Ultra-Wideband Radar-Based Indoor Activity Monitoring for Elderly Care.

Area of Science:

• Health Informatics
• Sensor Technology
• Machine Learning

Background:

• Traditional health monitoring systems often rely on cameras, raising privacy concerns.
• There is a need for unobtrusive methods to collect health-related parameters in home environments.

Purpose of the Study:

• To propose an unobtrusive method and architecture for simultaneous presence monitoring and health parameter collection.
• To utilize ultra-wideband (UWB) impulse-radar for non-camera-based human sensing.

Main Methods:

• Employing a single UWB impulse-radar as the primary sensing device.
• Developing algorithms to recognize individuals, preselected movements (e.g., coughing), and static postures.
• Implementing a k-nearest neighbor machine learning model for posture classification using UWB data.
• Analyzing statistical features like skewness, kurtosis, and received power for classification.

Main Results:

• Demonstrated remote detection of small chest movements indicative of coughing using UWB signals.
• Achieved over 99% accuracy in classifying static postures using the UWB radar and k-nearest neighbor model.
• Presented reliability and fault tolerance analyses for UWB radar network architectures.

Conclusions:

• UWB radar offers a viable, privacy-preserving technology for unobtrusive health monitoring and presence detection in homes.
• The developed system can accurately identify individuals, movements, and postures, with high classification accuracy.
• Reliability and fault tolerance analysis of UWB network architectures is crucial for practical implementation.