[Paper Review] A new PIR-based method for real-time tracking

Tianye Yang, Xuefeng Liu+3 more|arXiv (Cornell University)|Jan 30, 2019

Indoor and Outdoor Localization Technologies18 references4 citations

TL;DR

This paper proposes a novel PIR-based device-free localization method that leverages raw PIR sensor data to extract azimuth change—a physical-layer feature derived from the sensor's directional response—enabling high-accuracy real-time tracking without requiring extensive training data or dense sensor deployment. The approach achieves up to 50% higher accuracy than state-of-the-art methods using only four PIR sensors.

ABSTRACT

Pyroelectric infrared (PIR) sensors are considered to be promising devices for device-free localization due to its advantages of low cost, less intrusive, and the immunity from multi-path fading. However, most of the existing PIR-based localization systems only utilize the binary information of PIR sensors and therefore require a large number of PIR sensors and a careful deployment. A few works directly map the raw data of PIR sensors to one's location using machine learning approaches. However, these approaches require to collect abundant training data and suffer from environmental change. In this paper, we propose a PIR-based device-free localization approach based on the raw data of PIR sensors. The key of this approach is to extract a new type of location information called as the azimuth change. The extraction of the azimuth change relies on the physical properties of PIR sensors. Therefore, no abundant training data are needed and the system is robust to environmental change. Through experiments, we demonstrated that a device-free localization system incorporating the information of azimuth change outperforms the state-of-the-art approaches in terms of higher location accuracy. In addition, the information of the azimuth change can be easily integrated with other PIR-based localization systems to improve their localization accuracy.

Motivation & Objective

• To address the limitations of binary PIR-based localization systems, which require dense sensor deployment and suffer from low granularity.
• To reduce dependency on large-scale training data in data-driven PIR localization approaches.
• To improve robustness to environmental changes by leveraging physical properties of PIR sensors rather than learned patterns.
• To introduce a new location feature—azimuth change—derived from raw PIR output for enhanced localization accuracy.
• To enable practical, low-cost, privacy-preserving real-time tracking using minimal PIR sensors.

Proposed method

• The method extracts azimuth change from raw PIR voltage signals by analyzing the temporal variation in sensor output across multiple PIR sensors.
• It models the physical behavior of PIR sensors and Fresnel lens arrays to estimate the change in the angle of arrival (AoA) of infrared radiation from a moving person.
• The azimuth change is computed as the absolute difference in the estimated direction of a person relative to each PIR sensor at two consecutive positions.
• A particle filter is used to integrate azimuth change estimates over time for robust trajectory estimation.
• The system uses a cosine formula-based geometric model to relate azimuth change to physical distances between the person and PIR sensors.
• The approach avoids black-box learning by relying on physical modeling rather than data-driven training.

Experimental results

Research questions

• RQ1 Can raw PIR sensor data be used to extract a physical-layer feature—azimuth change—that improves localization accuracy without requiring training data?
• RQ2 How does the proposed azimuth change estimation compare to existing data-driven methods in terms of accuracy and robustness to environmental changes?
• RQ3 To what extent can a minimal number of PIR sensors (e.g., four) achieve high localization accuracy using azimuth change?
• RQ4 Can azimuth change alone or in combination with binary signals improve tracking performance in real-time?
• RQ5 How does the system perform under varying environmental conditions, such as presence of static or moving heat sources?

Key findings

• The proposed system achieves an accuracy rate of 0.54 and 0.78 in two experimental scenarios, representing approximately a 50% improvement over the state-of-the-art method.
• The system outperforms the data-driven method from [PIR Sensors Characterization and Novel Localization Technique] despite using only four PIR sensors instead of eight.
• The system requires no labeled training data, making it robust to environmental changes and easily deployable in new settings.
• Azimuth change estimation is most accurate when the person is not too close to the PIR sensor, with accuracy degrading at very short distances.
• The method demonstrates strong potential for integration into existing PIR-based systems to enhance their localization accuracy.
• The system shows sensitivity to environmental noise, particularly from static or moving heat sources, suggesting a need for dynamic thresholding in future work.