeds. This investigation presents a groundbreaking avenue for remote gait assessment, harnessing the potential of Seq2seq architectures inspired by natural language processing (NLP) to enhance pose estimation accuracy. When comparing BlazePose alone to the combination of BlazePose and 1D convolution Long Short-term Memory Network (1D-LSTM), Gated Recurrent Unit (GRU) and Long Short-Term Memory (LSTM), the average mean absolute errors decreased from 13.4° to 5.3° for fast gait, from 16.3° to 7.5° for normal gait, and from 15.5° to 7.5° for slow gait at the left ankle joint angle respectively. The strategic utilization of synchronized data and rigorous testing methodologies further bolsters the robustness and credibility of these findings.

Validation of machine vision and action sport cameras for 3D motion analysis model reconstruction

The study investigated the feasibility of using action sport cameras for motion analysis research. Data acquired from two different marker-based motion capture systems and six different camera combinations were analyzed for motion reconstruction accuracy. Two different calibration procedures were used to determine the influence on marker position reconstruction. Static and dynamic calibration mean merit score differences between the reference and experimental camera systems were 0.4 mm and 1.3 mm, respectively. Angular displacement difference between the reference and experimental camera systems range between 0.1 and 2.0 degrees. A systematic bias (- 0.54 to 0.19 degrees) was determined between the reference and the experimental camera systems for range of motion. The mean of the multi-trial findings suggests the machine vision camera system calibrated with a dynamic procedure generated highly accurate three-dimensional reconstructed ROM data (0.5 degree) followed closely by the four action sport cameras implementing a static calibration procedure (0.5 degree). The overall findings suggest the selected machine vision and action sport camera systems produced comparable results to the reference motion analysis system. However, the combination of camera type, processing software, and calibration procedure can influence motion reconstruction accuracy.

Variability analysis of muscle activation symmetry to identify indicators of individual motor strategy: a case series on elite Paralympic powerlifters

Introduction

In Paralympic powerlifting competitions, movement execution symmetry is a technical requirement influenced by individual athlete characteristics and motor strategies. Identifying the elements associated with individual motor strategies can offer valuable insight for improving sport performance. Therefore, this case series study aimed to explore muscle activation symmetry and its intra- and inter-individual variability to determine the muscles mostly related to individual motor strategies in elite Paralympic powerlifters.

Methods

Bilateral electromyographic activation of the anterior deltoid (AD), pectoralis major (PM), latissimus dorsi (LD), triceps (TRI) and external oblique (EO) muscles were analysed in five elite Paralympic powerlifters while performing four sets of one-repetition maximum of Paralympic bench press. Muscle activation symmetry indexes (SI) were obtained and transformed to consider individual-independent evaluation. The coefficient of variation (CV), variance ratio (VR), and mean deviation (MD) were computed to assess inter- and intra-individual variability in electromyographic waveforms and SI.

Results

Both transformed and non-transformed SI indicated overall symmetric activation in DA, PM, TRI, and LD. Transformed SI revealed asymmetrical muscle activation of EO when grouping data (mean bilateral difference: 10%). Athletes exhibited low intra-individual SI variability in all analysed muscles (CV < 10%) and low inter-individual variability in DA, PM, LD, and TRI (CV < 10%; VR: 4%-11%; MD: 29%-43%). In contrast, higher inter-individual variability was observed in EO (CV: 23%; VR: 23%; MD: 72%-81%).

Conclusion

The highest variability and asymmetry in abdominal muscle activation among athletes emphasize the importance of personalized training approaches for targeting these muscles due to their role in individualizing motor strategies.

The analysis of infrared high-speed motion capture system on motion aesthetics of aerobics athletes under biomechanics analysis

This paper uses an infrared high-speed motion capture system based on deep learning to analyze difficult movements, which helps aerobics athletes master difficult movements more accurately. Firstly, changes in joint angle, speed of movement, and ground pressure are used to analyze the impact and role of motion fluency and completion based on a biomechanical perspective. Moreover, based on the existing infrared high-speed motion capture systems, the Restricted Boltzmann Machine (RBM) model is introduced to construct an unsupervised similarity framework model. Next, the motion data is reorganized based on three-dimensional information to adapt to the model's input. Then, the framework performs similar frame matching to obtain a set of candidate frames that can be used as motion graph nodes. After the infrared high-speed motion capture system and inertial sensors are simultaneously applied to subjects, the multi-correlation coefficients (CMC) values of the hip, knee, and ankle angles are 0.94 ± 0.06, 0.98 ± 0.01, and 0.87 ± 0.09, respectively. The two systems show a high degree of correlation in the measurement results, and the knee joint is the most significant correlation. Finally, a motion graph is constructed to control its trajectory and adjust its motion pattern. The infrared high-speed motion capture system optimized for deep learning can extract features from human bone data and capture motion more accurately, helping trainers to fully understand difficult movements.

