Enhancing accuracy in dynamic pose estimation for sports competitions using HRPose: A hybrid approach integrating SinglePose AI

Human pose estimation plays a critical role in various applications, such as sports performance evaluation, rehabilitation, and human–computer interaction. Recent advancements in deep learning have significantly improved the accuracy and robustness of human pose estimation models. However, challenge...

Full description

Saved in:
Bibliographic Details
Main Authors: Rui Han, Mingnong Yi, Wei Feng, Feng Qi, Yining Zhou
Format: Article
Language:English
Published: Elsevier 2025-08-01
Series:Alexandria Engineering Journal
Subjects:
Online Access:http://www.sciencedirect.com/science/article/pii/S1110016825005587
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Human pose estimation plays a critical role in various applications, such as sports performance evaluation, rehabilitation, and human–computer interaction. Recent advancements in deep learning have significantly improved the accuracy and robustness of human pose estimation models. However, challenges remain in dynamic environments, especially in sports competitions, where high-speed movements, occlusions, and complex backgrounds often hinder accurate estimation. This paper proposes HRPose, a novel approach that combines HRNet for feature extraction and SinglePose AI for precise keypoint localization. It maintains high-resolution feature maps throughout the feature extraction process, enabling the model to capture fine-grained spatial details. SinglePose AI uses these features to generate and refine keypoint heatmaps, achieving accurate pose estimation even in challenging conditions. We evaluate HRPose on benchmark datasets, including the MPII Human Pose and PoseTrack datasets, and compare it with several models. Our results demonstrate that HRPose achieves superior performance in terms of mAP, precision, and robustness. Additionally, we discuss the real-time performance of HRPose and its potential applications in various domains, such as sports, healthcare, and rehabilitation. Future work will focus on improving the model’s robustness to extreme conditions, such as low lighting and motion blur, and exploring its integration with multimodal data for more comprehensive analysis.
ISSN:1110-0168