Human Indoor Localization Based on Ceiling
Mounted PIR Sensor Nodes
Xiaomu Luo
†∗
, Tong Liu
‡
, Baihua Shen
§
, Qinqun Chen
†
, Liwen Gao
†
and Xiaoyan Luo
†
†
School of Medical Information Engineering, Guangzhou University of Chinese Medicine, Guangzhou, China 510006
∗
Email: woodwood2000@163.com
‡
Department of Electronic Science, Huizhou University, Huizhou, China 516007
§
School of Information Engineering, GuangDong University of Technology, Guangzhou, China 510006
Abstract—This paper presents a human indoor localization
system using ceiling mounted pyroelectric infrared (PIR) sensors.
The field of views (FOVs) of the PIR sensors is modulated by two
degrees of freedom (DOF) of spatial segmentation. The localiza-
tion algorithm is proposed to fuse the data stream generated from
different sensor nodes within the wireless network. The Kalman
Filter and Kalman Smoother are utilized to refine the estimation
of the human position. We conduct experiments in a real office
environment, and the average root-mean-square error (RMSE)
of single human target tracking at different speed is about 0.6
meter. The promising results confirm the efficacy of our system.
I. INTRODUCTION
Recent years, indoor positioning systems (IPSs) have be-
come a growing field of research involving theoretical and ap-
plicative challenges [1]. Object detection and tracking are the
basis of many applications, such as surveillance and activity
recognition. Numerous techniques have been proposed. Active
badges is the first indoor location sensing system developed
by AT&T Cambridge [2], which belongs to the Infrared tech-
nology. To provide better and more accurate indoor position-
ing, they developed an ultrasonic tracking technology, name
bats [3]. Distributed Object Locating System for Physical
space Inter networking(DOLPHIN) [4] and Cricket [5] are
also ultrasonic tracking systems. To provide high temporal
resolution and accurate Time of Arrival (TOA) measurements
in multi-path environment, UltraWide band (UWB) impulse
radio signals are employed for indoor location and track-
ing [6]. Based on the Received Signal Strengh Information
(RSSI) attenuation, Radio Frequency Identification (RFID)
was employed to estimate the distances between transmitters
and receivers [7], and the position can be achieved with high
accuracy. These systems have been successfully used in many
applications such as asset tracking and inventory management.
To implement human indoor positioning, device-free tech-
niques are more suitable, because that means the human target
does not have to wear any device [8]. Recent advances in
CCD technologies, processing speed and image understanding
have been driving the development of the camera-based posi-
tioning systems. However, some disadvantages of continuous
surveillance by video cameras are difficult to overcome. It is
susceptible to light illumination, and its computational load
for continuous surveillance is high. Most importantly, it is
inevitable to violate the privacy, which will make the human
target feel uncomfortable.
Recently, more and more researchers employ the pyroelec-
tric infrared radial (PIR) sensors to achieve indoor positioning.
In [9], a hierarchical architecture of FOV spatial-modulated
scheme is employed to implement target tracking. The FOVs
of Fresnel lens array in a sensor node are modulated to achieve
a single degree of freedom (DOF) of spatial partition; then
the localization algorithm is proposed to coordinate multiple
sensors nodes to achieve two DOF spatial partitions. In [10],
PIR sensors, RFID and the outcome of a noise analysis
are combined to achieve multi-modal data fusion for indoor
localization. The PIR sensor is of low cost, and human motion
information is encoded into the low dimensional data streams
directly. PIR sensors will only provide sine-like wave output,
which means that they will not infringe the privacy, because
it is impossible to recover the image of the human target. It
has the promising advantages to overcome the drawbacks of
the camera-based method. The biggest challenge of the PIR
sensing system is the improvement of its spatial resolution.
In this paper, we proposed a human indoor localization
system based on ceiling mounted PIR sensor nodes. In our
system, five sensor nodes are utilized to form wireless sensor
networks (WSNs). In each sensor node, reference structure
[11] are used to modulate the FOV of each PIR sensor. So the
spatial information is embedded in the PIR signal triggered
by the human movement. Spatial information will be decode
based on the coding scheme. Then, the Kalman Filter and
the Kalman Smoother will be used to refine the location
estimation of the human target. We conduct experiments in
a real office environment, and the results confirm the efficacy
of our system.
II. S
ENSOR NODE DESIGN
A. Sensing model
To capture the spatio-temporal feature of the human tar-
get, the sensing model has to be designed deliberately. Our
model springs from the reference structure tomography (RST)
paradigm, which uses multidimensional modulations to encode
mappings between radiating objects and measurements [11].
The schematic diagram of our sensing model is shown in
Fig. 1. The object space refers to the space where the thermal
object moves. The measurement space refers to the space
2016 13th IEEE Annual Consumer Communications & Networking Conference (CCNC)
978-1-4673-9292-1/16/$31.00 ©2016 IEEE