Privacy-concerned averaged human activeness monitoring and normal pattern recognizing with single passive infrared sensor using one-dimensional modeling

Abstract

Detecting human activity through cameras and machine learning methods raises significant privacy concerns, while alternatives like thermal cameras can be expensive. Passive infrared (PIR) sensors present a cost-effective and privacy-preserving solution, commonly used in home settings for motion detection. This study introduces a system for monitoring human activeness using a single PIR sensor, focusing on privacy preservation. The proposed one-dimensional model, based on the Laplace distribution, emphasizes the role of the parameter $\mu$ in defining velocity distributions. Through real-world experiments with a Raspberry Pi and PIR sensor, the effectiveness of the model in capturing human activeness is validated. The study investigates how different $\mu$ values correlate with activity levels and detect abnormalities. Additionally, the paper addresses the stochastic nature of human behavior, and the impact of $\mu$ on predictability and variability, and provides insights into detection thresholds and interval times. The findings highlight the potential for enhancing abnormality detection and suggest a comprehensive understanding of human activeness.