Spatio-Temporal Constrained Human Trajectory Generation from the PIR Motion Detector Sensor Network Data: A Geometric...

Area of Science:

• Computer Vision
• Sensor Networks
• Robotics

Background:

• Passive infrared (PIR) motion detectors are crucial for continuous human motion analysis.
• Existing methods struggle to generate all possible human trajectories due to limited PIR sensor data (no location or individual info).

Purpose of the Study:

• To develop a geometric algebra (GA)-based approach for comprehensive human trajectory generation from PIR sensor networks.
• To overcome the limitations of current methods in extracting all possible spatio-temporal human motion paths.

Main Methods:

• Representing geographical networks, sensor activations, and human motion using GA.
• Labeling sensor activations with GA-based trajectory tracking.
• Employing matrix multiplication for dynamic trajectory generation based on sensor logs and constraints.

Main Results:

• The GA-based method successfully extracts major statistical patterns of human motion.
• It effectively generates all possible human trajectories, outperforming direct statistical analysis and tracklet graph methods in accuracy.
• The approach demonstrated flexibility in extracting motion patterns from the MERL motion database.

Conclusions:

• The developed GA approach provides a more accurate and comprehensive method for human trajectory extraction from PIR sensor data.
• This method offers a new pathway for filtering passive sensor log data in various sensor network applications.