Validation of an Automatic Inertial Sensor-Based Methodology for Detailed Barbell Velocity Monitoring during Maximal Paralympic Bench Press

Current technologies based on inertial measurement units (IMUs) are considered valid and reliable tools for monitoring barbell velocity in strength training. However, the extracted outcomes are often limited to a few velocity metrics, such as mean or maximal velocity. This study aimed at validating a single IMU-based methodology to automatically obtain the barbell velocity full profile as well as key performance metrics during maximal Paralympic bench press. Seven Paralympic powerlifters (age: 30.5 ± 4.3 years, sitting height: 71.6 ± 6.8 cm, body mass: 72.5 ± 16.4 kg, one-repetition maximum: 148.4 ± 38.6 kg) performed four attempts of maximal Paralympic bench press. The barbell velocity profile and relevant metrics were automatically obtained from IMU linear acceleration through a custom-made algorithm and validated against a video-based reference system. The mean difference between devices was 0.00 ± 0.04 m·s−1 with low limits of agreement (<0.09 m·s−1) and moderate-to-good reliability (ICC: 0.55−0.90). Linear regression analysis showed large-to-very large associations between paired measurements (r: 0.57−0.91, p < 0.003; SEE: 0.02−0.06 m·s−1). The analysis of velocity curves showed a high spatial similarity and small differences between devices. The proposed methodology provided a good level of agreement, making it suitable for different applications in barbell velocity monitoring during maximal Paralympic bench press.

Pose2Sim: An End-to-End Workflow for 3D Markerless Sports Kinematics—Part 2: Accuracy

Two-dimensional deep-learning pose estimation algorithms can suffer from biases in joint pose localizations, which are reflected in triangulated coordinates, and then in 3D joint angle estimation. Pose2Sim, our robust markerless kinematics workflow, comes with a physically consistent OpenSim skeletal model, meant to mitigate these errors. Its accuracy was concurrently validated against a reference marker-based method. Lower-limb joint angles were estimated over three tasks (walking, running, and cycling) performed multiple times by one participant. When averaged over all joint angles, the coefficient of multiple correlation (CMC) remained above 0.9 in the sagittal plane, except for the hip in running, which suffered from a systematic 15° offset (CMC = 0.65), and for the ankle in cycling, which was partially occluded (CMC = 0.75). When averaged over all joint angles and all degrees of freedom, mean errors were 3.0°, 4.1°, and 4.0°, in walking, running, and cycling, respectively; and range of motion errors were 2.7°, 2.3°, and 4.3°, respectively. Given the magnitude of error traditionally reported in joint angles computed from a marker-based optoelectronic system, Pose2Sim is deemed accurate enough for the analysis of lower-body kinematics in walking, cycling, and running.

Automatic Markerless Motion Detector Method against Traditional Digitisation for 3-Dimensional Movement Kinematic Analysis of Ball Kicking in Soccer Field Context

Kicking is a fundamental skill in soccer that often contributes to match outcomes. Lower limb movement features (e.g., joint position and velocity) are determinants of kick performance. However, obtaining kicking kinematics under field conditions generally requires time-consuming manual tracking. The current study aimed to compare a contemporary markerless automatic motion estimation algorithm (OpenPose) with manual digitisation (DVIDEOW software) in obtaining on-field kicking kinematic parameters. An experimental dataset of under-17 players from all outfield positions was used. Kick attempts were performed in an official pitch against a goalkeeper. Four digital video cameras were used to record full-body motion during support and ball contact phases of each kick. Three-dimensional positions of hip, knee, ankle, toe and foot centre-of-mass (CM_foot) generally showed no significant differences when computed by automatic as compared to manual tracking (whole kicking movement cycle), while only z-coordinates of knee and calcaneus markers at specific points differed between methods. The resulting time-series matrices of positions (r² = 0.94) and velocity signals (r² = 0.68) were largely associated (all p < 0.01). The mean absolute error of OpenPose motion tracking was 3.49 cm for determining positions (ranging from 2.78 cm (CM_foot) to 4.13 cm (dominant hip)) and 1.29 m/s for calculating joint velocity (0.95 m/s (knee) to 1.50 m/s (non-dominant hip)) as compared to reference measures by manual digitisation. Angular range-of-motion showed significant correlations between methods for the ankle (r = 0.59, p < 0.01, large) and knee joint displacements (r = 0.84, p < 0.001, very large) but not in the hip (r = 0.04, p = 0.85, unclear). Markerless motion tracking (OpenPose) can help to successfully obtain some lower limb position, velocity, and joint angular outputs during kicks performed in a naturally occurring environment.

