Indoor Group Identification and Localization Using Privacy-Preserving Edge Computing Distributed Camera Network.

Area of Science:

• Computer Vision
• Behavioral Monitoring
• Edge Computing

Background:

• Social interaction changes can indicate physical or psychiatric health issues.
• Monitoring patient group activity is crucial for mental health assessment in clinical settings.
• Existing methods lack passive, real-time social behavior analysis in therapeutic environments.

Purpose of the Study:

• To develop and evaluate a system for detecting and localizing group formations in a therapeutic environment.
• To enable passive, real-time monitoring of social behaviors for patient mental health.
• To create a scalable and low-cost solution using edge computing and camera networks.

Main Methods:

• Utilized a distributed edge-computing camera network within a 1700 m² therapeutic space.
• Employed a lightweight multiperson 2-D pose detection model for individual position estimation.
• Developed a group identification method to localize formations from noisy individual data.

Main Results:

• Achieved an F1 score of up to 90% for group identification.
• Demonstrated a mean absolute error of 1.25 m for group localization.
• Validated on a benchmark dataset with 2-7 subjects over 35 minutes.

Conclusions:

• The proposed system effectively identifies and localizes social group formations using edge computing.
• This technology offers a scalable, low-cost solution for transforming buildings into smart environments for health monitoring.
• The system can enhance therapeutic units by passively monitoring social behaviors and facilitating real-time interventions.