Optimization of an intelligent evaluation model for aerobics performance using artificial intelligence and multimodal deep representation

Manual aerobics scoring is sensitive to evaluator experience, while a single sensing modality cannot fully represent posture, dynamics, and musical timing. This study develops an intelligent evaluation model that encodes video, skeleton, inertial, and music-rhythm data in parallel and integrates them through cross-modal attention, reliability-aware fusion, temporal attention, and multi-task scoring. The dataset contains 8,466 valid sequences from 60 adult athletes. On a participant-independent test set, the model achieved an MAE of 3.18 points, R 2 of 0.946, and macro-F1 of 0.925 for four-class primary deviation recognition. Performance declined under cross-site and cross-device testing, with MAE increasing to 4.28 and 4.47 points, respectively. A four-week pilot involving 18 athletes and four teachers indicated shorter video-review time, but it was not designed to establish long-term technical improvement or deployment readiness. The model therefore provides an auxiliary source of quantitative evidence for training diagnosis rather than a substitute for expert judgment.

Authors

Institutions

Publication Details

Journal
Discover Artificial Intelligence
Published
2026-11-13
DOI
https://doi.org/10.1007/s44163-026-02754-3
Primary Topic
Human Pose and Action Recognition
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Optimization of an intelligent evaluation model for aerobics performance using artificial intelligence and multimodal deep representation

Haibin Zong, Beibei Xue, Tian Wang
2 links
Discover Artificial Intelligence
Human Pose and Action Recognition
article

Optimization of an intelligent evaluation model for aerobics performance using artificial intelligence and multimodal deep representation

Haibin Zong, Beibei Xue, Tian Wang
article en

Abstract

Manual aerobics scoring is sensitive to evaluator experience, while a single sensing modality cannot fully represent posture, dynamics, and musical timing. This study develops an intelligent evaluation model that encodes video, skeleton, inertial, and music-rhythm data in parallel and integrates them through cross-modal attention, reliability-aware fusion, temporal attention, and multi-task scoring. The dataset contains 8,466 valid sequences from 60 adult athletes. On a participant-independent test set, the model achieved an MAE of 3.18 points, R 2 of 0.946, and macro-F1 of 0.925 for four-class primary deviation recognition. Performance declined under cross-site and cross-device testing, with MAE increasing to 4.28 and 4.47 points, respectively. A four-week pilot involving 18 athletes and four teachers indicated shorter video-review time, but it was not designed to establish long-term technical improvement or deployment readiness. The model therefore provides an auxiliary source of quantitative evidence for training diagnosis rather than a substitute for expert judgment.

Discover Artificial IntelligenceVol. 6(1)
Changji University (CN)
Good health and well-being
Human Pose and Action Recognition
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon