Sports video analysis has evolved from niche methodologies to expansive technological frameworks that integrate multiple sports disciplines. This research provides a comprehensive comparative analysis of four deep learning architectures: ConvLSTM, Long-term Recurrent Convolutional Networks (LRCNN), Optical Flow with LSTM, and C3D-Lite, specifically for detailed tennis action recognition using the Three dimEnsional TennIs Shots THETIS dataset. The experiments were performed over five random seeds and three stratifications of data into three skill sets: S1 Mixed (all players), BS2 Beginner-Only (Players 1–31), and BS3 Expert-Only (Players 32–55), using common train-validation-test splits to ensure the reproducibility of results over different architectures. The LRCNN provided the best performance (Verified with results in Table 55 65%) and a generalization gap of 14 percentage points due to the use of pretrained features of the InceptionV3 architecture. The C3D-Lite architecture demonstrated the most accurate-efficiency trade-off in terms of mean test accuracy 55.25% ± 3.5% for S1. The Optical Flow + LSTM combination provided validation accuracy of 57.8% (S1), 58.2% (BS2), and 63.8% (BS3), whereas training accuracy was 100% for each seed and strategy, which implies full memorization of the training set. ConvLSTM suffered severe overfitting (∼40% training-to-test gap) and is not recommended for limited-scale video datasets. Cross-seed McNemar analysis confirmed result reproducibility across all strategies. Performance remained consistent across skill-stratified partitions, indicating that mixed-skill datasets are sufficient for deployment. This work establishes the first multi-architecture benchmark on the THETIS dataset with full computational profiling, cross-sport applicability analysis, and deployment guidance for racquet sports analytics.
Anthropometric test devices (ATDs) are human surrogate devices that are used in impact testing for the simulation and evaluation of helmet impact performance in American football. However, these ATD's are often used outside of the scope of their original testing purposes, leading to an increased potential for misinterpretation of laboratory testing data. In this study a computational analysis was conducted of human neck muscular equivalence to that of the Hybrid III neckform using the HYOID OpenSim model scaled to 50th and 95th percentiles of human muscle mechanical properties. Neck bending was simulated in four directions (sagittal, coronal, 45 degrees off-center, and transverse planes) and compared to experimental data on the Hybrid III neckform. In a muscle parameter optimization, maximum isometric force () and passive fiber strain at () were altered to match the stiffness responses of the HYOID model to the Hybrid III neckform. The Hybrid III sagittal plane flexion response was found to be non-physiologic, requiring a 275% increase in from the HYOID 50th percentile model. The extension response was moderately more physiologic, requiring only a 66% increase in . The Hybrid III neckform was consistently closer to the HYOID 95th percentile model in all planes of motion, suggesting a better match with collegiate/professional populations than the average male. Results from this work can be used as a baseline for future dynamic comparisons to improve the biofidelity of surrogate devices used in the laboratory testing of head protective equipment for American football.
High-fidelity biomechanical assessment typically relies on instrumented systems (e.g. optical motion capture or inertial measurement units (IMUs)), which can be costly, intrusive, and difficult to deploy at scale. This study presents a contactless, smartphone-based framework for preliminary kinematic screening of a ballet Grand Plié task using monocular video. Using MediaPipe BlazePose, the system derives 2D pose-vector indicators for knee-flexion and alignment-related features and applies dynamic time warping (DTW) to compare input sequences with reference template trajectories derived from the training set after temporal alignment and averaging. A retrospective dataset of 250 videos was curated from public sources; after quality control and segmentation, 220 valid motion instances (video clips) were retained for analysis. On a contemporary mobile device (Snapdragon 8 Gen 1), the pipeline runs in real time, achieving 22 frames per second (FPS). Agreement with an expert-derived 2D reference (frame-based digital goniometry; n = 50 key frames) yielded a mean bias of +1.2° and 95% limits of agreement (LoA) of −3.3° to +5.6°. A Monte Carlo sensitivity analysis indicated resilience to simulated landmark perturbations (additive Gaussian noise with σ = 0.01 applied to normalized coordinates), resulting in a mean angular error of approximately 4.0°. In a subject-level pilot evaluation ( N = 5), the framework achieved an F 1-score of 0.921 for detecting gross form deviations under the specified capture protocol. Overall, these results support the feasibility of mobile, markerless video-based screening as a practical complement to instrumented measurement for preliminary assessment under resource constraints.
The performance of a cyclist can be optimized by selecting the most appropriate gear ratio for a rider’s physiological profile, bike-rider setup, and environmental conditions. Currently, there is no analytic model available to guide gear ratio selection, and the aim of this study is to fill this gap. We study the kinematics of a cyclist using a physiologically motivated human power model. Analytic expressions for the position and velocity of the cyclist as functions of time are presented. We show how these formulas can be used to optimize the choice of gear ratio, taking into account the transient behavior of sprint-like efforts. The model can be used to guide gear ratio selection in a variety of cycling disciplines.
The majority of the arrows used on the archery target range are either simple carbon fibre composite tubes, or carbon fibre composite tubes with an inner tube of high strength aluminium. The addition of the aluminium component markedly increases the arrow's mass and reduces the arrow's lateral drift in wind. This paper shows that the optimal wall thickness of the aluminium layer to both minimise wind drift and to retain the characteristics of a predominantly carbon fibre arrow shaft is approximately 0.008 inch, which is comparable to commonly available high-performance arrow shafts. The paper also provides a useful metric for comparing the expected wind drifts of different arrows.
