Synthetic Generation of Passive Infrared Motion Sensor Data Using a Game Engine.

Area of Science:

• Computer Science
• Sensor Technology
• Human-Computer Interaction

Background:

• Accurate indoor occupancy counting is vital for various applications.
• Existing methods rely on passive infrared (PIR) motion sensor data and supervised learning.
• Acquiring labeled datasets for PIR motion sensing and ground truth people counts is labor-intensive and time-consuming.

Purpose of the Study:

• To propose a method for synthetically generating labeled PIR motion sensor data for occupancy counting.
• To develop and validate a simulator for producing realistic PIR sensor data.
• To investigate the spatial non-linearity of PIR sensor responses.

Main Methods:

• Development of a simulator in the Unity game engine to generate synthetic PIR motion sensor data.
• Replication of a real-world meeting room within the simulator.
• Experimental validation by comparing simulated PIR sensor data with real-world data during choreographed movements.

Main Results:

• The developed simulator achieved comparable output to real-world PIR sensors in 34 out of 50 tested scenarios.
• The study revealed a highly non-linear relationship between PIR sensor output and the spatial position of motion.
• The simulator successfully generated synthetic PIR motion sensor data by detecting simulated occupants.

Conclusions:

• Synthetic data generation using the developed simulator offers a viable alternative to time-consuming real-world data collection for PIR-based occupancy counting.
• The identified non-linear relationship between sensor output and motion location highlights challenges in directly mapping PIR data to occupancy counts.
• The simulator provides a valuable tool for studying PIR sensor behavior and advancing occupancy sensing technologies.