A depth video sensor-based life-logging human activity recognition system for elderly care in smart indoor...

Area of Science:

• Computer Vision
• Artificial Intelligence
• Biomedical Engineering

Background:

• Conventional human activity recognition (HAR) relies on RGB sensors.
• Depth video sensors offer improved data by providing distance information.
• Elderly monitoring applications can benefit significantly from advanced HAR.

Purpose of the Study:

• To design a depth-based life logging HAR system for elderly activity recognition.
• To transform living environments into intelligent spaces for elderly care.
• To evaluate the system's effectiveness in recognizing daily activities.

Main Methods:

• Utilizing depth imaging sensors to capture depth silhouettes.
• Extracting human skeleton and joint information for activity analysis.
• Employing Hidden Markov Models (HMMs) for activity training and recognition.
• Implementing a two-process system: training and recognition engine.

Main Results:

• The depth-based system achieved satisfactory recognition rates compared to conventional methods.
• Life logging features demonstrated effectiveness against principal component and independent component features.
• Experiments on benchmark datasets (e.g., MSRDailyActivity3D) showed promising results.

Conclusions:

• The proposed depth-based HAR system is effective for elderly monitoring.
• The system enables the creation of intelligent living environments for enhanced elder care.
• Direct applicability to healthcare monitoring and indoor activity analysis in various settings.