Citation

Ang, Khai Pin and Low, Iven Zi Yin and Hooi, Yumun and Loh, Yuen Peng (2025) Syncscore: A Framework for Synchronization Scoring in Group Sports Via Human Pose Estimation. In: 2025 Asia Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA ASC), 22-24 October 2025, Singapore, Singapore.

Text 481.pdf - Published Version Restricted to Repository staff only Download (3MB)

Abstract

Human Pose Estimation (HPE) plays a critical role in performance analysis for synchronized sports such as Taekwondo Poomsae, where timing and coordination between participants are essential. Traditional synchronization scoring methods rely on manual expert judgment, which can be subjective and inconsistent. This paper proposes an automated framework for synchronization scoring, leveraging the ViTPose model for high-precision keypoint detection and robust generalization. To estimate synchronization scores, we compute inter-person landmark distances using Euclidean distance applied to aligned and temporally smoothed 2 D landmarks. These distances are then used as input features to regression models that estimate synchronization scores aligned with expert assessments. The framework was validated using the benchmark LSP dataset for HPE evaluation and a custom-annotated Taekwondo Poomsae dataset comprising 70 videos. Experimental results show that ViTPose, fine-tuned on the custom dataset provided good performance for complex HPE. Among the regression approaches, the combination of ViTPose-L with Support Vector Regression (SVR) achieved the highest synchronization score estimation with an R2 value of 0.3915. These findings demonstrate the potential of the proposed framework for objective and precise scoring in synchronized sports performance.

Item Type:	Conference or Workshop Item (Paper)
Uncontrolled Keywords:	Support vector machines, Accuracy, Computational modeling, Pose estimation, Predictive models, Benchmark testing, Vectors, Synchronization, Sports, Videos
Subjects:	H Social Sciences > HT Communities. Classes. Races
Divisions:	Faculty of Computing and Informatics (FCI)
Depositing User:	Ms Suzilawati Abu Samah
Date Deposited:	17 Mar 2026 04:34
Last Modified:	19 Mar 2026 01:41
URII:	http://shdl.mmu.edu.my/id/eprint/15479

Downloads

Edit (login required)