Related Experiment Video
Updated: Jan 16, 2026
06:37
Author Spotlight: Addressing Technical and Subjective Challenges in Measuring Classroom Attention
Published on: December 15, 2023
5.3K
Dual Attention-Based recurrent neural network and Two-Tier optimization algorithm for human activity recognition in
Hend Khalid Alkahtani1, Gouse Pasha Mohammed2, Radwa Marzouk3,4
1Department of Information Systems, College of Computer and Information Sciences, Princess Nourah Bint Abdulrahman University, P.O. Box 84428, Riyadh, 11671, Saudi Arabia. hkAlqahtani@pnu.edu.sa.
Scientific Reports
|September 29, 2025
Summary
This study introduces a new model for Human Activity Recognition (HAR) to aid disabled individuals. The Dual Attention-Based Two-Tier Metaheuristic Optimization Algorithm achieved 98.66% accuracy, significantly improving HAR performance.
Area of Science:
- Computer Science
- Artificial Intelligence
- Biomedical Engineering
Background:
- Human Activity Recognition (HAR) is crucial for applications like remote monitoring, healthcare, and security.
- Existing HAR methods utilize diverse sensors and techniques, including wearable, object-tagged, and device-free approaches.
- Deep learning (DL) and machine learning (ML) have shown significant promise in enhancing HAR accuracy.
Purpose of the Study:
- To propose a novel Dual Attention-Based Two-Tier Metaheuristic Optimization Algorithm for Human Activity Recognition with Disabilities (DATTMOA-HARD).
- To specifically improve HAR systems to better assist individuals with disabilities.
- To enhance the accuracy and efficiency of human activity detection.
Main Methods:
- The DATTMOA-HARD model employs Z-score normalization for data preprocessing.
- Feature selection is performed using the binary firefly algorithm (BFA).
- Classification is achieved through a dual attention bidirectional gated recurrent unit (DABiG) technique, with hyperparameters optimized by the Tasmanian devil optimizer (TDO).
Main Results:
- The DATTMOA-HARD model demonstrated superior performance on the HAR dataset.
- Experimental evaluation showed a high accuracy rate of 98.66%.
- The proposed model significantly outperformed existing HAR methods in detection accuracy.
Conclusions:
- The DATTMOA-HARD model offers a significant advancement in Human Activity Recognition, particularly for assisting disabled individuals.
- The combination of dual attention mechanisms, metaheuristic optimization, and advanced DL techniques leads to highly accurate activity detection.
- This research highlights the potential of sophisticated AI models in creating more inclusive and supportive technological solutions.
