Article: An athlete motion recognition model based on machine learning and the internet of things Journal: International Journal of Information and Communication Technology (IJICT) 2026 Vol.27 No.4 pp.48 - 68 Abstract: With the continuous improvement of sports training and competitive levels, athletes' demands for motion recognition and motion monitoring during training are increasing day by day. Based on a multi-node sensor platform and the internet of things environment, this study constructed an action data acquisition system and ensured high-quality data input through pre-processing and feature extraction. In terms of model construction and optimisation, the performance of LSTM, CNN, SVM and the fusion model was compared and analysed. The results show that the fusion model is significantly superior to the single model in terms of recognition accuracy, system delay, stability and energy consumption, especially in the recognition of complex actions such as rotation and bending, the accuracy exceeds 95%. Further three-dimensional surface analysis shows that the fusion model still maintains a latency of less than 120 milliseconds and a stability index higher than 0.85 in a high-load environment, demonstrating good robustness. Inderscience Publishers - linking academia, business and industry through research

Title: An athlete motion recognition model based on machine learning and the internet of things

Authors: Hua Liu

Addresses: Department of Physical Education, Xinxiang University, Xinxiang, 453003, China

Abstract: With the continuous improvement of sports training and competitive levels, athletes' demands for motion recognition and motion monitoring during training are increasing day by day. Based on a multi-node sensor platform and the internet of things environment, this study constructed an action data acquisition system and ensured high-quality data input through pre-processing and feature extraction. In terms of model construction and optimisation, the performance of LSTM, CNN, SVM and the fusion model was compared and analysed. The results show that the fusion model is significantly superior to the single model in terms of recognition accuracy, system delay, stability and energy consumption, especially in the recognition of complex actions such as rotation and bending, the accuracy exceeds 95%. Further three-dimensional surface analysis shows that the fusion model still maintains a latency of less than 120 milliseconds and a stability index higher than 0.85 in a high-load environment, demonstrating good robustness.

Keywords: machine learning; internet of things; IoT; action recognition; athlete training; stability.

DOI: 10.1504/IJICT.2026.151525

International Journal of Information and Communication Technology, 2026 Vol.27 No.4, pp.48 - 68

Received: 28 Sep 2025
Accepted: 10 Nov 2025
Published online: 04 Feb 2026 *

Title: An athlete motion recognition model based on machine learning and the internet of things

Keep up-to-date