Title: Sparse-LSTM for sports fatigue assessment in a wearable sensing network

Authors: Ziping Meng; Jingya An; Linpeng Xiao

Addresses: School of Social Sports, Tianjin University of Sport, Tianjin, 300000, China ' School of Social Sports, Tianjin University of Sport, Tianjin, 300000, China ' School of Management, Beijing Sport University, Beijing, 100000, China

Abstract: Accurate assessment of exercise fatigue has become a critical requirement for enhancing the scientific rigor of athletic training. However, traditional methods face challenges such as insufficient feature extraction capabilities and low evaluation accuracy. To address this, this paper first collects exercise fatigue data through a wearable sensor network, then performs window segmentation to obtain effective keyframe data. Building upon an enhanced Inception network architecture, the feature map dimensions are expanded to enable estimation of exercise fatigue actions. To capture key motion trajectories, a sparse distribution-enhanced long short-term memory network is employed for temporal feature extraction. Finally, a similarity evaluation method based on dynamic time warping and the longest common subsequence is designed to analyse angular distance differences, thereby enabling the assessment of athletic fatigue. Experimental results demonstrate that the proposed model achieves an improvement in evaluation accuracy of 4.08% to 13.97%.

Keywords: exercise fatigue assessment; wearable sensor network; sparse distribution-enhanced long short-term memory network; LSTM; dynamic time warping; DTW; longest common subsequence; LCS.

DOI: 10.1504/IJSNET.2026.153106

International Journal of Sensor Networks, 2026 Vol.50 No.4, pp.260 - 272

Received: 29 Sep 2025
Accepted: 30 Sep 2025
Published online: 22 Apr 2026 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article