Sep 2026· 2026 International Conference on Data Science and Its Applications (ICoDSA)· pp. 395-400· 0 citations· 25 references
TL;DR
A push-up posture classification system using a low-capacity lightweight Long Short-Term Memory architecture to prevent overfitting and demonstrates the most consistent cross-subject generalization, evidenced by the lowest standard deviation across accuracy, recall, and F1-score.
Abstract
Automated evaluation of push-up posture quality via computer vision has the potential to prevent physical injuries without requiring the supervision of a professional instructor. However, current literature still struggles with a reliance on the spatial coordinate domain, which is vulnerable to anthropometric bias and camera scale distortion; massive deep learning architectures that are prone to overfitting on small-scale datasets; and random split evaluation based on sliding windows that triggers the phenomenon of performance overestimation. This study proposes a push-up posture classification system using a low-capacity lightweight Long Short-Term Memory (LSTM) architecture to prevent overfitting. The system utilizes four joint angle features (elbow, shoulder, hip, and knee) extracted from MediaPipe PoseLandmarker, converted via the arctan2 function to produce a scale-and-translation-invariant representation. Temporal data is standardized into a constant length of 60 frames and augmented using proportional noise injection. To prevent data leakage, sequences are extracted in isolation per full repetition cycle and evaluated using a Group 4-Fold Cross-Validation scheme with an independent validation set. The proposed LSTM model achieves the highest average performance with 86.12% accuracy and an 85.79% F1-score against GRU (85.12%) and SVM (83.75%) baseline models. Although pairwise McNemar's tests indicate that the margins do not reach statistical significance on this small-scale dataset, LSTM demonstrates the most consistent cross-subject generalization, evidenced by the lowest standard deviation across accuracy, recall, and F1-score.
Maintaining correct posture during strength training is essential for optimizing performance and preventing injuries. However, automated posture evaluation systems often face challenges in real-world gym environments due to variations in lighting, clothing, and body types. This study presents a real-time posture clas...
Prokash Gogoi, J. A. Valan, Tovi Achumi et al.· Proceedings of the Instituti...· 0 citations
: Badminton is one of the most popular sports globally, yet coaches and sports analysts still rely on conventional methods to evaluate players from match videos. As a first step toward a comprehensive badminton analytics platform, this paper proposes an end-to-end system to classify four badminton strokes (serve, smash...
M. Fadhlan, U. Wiil· Proceedings of the 14th Inte...· 0 citations
Musculoskeletal disorders impact more than 1.3 billion people globally, and rehabilitation in the home setting is crucial to the process. Due to the lack of professional oversight, patient compliance usually dips under 35%. This study describes RehabSense, an AI-powered software framework aiming to fill this void. Medi...
Ravi Kishan Kumar, Rachit Mittal, Soumya Jha et al.· 2026 Forth International Con...· 0 citations
The feasibility of a deep learning model based on smartphone video recordings to enable real-time detection and feedback of basketball free-throw technique execution is validated, and the integration of edge computing and multimodal fusion to enhance robustness is explored.
Jin-Yu Wang· Computers and artificial int...· 0 citations
To address the challenges of small motion amplitudes, high similarity among upper-limb postures, and substantial variations in the duration of different driver distraction behaviors, this paper proposes a driver distraction behavior recognition method based on pose estimation and an improved PoseC3D model. First, YOLO1...
Human posture classification is an important computer vision task for intelligent monitoring, rehabilitation, assistive systems, and human–machine interaction. This study presents a comparative evaluation of four transfer learning configurations, TL-ResNet-18, TL-ResNet-50, TL-MobileNet, and TL-Xception, under a common...
A. Pradhan, Sarita Tripathy, K. Anbananthen et al.· Future Internet· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.