sensors

Article
Passive Infrared (PIR)-Based Indoor Position
Tracking for Smart Homes Using Accessibility
Maps and A-Star Algorithm
Dan Yang 1 , Bin Xu 1, *, Kaiyou Rao 2 and Weihua Sheng 3
1 School of Information Science and Engineering, Northeastern University, Shenyang 110819, China;
yangdan@mail.neu.edu.cn
2 School of Physics, Northeast Normal University, Changchun 130000, China; raoky290@nenu.edu.cn
3 School of Electrical and Computer Engineering, Oklahoma State University, Stillwater, OK 74074, USA;
weihua.sheng@okstate.edu
* Correspondence: xubin@mail.neu.edu.cn

Received: 13 December 2017; Accepted: 22 January 2018; Published: 24 January 2018

Abstract: Indoor occupants’ positions are significant for smart home service systems, which usually
consist of robot service(s), appliance control and other intelligent applications. In this paper,
an innovative localization method is proposed for tracking humans’ position in indoor environments
based on passive infrared (PIR) sensors using an accessibility map and an A-star algorithm, aiming at
providing intelligent services. First the accessibility map reflecting the visiting habits of the occupants
is established through the integral training with indoor environments and other prior knowledge.
Then the PIR sensors, which placement depends on the training results in the accessibility map,
get the rough location information. For more precise positioning, the A-start algorithm is used to
refine the localization, fused with the accessibility map and the PIR sensor data. Experiments were
conducted in a mock apartment testbed. The ground truth data was obtained from an Opti-track
system. The results demonstrate that the proposed method is able to track persons in a smart home
environment and provide a solution for home robot localization.

Keywords: smart home; indoor position tracking; PIR sensors; accessibility map; A-start algorithm

1. Introduction
The elderly population is rapidly increasing worldwide. In 2050, the number of elderly people
(60 and above) is forecast to reach a staggering 2 billion [1]. For personal comfort and due to limited
medical resources, most of them live alone in their own houses instead of nursing homes. Elders usually
prefer to stay in the comfort of their home where they feel more confident than moving to an expensive
adult care or healthcare facility. Hence, how to help them to live conveniently and safely has become
an important social issue [2]. Real-time indoor human monitoring plays an important role in assisted
living, emergency detection, among many other aspects. A fundamental problem in human monitoring
is how to localize humans in indoor environments.
Researchers have developed various techniques to localize residents in indoor environments [3].
Active Badge [4] was developed at the Olivetti Research Laboratory using diffuse infrared technology.
It employs a network of sensors placed around the building with each room having at least one sensor.
The subjects must carry a tag that can emit a unique code for approximately a tenth of a second every
15 s (a beacon) in order for the sensors to detect it. A master station, also connected to the network,
polls the sensors for badge detections, processes the data, and then makes it available to clients that
can display it in a useful visual form. When a sensor receives signals from the tag, the human was
known to be in that room. Cricket [5] is an example that uses ultrasound technology to localize