New Procedure for the Kinematic and Power Analysis of Cyclists in Indoor Training

In this research, the performance and movements of amateur and professional cyclists were analyzed. For this, reflective markers have been used on different parts of the body of the participants in conjunction with sports cameras and a mobile power meter. The trajectories of the markers have been obtained with the software Kinovea and subsequently analyzed using error ellipses. It is demonstrated that the error ellipses help determine movement patterns in the knees, back, and hip. The covariance of the error ellipses can be indicative of the alignment and symmetry of the frontal movement of the knees. In addition, it allows verifying the alignment of the spine and the symmetry of the hip. Finally, it is shown that it is necessary to consider the uncertainty of the power devices since it considerably affects the evaluation of the cyclists’ performance. Devices with high uncertainty will demand a greater effort from the cyclist to meet the power required in the endurance test developed. The statistical magnitudes considered help to analyze power and evaluate the cyclists’ performance.

Marker-Based Movement Analysis of Human Body Parts in Therapeutic Procedure

Movement analysis of human body parts is momentous in several applications including clinical diagnosis and rehabilitation programs. The objective of this research is to present a low-cost 3D visual tracking system to analyze the movement of various body parts during therapeutic procedures. Specifically, a marker based motion tracking system is proposed in this paper to capture the movement information in home-based rehabilitation. Different color markers are attached to the desired joints' locations and they are detected and tracked in the video to encode their motion information. The availability of this motion information of different body parts during the therapy can be exploited to achieve more accurate results with better clinical insight, which in turn can help improve the therapeutic decision making. The proposed framework is an automated and inexpensive motion tracking system with execution speed close to real time. The performance of the proposed method is evaluated on a dataset of 10 patients using two challenging matrices that measure the average accuracy by estimating the joints' locations and rotations. The experimental evaluation and its comparison with the existing state-of-the-art techniques reveals the efficiency of the proposed method.

Moving system with action sport cameras: 3D kinematics of the walking and running in a large volume

Traditionally, motion analysis in clinical laboratories using optoelectronic systems (MOCAP) is performed in acquisition volumes of limited size. Given the complexity and cost of MOCAP in larger volumes, action sports cameras (ASC) represent an alternative approach in which the cameras move along with the subject during the movement task. Thus, this study aims to compare ASC against a traditional MOCAP in the perspective of reconstructing walking and running movements in large spatial volumes, which extend over the common laboratory setup. The two systems, consisting of four cameras each, were closely mounted on a custom carrying structure endowed with wheels. Two different acquisition setups, namely steady and moving conditions, were taken into account. A devoted calibration procedure, using the same protocol for the two systems, enabled the reconstruction of surface markers, placed on voluntary subjects, during the two acquisition setups. The comparison was quantitatively expressed in terms of three-dimensional (3D) marker reconstruction and kinematic computation quality. The quality of the marker reconstruction quality was quantified by means of the mean absolute error (MAE) of inter-marker distance and two-stick angle. The kinematic computation quality was quantified by means of the measure of the knee angle reconstruction during walking and running trials. In order to evaluate the camera system and moving camera effects, we used a Wilcoxon rank sum test and a Kruskal Wallis test (post-hoc Tukey), respectively. The Spearman correlation coefficient (ρ) and the Wilcoxon rank sum test were applied to compare the kinematic data obtained by the two camera systems. We found small ASC MAE values (< 2.6mm and 1.3°), but they were significantly bigger than the MOCAP (< 0.7mm and 0.6°). However, for the human movement no significant differences were found between kinematic variables in walking and running acquisitions (p>0.05), and the motion patterns of the right-left knee angles between both systems were very similar (ρ>0.90, p<0.05). These results highlighted the promising results of a system that uses ASC based on the procedure of mobile cameras to follow the movement of the subject, allowing a less constrained movement in the direction in which the structure moves, compared to the traditional laboratory setup.

A Textile Sensor for Long Durations of Human Motion Capture

Human posture and movement analysis is important in the areas of rehabilitation, sports medicine, and virtual training. However, the development of sensors with good accuracy, low cost, light weight, and suitability for long durations of human motion capture is still an ongoing issue. In this paper, a new flexible textile sensor for knee joint movement measurements was developed by using ordinary fabrics and conductive yarns. An electrogoniometer was adopted as a standard reference to calibrate the proposed sensor and validate its accuracy. The knee movements of different daily activities were performed to evaluate the performance of the sensor. The results show that the proposed sensor could be used to monitor knee joint motion in everyday life with acceptable accuracy.