Towards Breathing as a Sensing Modality in Depth-Based Activity Recognition

Area of Science:

• Computer Vision
• Human-Computer Interaction
• Biomedical Engineering

Background:

• Depth cameras are increasingly used for activity recognition due to precise joint and posture detection.
• Advancements in depth camera resolution enable novel applications like remote physiological monitoring.
• Respiration monitoring is crucial for various health and activity recognition tasks.

Purpose of the Study:

• To develop a robust method for monitoring respiration using depth imaging by analyzing chest elevation.
• To assess the accuracy and robustness of the proposed method for activity recognition.
• To compare the performance against existing depth imaging techniques and commercial respiration monitors.

Main Methods:

• A novel method modeling chest elevation changes over time to detect respiration.
• Utilizing depth camera data to capture subtle torso movements indicative of breathing.
• Testing the method's efficacy in various scenarios, including non-sedentary individuals and partial torso occlusion.

Main Results:

• The method achieves high accuracy (92-97%) in detecting breathing rates from a distance of two meters.
• Performance rivals that of commercial respiration chest monitor belts.
• Outperforms state-of-the-art depth imaging methods, particularly for non-sedentary individuals.

Conclusions:

• Remote respiration monitoring via depth cameras is feasible and accurate.
• The proposed chest elevation modeling method offers a robust and effective approach for activity recognition.
• This technology has potential applications in health monitoring, fitness tracking, and human-computer interaction.