Sensors 2018, 18, 332; doi:10.3390/s18020332 www.mdpi.com/journal/sensors

Sensors 2018, 18, 332 2 of 12

people. The Cricket indoor location system consists of location beacons and listeners. Beacons are
transmitters attached to the ceiling of buildings and receivers called listeners are attached to the
devices (or people) requiring location information. Active beacons transmit location information and
an ultrasonic pulse. The passive listeners capture this information to calculate its distance from the
beacon. This method would work for a large office environment. In [6], WiFi Received Signal Strength
(RSS) fingerprints were used to detect persons’ location and estimate their velocity. Signal strength
attenuation will cause lower localization accuracy. In [7], the authors’ solution automatically identifies
the rooms where the smart objects are placed, by using the information contained within the RSS,
the knowledge of the map of the house and the use of a fuzzy inference system without any a priori
knowledge of the positions of the anchor nodes (ANs). In [8] multiple cameras and floor sensors were
used to localize a human in a smart home environment. The system consisted of three components:
camera localization, sensory floor localization and condensation tracker. The presence of a person was
detected by the sensory floor by measuring the pressure variation. The cameras detected the person by
using background subtraction and human template matching. Obviously, a large number of sensors
will increase the cost of installation and there are always privacy concerns when cameras are used.
In [9], they used ultrawideband (UWB) technology for indoor positioning. They presented an IPS based
on UWB, which consists of four fixed transmitters and some mobile users. The transmitters, which are
synchronized using a time division scheme, send a UWB signal to the mobile users. The estimation of
location was calculated using a triangulation method based on the measure the Time Difference of
Arrival (TDoA) between transmitters and receivers. In [10], they make used of Earth’ natural magnetic
field orientation to perform the localization process. They measures location using disturbances of
Earth’s magnetic field caused by structural steel elements in a building. Magnetic maps are developed
taking measurements along corridors and landmarks and then places identified based on the magnetic
signature. Ojeda et al. [11] developed a Zero Velocity Update (ZUPT)-based method that compensates
the accelerometer bias and achieves an error of about 2% for short walks. These aforementioned
methods all have their own limitations when providing a practical personal-tracking solution.
Recently, passive infrared (PIR) sensors have been used to detect and localize humans because
of their simplicity and less privacy concerns. Luo and Chen [12] proposed a technique to track the
path of a human using PIR sensors. An array of PIR sensors was placed in a room so that the sensing
areas of two or more sensors overlap. The experiment was conducted on a test bed in which a total of
12 sensors are placed to ensure the maximum overlap and the full coverage of the test area. A location
accuracy of 0.5 m was achieved by the system. The accuracy of these PIR-based location systems
greatly depends on the distribution density of sensors and the intersection of the field of views (FOVs)
of the sensors. To reduce sensory uncertainty, multi-sensor techniques and data fusion algorithms are
usually adopted. In [13], Yang et al. developed a low-cost and small-size human tracking system based
on PIRs and wireless sensor network. A localization method based on detecting angle bisectors of PIR
sensors and data fusion was presented. Then, Kalman filtering and particle filtering were employed
for target tracking. In [14], Al-Naimi presented an automatic identification and tracking method by
combining data from PIR sensors and floor pressure sensors. In [15], a method using PIR and RF for
localization was proposed to deal with the RSS variation issue that caused a significant localization
error. PIRs were used to identify the rough area, then the position of the person was estimated by
applying the K Nearest Neighbor (K-NN) algorithm to the fingerprints inside that area. That method
can reduce the localization error. In [16], they proposed a region-based human tracking algorithm
based on the output signals of several PIR sensors. A mathematical abstraction of a PIR sensor as a
building block for the algorithm is provided. In [17], a human indoor localization system based on
ceiling mounted PIR sensor nodes was proposed. In the system, five sensor nodes are utilized to form
a wireless sensor network (WSN). The spatial information is embedded in the PIR signal triggered by
the human movement and is decoded based on the coding scheme. The Kalman Filter and the Kalman
Smoother are used to refine the location estimation of the human target.
Sensors 2018, 18, 332 3 of 12
Sensors 2018, 18, x FOR PEER REVIEW 3 of 12

In this
In this paper,
paper, we we propose
propose aa new
new solution
solution toto PIR-based
PIR-based indoor
indoor human
human localization
localization byby using
using anan
A-algorithm and a grid-based accessibility map. In our solution, the grid-based
A-algorithm and a grid-based accessibility map. In our solution, the grid-based accessibility map is accessibility map is
built which
built which reflects
reflects human
human evaluation
evaluation of of the
the degree
degree of of accessibility
accessibility inin aa smart
smart home.
home. We distinguish
We distinguish
these areas based on obstacles (walls), furniture, unoccupied free space, accessibility-related
these areas based on obstacles (walls), furniture, unoccupied free space, accessibility-related daily daily
habits, and
habits, andsosoforth. PIRPIR
forth. sensors data data
sensors give agive
rough estimate
a rough of the human
estimate of thelocation
human without
locationcapturing
without
any user images. They are deployed according to the area accessibility as
capturing any user images. They are deployed according to the area accessibility as the standard. the standard. To further
improve the accuracy, we also exploit the A-star algorithm for indoor
To further improve the accuracy, we also exploit the A-star algorithm for indoor human tracking. human tracking. First, the
continuous
First, triggering triggering
the continuous of two PIRof sensor
two PIRareas are defined
sensor areas areas the givenas
defined initial node and
the given goal
initial node
node of
and
A-star algorithm. The weight function will be as the path-cost evaluation
goal node of A-star algorithm. The weight function will be as the path-cost evaluation function. function. Eight candidate
state nodes
Eight candidateare state
selected
nodesaround the current
are selected around initial node. initial
the current Their node.
weightTheir
coefficients are calculated
weight coefficients are
according to their direction and distance related the current initial node, and
calculated according to their direction and distance related the current initial node, and corresponding corresponding area
accessibility
area value.
accessibility The target
value. nodenode
The target is theismaximum
the maximum of theofweight function.
the weight The A-star
function. algorithm
The A-star fuse
algorithm
the PIR
fuse sensor
the PIR datadata
sensor andandgrid-based
grid-basedaccessibility
accessibilitymap map information
informationtotoobtain
obtainmore
moreaccurate
accurate position
position
estimate. The rest of the paper is organized as follows: Section 2 details our
estimate. The rest of the paper is organized as follows: Section 2 details our tracking methodology, tracking methodology,
including how
including how to tobuild
buildananaccessibility
accessibilitymapmapforfora specific
a specific home,
home, thethe details
details of the
of the A-star
A-star algorithm
algorithm for
tracking and the PIR sensor model. Section 3 will present the experiments results. Section 4 concludes4
for tracking and the PIR sensor model. Section 3 will present the experiments results. Section
concludes
this this paper
paper and suggests andfuture
suggests
work.future work.

2. Methodology
2. Methodology

2.1. Grid-Based Accessibility Map Building

In this
this paper,
paper, we weuseusea agrid-based
grid-based accessibility
accessibility mapmap to model
to model the tracking
the tracking environment.
environment. The
The grid-based
grid-based accessibility
accessibility map map is subdivide
is subdivide into ainto a finite
finite position
position grid
grid for theforgiven
the given environment.
environment. Each
Each grid provides
grid provides two two pieces
pieces of information.
of information. TheThe firstfirst
oneone is the
is the locationofofthe
location thegrid
gridin
inthe
the environment
environment
and the second one is its probability of a person being present at that location. Detailed explanations
are presented in the following
following paragraphs.
paragraphs.AAmock mockapartment
apartment(12 (12× × 2 ) used to test our ideas,
7.27.2
m2m ) used to test our ideas, as
as shown
shown as as Figure
Figure 1. 1.
TheThe functionAA_map
function _ map( x(,x,y )y)isisdefined
definedforfor describing
describing thethe accessibility
accessibility value at the
location (x, y)
location (x, y) in 2D x-y
in 2D x-y plane
plane of
of the
the given
given environment.
environment.

Figure 1. The floor plan of the mock apartment.

Figure 1. The floor plan of the mock apartment.

The different colors are used to express different regions. The blue areas and the black areas
The different colors arewalls,
used respectively.
to express different regions. Thebeblue areas
in,and
and the black areas
denote the furniture and the These areas cannot walked A _ map ( xt , yt )
denote the furniture and the walls, respectively. These areas cannot be walked in, and A_map( xt , yt )
is set as “0”. Free space areas have no obstacle for access, so we store “0.5” as the original value. The
is set as “0”. Free space areas have no obstacle for access, so we store “0.5” as the original value.
original two-dimensional floorplan is converted into a 3D depth map according to the furniture
The original two-dimensional floorplan is converted into a 3D depth map according to the furniture
layout, as shown as Figure 2.
layout, as shown as Figure 2.
 Sensors 2018, 18, 332 4 of 12
Sensors 2018,
Sensors 18,18,
2018, x FOR PEER
x FOR REVIEW
PEER REVIEW 4 of 12 12
4 of

Figure 2. 2.
Figure
Figure The initial
2.The
The value
initial value
initial of of
value the
oftheA _A
the map
A_map( x(t(x,xyt, ,ty)y)t. .) .
_ map
t t

ToTobuild the
build accessibility
the accessibility map,
map, wewe
useuseanother
another source
source ofof
information,
information, i.e., the
i.e., thedaily visiting
daily visiting habits
habits
To build the accessibility map, we use another source of information, i.e., the daily visiting habits
ofof
human
human subjects
subjectswhen
when they
theyinteracts
interacts with
with the environment.
the environment. Generally,
Generally, thethehuman
human subject
subject visits
visits
of human subjects when they interacts with the environment. Generally, the human subject visits
more
more frequently
frequently around
around certain
certainfurniture.
furniture. The
The mock
mock smart
smart home
homewewe designed
designed hashassome
some keykeyfurniture:
furniture:
more frequently around certain furniture. The mock smart home we designed has some key furniture:
bedroom
bedroom bed, living
bed, livingroom
room couch,
couch, dining
diningroom
room chair, living
chair, livingroom
room table, kitchen
table, kitchen basin
basinandand bathroom
bathroom
bedroom bed, living room couch, dining room chair, living room table, kitchen basin and bathroom
basin.
basin.TheThefurniture
furniture locations
locationsininthe indoor
the indoor environment
environment areare
known.
known. The
The areas
areas around
around thethefurniture
furniture
basin. The furniture locations in the indoor environment are known. The areas around the furniture
has
hasmore
morevisiting
visitingprobability
probabilitythan thanother
otherinner
innerareas
areasininthetheindoor
indoorenvironment,
environment,asasshown showninin
has more visiting probability than other inner areas in the indoor environment, as shown in Figure 3a.
Figure
Figure 3a.3a.
Applying
Applying a Gaussian
a Gaussian kernel
kernelalso yields
also yieldssmooth
smooth transitions
transitionsbetween
between the more
the more visited
visitedareas
areas
Applying a Gaussian kernel also yields smooth transitions between the more visited areas and the
and
andthethesurrounding
surrounding areas,
areas,asas
described
described ininFigure
Figure3b.3b.
surrounding areas, as described in Figure 3b.

