Improving data acquisition speed and accuracy in sport using neural networks

Identification of pattern mining algorithm for rugby league players positional groups separation based on movement patterns

The application of pattern mining algorithms to extract movement patterns from sports big data can improve training specificity by facilitating a more granular evaluation of movement. Since movement patterns can only occur as consecutive, non-consecutive, or non-sequential, this study aimed to identify the best set of movement patterns for player movement profiling in professional rugby league and quantify the similarity among distinct movement patterns. Three pattern mining algorithms (l-length Closed Contiguous [LCCspm], Longest Common Subsequence [LCS] and AprioriClose) were used to extract patterns to profile elite rugby football league hookers (n = 22 players) and wingers (n = 28 players) match-games movements across 319 matches. Jaccard similarity score was used to quantify the similarity between algorithms' movement patterns and machine learning classification modelling identified the best algorithm's movement patterns to separate playing positions. LCCspm and LCS movement patterns shared a 0.19 Jaccard similarity score. AprioriClose movement patterns shared no significant Jaccard similarity with LCCspm (0.008) and LCS (0.009) patterns. The closed contiguous movement patterns profiled by LCCspm best-separated players into playing positions. Multi-layered Perceptron classification algorithm achieved the highest accuracy of 91.02% and precision, recall and F1 scores of 0.91 respectively. Therefore, we recommend the extraction of closed contiguous (consecutive) over non-consecutive and non-sequential movement patterns for separating groups of players.

"Selecting the right tool for the job" a narrative overview of experimental methods used to measure or estimate active and passive drag in competitive swimming

Free-swimming performance depends strongly on the ability to develop propulsive force and minimise resistive drag. Therefore, estimating resistive drag (passive or active) may be important to understand how free-swimming performance can be improved. The purpose of this narrative overview was to describe and discuss experimental methods of measuring or estimating active and passive drag relevant to competitive swimming. Studies were identified using a mixed-model approach comprising a search of SCOPUS and Web of Science data bases, follow-up of relevant studies cited in manuscripts from the primary search, and additional studies identified by the co-authors based on their specific areas of fluid dynamics expertise. The utility and limitations of active and passive drag methods were critically discussed with reference to primary research domains in this field, 'swimmer morphology' and 'technique analysis'. This overview and the subsequent discussions provide implications for researchers when selecting an appropriate method to measure resistive forces (active or passive) relevant to improving performance in free-swimming.

The role of the biomechanics analyst in swimming training and competition analysis

Swimming analysts aid coaches and athletes in the decision-making by providing evidence-based recommendations. The aim of this narrative review was to report the best practices of swimming analysts that have been supporting high-performance athletes. It also aims to share how swimming analysts can translate applied research into practice. The role of the swimming analyst, as part of a holistic team supporting high-performance athletes, has been expanding and is needed to be distinguished from the job scope of a swimming researcher. As testing can be time-consuming, analysts must decide what to test and when to conduct the evaluation sessions. Swimming analysts engage in the modelling and forecast of the performance, that in short- and mid-term can help set races target-times, and in the long-term provide insights on talent and career development. Races can be analysed by manual, semi-automatic or fully automatic video analysis with single or multi-cameras set-ups. The qualitative and quantitative analyses of the swim strokes, start, turns, and finish are also part of the analyst job scope and associated with race performance goals. Land-based training is another task that can be assigned to analysts and aims to enhance the performance, prevent musculoskeletal injuries and monitor its risk factors.

A markerless motion capture system can reliably determine peak trunk flexion while squatting with and without a weighted vest

Markerless motion capture has improved physical screening efficiency in sport and occupational settings; however, reliability of kinematic measurements from commercial systems must be established. Further, the impact of torso-borne equipment on these measurements is unclear. The purpose of this study was to evaluate the reliability of HumanTrak, a markerless motion capture system, for estimating peak trunk flexion in squat movements with and without a weighted vest. Eighteen participants completed body weight squats (BWSQ) and overhead squats (OHSQ) to their maximum depth (unrestricted-range) and to a plyometric box (fixed-range) while wearing no body armour (NBA) or 9 kg body armour (BA9). Peak trunk flexion was measured using HumanTrak. Testing was performed in two sessions on one day (intra-day) and one session on a separate day (inter-day) to assess reliability. HumanTrak had a standard error of measurement < 3.74° across all movements and conditions. Reliability was good to excellent (ICC = 0.82-0.96) with very large to nearly perfect Pearson correlations (r > 0.80) for all comparisons except unrestricted-range BWSQ with BA9 (ICC = 0.60-0.71, r = 0.71). HumanTrak was more reliable for intra- than inter-day, but reliability was still excellent for almost all inter-day comparisons (ICC > 0.82). HumanTrak is reliable for detecting differences in peak trunk flexion > 8.5° when body armour is not worn and > 10.5° when body armour is worn. Practitioners can assess meaningful changes in sagittal plane trunk motion when screening squat movements regardless of whether body armour is worn.

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.

