Deep learning-based gym activity recognition using wearable sensors with potential applications in vocational physical education
Wearable-sensor-based human activity recognition can support objective monitoring of physical exercise and may provide a foundation for future motor-skill assessment in physical education. This study presents a deep learning framework for gym activity recognition using the RecGym dataset, which contains wearable-sensor-derived features. The framework applies noise removal, mean imputation, and Z-score normalization during preprocessing, followed by a Deep Autoencoder for feature reduction and a hybrid Temporal Convolutional Network (TCN) with an Attention mechanism for activity recognition. A Softmax classifier predicts the activity classes. The model achieved an accuracy of 0.9903, precision of 0.9906, recall of 0.9903, and F1-score of 0.9903 on the evaluated dataset. These results demonstrate the model’s performance in gym activity classification. However, the dataset does not provide validated motor-skill proficiency labels for higher vocational physical education students; therefore, the reported results do not establish the validity of student skill-level assessment. Validation with vocational students and independently measured motor-skill outcomes is required before the framework can be used to assess educational performance.
Authors
- Xiaofan Yang
Institutions
- Xinzhou Teachers University (CN)
- Xinzhou Vocational and Technical College (CN)
Publication Details
- Journal
- Discover Artificial Intelligence
- Published
- 2026-11-18
- DOI
- https://doi.org/10.1007/s44163-026-02762-3
- Primary Topic
- Context-Aware Activity Recognition Systems
- Type
- article
- Field-Weighted Citation Impact
- 0.00