Human Activity Recognition Using Hybrid Coronavirus Disease Optimization Algorithm for Internet of Medical Things

Area of Science:

• Computer Science
• Biomedical Engineering
• Machine Learning

Background:

• Smartphones are increasingly used in healthcare for data collection via sensors, supporting applications like elderly care and disease diagnosis.
• High-dimensional sensor data from smartphones presents challenges for accurate human activity recognition (HAR).
• Not all features derived from smartphone sensors are equally valuable for HAR.

Purpose of the Study:

• To propose an efficient feature selection algorithm for human activity recognition (HAR).
• To enhance the performance of HAR by identifying the most relevant sensor data features.
• To address the challenge of high dimensionality in smartphone-derived healthcare data.

Main Methods:

• A hybrid algorithm combining the Coronavirus Disease Optimization Algorithm (COVIDOA) with Simulated Annealing (SA) was developed.
• The hybrid approach aims to improve feature selection efficiency and overcome local optima in optimization.
• The algorithm focuses on selecting a relevant subset of features for improved HAR.

Main Results:

• The proposed algorithm was evaluated using the UCI-HAR dataset.
• Performance was compared against seven established feature selection algorithms.
• Key metrics included fitness, standard deviation, accuracy, subset size, and processing time.

Conclusions:

• The developed hybrid optimization algorithm significantly outperforms existing HAR techniques.
• Achieved an average accuracy of 97.82% in human activity recognition.
• Demonstrated a substantial feature selection reduction ratio of 52.7%.