Motion feature extraction based on semi-supervised learning and long short-term memory network in digital dance
Frontiers in Neurorobotics · 29.01.2026 · 20
DOI: 10.3389/fnbot.2026.1743288 ↗Kurz gefasst
Fachartikel in der Zeitschrift Frontiers in Neurorobotics, online erschienen am 29.01.2026. Verfasst von Xue Yang und 3 weiteren Autor:innen. Beteiligte Einrichtung: College of Music, Huaiyin Normal University, Huai'an, China. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 48 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Jiangsu university.
Abstract (Originalsprache)
Digital-image technology has broadened the creative space of dance, yet accurately capturing the semantic correspondence between low-level motion data and high-level dance key-points remains challenging, especially when labeled data are scarce. We aim to establish a lightweight, semi-supervised pipeline that can extract discriminative motion features from depth sequences and map them to 3-D key-points of dancers in real time. To achieve pixel-level alignment between dance movement targets and high-dimensional sensory data, we propose a novel LSTM-CNN (Long Short Term Memory-Convolutional Neural Network) framework. Temporal-context features are first extracted by LSTM, after which multi-dimensional spatial features are captured by three convolutional layers and one max-pooling layer; the fused representation is finally regressed to 3-D body key-points. To relieve class imbalance caused by complex postures, an online hard-example mining (OHEM) strategy together with a Dice-cross-entropy weighted loss (3:1) is embedded into semi-supervised learning, enabling the network to converge with only 20% labeled samples. Experiments on the public MSR-Action3D dataset (567 sequences, 20 actions) yielded an average recognition rate of 96.9%, surpassing the best comparison method (MSST) by 1.1%. On our self-established dataset (99 sequences, 11 actions) the accuracy reached 97.99% while training time was reduced by 35% compared with the previous best Multi_perspective_MHPCs approach. Both datasets show low RMSE (≤ 0.032) between predicted and ground-truth key-points, confirming spatial precision. The results demonstrate that the proposed model can reliably track subtle dance gestures under limited annotation, offering an efficient, low-cost solution for digital choreography, motion-style transfer and interactive stage performance.
Schlagwörter
Stichwörter der Autor:innen
Autor:innen und Institutionen
- College of Music, Huaiyin Normal University, Huai'an, China.
- Weihai Art College, Shandong University, Weihai, China.
- Department of Dance School of Dance, Taipei National University of the Arts, Taipei City, Taiwan.
- College of Artificial Intelligence, Shenyang Normal University, Shenyang, China.
Förderung
- Jiangsu university
Zitieren
APA
Yang, X., Sun, H., Lyu, Y., & Sun, Y. (2026). Motion feature extraction based on semi-supervised learning and long short-term memory network in digital dance. Frontiers in Neurorobotics, 20. https://doi.org/10.3389/fnbot.2026.1743288
Vancouver
Yang X, Sun H, Lyu Y, Sun Y. Motion feature extraction based on semi-supervised learning and long short-term memory network in digital dance. Frontiers in Neurorobotics. 2026;20. doi:10.3389/fnbot.2026.1743288
BibTeX
@article{Yang2026,
author = {Yang, Xue and Sun, Hanmin and Lyu, Yin and Sun, Yang},
title = {Motion feature extraction based on semi-supervised learning and long short-term memory network in digital dance},
journal = {Frontiers in Neurorobotics},
year = {2026},
volume = {20},
doi = {10.3389/fnbot.2026.1743288}
}
EUVIMED