Survey on Video-Based Biomechanics and Biometry Tools for Fracture and Injury Assessment in Sports

This work presents a survey literature review on biomechanics, specifically aimed at the study of existent biomechanical tools through video analysis, in order to identify opportunities for researchers in the field, and discuss future proposals and perspectives. Scientific literature (journal papers and conference proceedings) in the field of video-based biomechanics published after 2010 were selected and discussed. The most common application of the study of biomechanics using this technique is sports, where the most reported applications are american football, soccer, basketball, baseball, jumping, among others. These techniques have also been studied in a less proportion, in ergonomy, and injury prevention. From the revised literature, it is clear that biomechanics studies mainly focus on the analysis of angles, speed or acceleration, however, not many studies explore the dynamical forces in the joints. The development of video-based biomechanic tools for force analysis could provide methods for assessment and prediction of biomechanical force associated risks such as injuries and fractures. Therefore, it is convenient to start exploring this field. A few case studies are reported, where force estimation is performed via manual tracking in different scenarios. This demonstration is carried out using conventional manual tracking, however, the inclusion of similar methods in an automated manner could help in the development of intelligent healthcare, force prediction tools for athletes and/or elderly population. Future trends and challenges in this field are also discussed, where data availability and artificial intelligence models will be key to proposing new and more reliable methods for biomechanical analysis.

Agreement Between Sagittal Foot and Tibia Angles During Running Derived From an Open-Source Markerless Motion Capture Platform and Manual Digitization

Several open-source platforms for markerless motion capture offer the ability to track 2-dimensional (2D) kinematics using simple digital video cameras. We sought to establish the performance of one of these platforms, DeepLabCut. Eighty-four runners who had sagittal plane videos recorded of their left lower leg were included in the study. Data from 50 participants were used to train a deep neural network for 2D pose estimation of the foot and tibia segments. The trained model was used to process novel videos from 34 participants for continuous 2D coordinate data. Overall network accuracy was assessed using the train/test errors. Foot and tibia angles were calculated for 7 strides using manual digitization and markerless methods. Agreement was assessed with mean absolute differences and intraclass correlation coefficients. Bland-Altman plots and paired t tests were used to assess systematic bias. The train/test errors for the trained network were 2.87/7.79 pixels, respectively (0.5/1.2 cm). Compared to manual digitization, the markerless method was found to systematically overestimate foot angles and underestimate tibial angles (P < .01, d = 0.06-0.26). However, excellent agreement was found between the segment calculation methods, with mean differences ≤1° and intraclass correlation coefficients ≥.90. Overall, these results demonstrate that open-source, markerless methods are a promising new tool for analyzing human motion.

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.

A Study of Athlete Pose Estimation Techniques in Sports Game Videos Combining Multiresidual Module Convolutional Neural Networks

In this paper, we propose a multiresidual module convolutional neural network-based method for athlete pose estimation in sports game videos. The network firstly designs an improved residual module based on the traditional residual module. Firstly, a large perceptual field residual module is designed to learn the correlation between the athlete components in the sports game video within a large perceptual field. A multiscale residual module is designed in the paper to better solve the inaccuracy of the pose estimation due to the problem of scale change of the athlete components in the sports game video. Secondly, these three residual modules are used as the building blocks of the convolutional neural network. When the resolution is high, the large perceptual field residual module and the multiscale residual module are used to capture information in a larger range as well as at each scale, and when the resolution is low, only the improved residual module is used. Finally, four multiresidual module convolutional neural networks are used to form the final multiresidual module stacked convolutional neural network. The neural network model proposed in this paper achieves high accuracy of 89.5% and 88.2% on the upper arm and lower arm, respectively, so the method in this paper reduces the influence of occlusion on the athlete's posture estimation to a certain extent. Through the experiments, it can be seen that the proposed multiresidual module stacked convolutional neural network-based method for athlete pose estimation in sports game videos further improves the accuracy of athlete pose estimation in sports game videos.

Augmented feedback can change body shape to improve glide efficiency in swimming

Curvatures of the body can disrupt fluid flow and affect hydrodynamic resistance. The purpose of this study was to evaluate the effect of a feedback intervention on glide performance and torso morphology. Eleven male and female national swimmers performed glides before and after augmented feedback. Feedback consisted of self-modelling visual feedback and verbal cuing, to manipulate body curvatures that affect hydrodynamic resistance. Two-dimensional landmark position data (knee, hip and shoulder) were used to enable computation of glide factor and glide coefficient as indicators of glide efficiency; posture (trunk incline and hip angle); and performance (horizontal velocity). Underwater images of the swimmers were manually traced to derive transverse and sagittal diameters, cross-sectional areas, and continuous form outlines (anterior and posterior) of the torso. Maximum rate of change in cross-sectional area and form gradient progressing caudally, were calculated for torso segments: shoulder-chest, chest-waist, waist-hip. Mean velocity, glide factor and glide coefficient values significantly (p< 0.001) improved due to the intervention, with large effect size (d) changes 0.880 (p= 0.015), 2.297 and 1.605, respectively. Significant changes to form gradients were related to reductions in lumbar lordosis and chest convexity. The study provides practical cuing phrases for coaches and swimmers to improve glide efficiency and performance.