Deep learning for stress oriented human activity recognition

1

This study focuses on Human Activity Recognition (HAR) to assess stress-related behaviors using sensor-generated time-series data.
2

Multiple deep learning architectures, including RNNs, LSTMs, and Transformers, were employed for feature extraction and classification.
3

Transformer models achieved the highest classification accuracy of 97.83%, outperforming LSTM and RNN models in stress detection.
4

The research highlights the potential of HAR in monitoring stress through non-intrusive wearable devices.
5

Optimal window size and overlap ratio significantly influence the classification accuracy of stress-related activities.

Frontiers In Digital Health

by Muhammad Hamza, Nasir Uddin, Gulnaz Anjum, Mohammad Anas, Uzair Gabol, Nida Saddaf Khan
May 18, 2026

1