Deep Learning and Metaheuristic Optimization for IoT-based Wearable Sensor Network in Rehabilitation Movement Assessment
Journal of Al-Farabi for Engineering Sciences · 2026
DOI: 10.59746/8mvpns51 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Al-Farabi for Engineering Sciences, 2026. Verfasst von Muna Al-Saadi.
Abstract (Originalsprache)
A clear assessment of rehabilitation exercise performance is important for safe and personal training. However, common assessment still depends on visual observation and may differ between assessors. This study develops an Escaping Bird Search-Optimized Long Short-Term Memory (EBS-LSTM) framework for subject-independent classification of correct and incorrect rehabilitation movements from multi-joint wearable-sensor data. The PHYTMO dataset contains recordings from 30 adults. Signals were processed using a fourth-order 4 Hz Butterworth low-pass filter, linear interpolation, training-derived min-max normalization, and five-level discrete wavelet transform features. An LSTM temporal model was coupled with EBS using a composite fitness function that combines classification error and root mean square error (RMSE), and performance was compared with BN, KNN, SVM, MLP, and AE-CNN under the same subject-independent evaluation protocol. EBS-LSTM achieved 96.81 ± 0.18% accuracy, 95.85 ± 0.23% precision, 97.40 ± 0.22% recall, 96.35 ± 0.25% specificity, 96.62 ± 0.20% F1-score, and an RMSE of 0.018. Its mean accuracy exceeded AE-CNN, the strongest reported baseline, by 3.47 percentage points. The results support EBS-guided recurrent modelling for wearable-sensor assessment of rehabilitation movement-execution quality. The output should be interpreted as a measure of execution correctness rather than a complete measure of teaching effectiveness, and external validation in patient and home-rehabilitation settings remains necessary.
Zitieren
APA
Al-Saadi, M. (2026). Deep Learning and Metaheuristic Optimization for IoT-based Wearable Sensor Network in Rehabilitation Movement Assessment. Journal of Al-Farabi for Engineering Sciences. https://doi.org/10.59746/8mvpns51
Vancouver
Al-Saadi M. Deep Learning and Metaheuristic Optimization for IoT-based Wearable Sensor Network in Rehabilitation Movement Assessment. Journal of Al-Farabi for Engineering Sciences. 2026. doi:10.59746/8mvpns51
BibTeX
@article{AlSaadi2026,
author = {Al-Saadi, Muna},
title = {Deep Learning and Metaheuristic Optimization for IoT-based Wearable Sensor Network in Rehabilitation Movement Assessment},
journal = {Journal of Al-Farabi for Engineering Sciences},
year = {2026},
doi = {10.59746/8mvpns51}
}
EUVIMED