Intelligent Reflecting Surface-Based Non-LOS Human Activity Recognition for Next-Generation 6G-Enabled Healthcare...

Area of Science:

• Human-Computer Interaction
• Signal Processing
• Machine Learning

Background:

• Human activity monitoring is crucial for supporting independent living in elderly and disabled populations.
• Existing methods like cameras and wearables raise privacy and comfort concerns.
• Microwave sensing offers a privacy-preserving alternative but struggles with non-line-of-sight and multi-floor environments.

Purpose of the Study:

• To evaluate the effectiveness of Intelligent Reflecting Surface (IRS) systems for human activity monitoring.
• To apply and assess machine learning algorithms on IRS data for improved activity detection.
• To analyze the computational processing time of these algorithms in the context of IRS data.

Main Methods:

• Utilized a publicly available dataset from an IRS system.
• Applied machine learning algorithms including Support Vector Machine (SVM), Bagging, and Decision Tree.
• Evaluated classification accuracy and processing time on specific human activities.

Main Results:

• Achieved improved accuracy in human activity detection using IRS data with specific machine learning algorithms.
• Demonstrated superior performance of IRS systems in non-line-of-sight and multi-floor scenarios compared to traditional radar.
• Provided novel insights into the processing time of classifiers trained on IRS data.

Conclusions:

• IRS systems show significant promise for advanced human activity monitoring, particularly in challenging environments.
• Machine learning algorithms applied to IRS data can enhance accuracy and efficiency for healthcare applications.
• This research opens new avenues for privacy-preserving and effective assistive technologies.