(a)(a) (b)(b)

Figure 3. 3.
Figure
Figure Hot
3. Hot
Hotareas related
areas
areas with
related
related with
with the furniture
the
the layout:
furniture
furniture (a)(a)
layout:
layout: hot
(a) areas
hot
hot related
areas
areas with
related
related furniture;
with
with (b)(b)
furniture;
furniture; after
(b) after
after
processing
processing
processing with
withthe
with Gaussian
the
the Gaussian
Gaussianfunction.
function.
function.

Besides
Besidesthese considerations,
these considerations, each
each twotwolandmarks
landmarks can
cangenerate
generate ananarbitrary
arbitrary route. The
route. The visiting
visiting
Besides these considerations, each two landmarks can generate an arbitrary route. The visiting
probability
probability ofof
some
some routes
routesis is
based
basedonon thethe
observations
observations about
aboutthe specific
the specificperson’s
person’s daily habits.
daily The
habits. The
probability of some routes is based on the observations about the specific person’s daily habits.
starting
startingand
andending
ending landmarks
landmarks can
canbebe(p1) entrance
(p1) entrancetoto
kitchen,
kitchen,(p2) kitchen
(p2) kitchentoto
living
living room
roomtable, (p3)
table, (p3)
The starting and ending landmarks can be (p1) entrance to kitchen, (p2) kitchen to living room table,
major
majorbedroom
bedroom toto
guest
guest room,
room, (p4) living
(p4) livingroom
roomsofa
sofatotobathroom,
bathroom, and
and (p5) guest
(p5) guest room
room toto
balcony
balconyasas
(p3) major bedroom to guest room, (p4) living room sofa to bathroom, and (p5) guest room to balcony
shown
shown inin
Figure
Figure 4. 4.
Tests
Testswith
with different
different trajectories
trajectorieswere
were processed.
processed. The
Theareas
areas ininthese trajectories
these trajectories
as shown in Figure 4. Tests with different trajectories were processed. The areas in these trajectories
have
havemore
more accessible
accessible frequencies
frequencies than
thanother
otherareas ofof
areas thethe
apartment.
apartment. These
Thesevalues
values cancanbebe
obtained
obtained byby
have more accessible frequencies than other areas of the apartment. These values can be obtained
observation.
observation.
by observation.
Sensors 2018, 18, 332 5 of 12
Sensors 2018, 18, x FOR PEER REVIEW 5 of 12

Sensors 2018, 18, x FOR PEER REVIEW 5 of 12

(a) (b)

(a) (b)

(c) (d)

(c) (d)

(e) (f)
Figure 4. The accessible(e) area heat map with several training. (a) result with(f) p1 training; (b) result with
Figure 4. The accessible area heat map with several training. (a) result with p1 training; (b) result with
p2 training; (c) result with p3 training; (d) result with p4 training; (e) result with p5 training; (f) added
p2 training; (c)The
Figure 4. result with p3
accessible training;
area heat map(d) result
with with
several p4 training;
training. (e)with
(a) result result
p1with p5 training;
training; (b) result (f)
withadded
results with p1-p5 40 times training.
p2 training; (c) result with p3
results with p1-p5 40 times training. training; (d) result with p4 training; (e) result with p5 training; (f) added
results with p1-p5 40 times training.
After all these processing steps, the accessiblity map in 2D view is shown in Figure 5.
AfterAfter
all these processing
all these steps,
processing steps,the
theaccessiblity mapin
accessiblity map in2D
2Dview
viewis is shown
shown in Figure
in Figure 5. 5.

.
.
Figure 5. The accessible map considering the daily visiting route.
Figure
Figure 5. The
5. The accessiblemap
accessible map considering
considering the
thedaily
dailyvisiting route.
visiting route.
Sensors 2018, 18, 332 6 of 12
Sensors 2018, 18, x FOR PEER REVIEW 6 of 12

Based
Based onon this
this prior
prior knowledge,
knowledge, thethe accessibility
accessibility map map can can bebe represented
represented as as aa heatmap
heatmap of of the
the
counts of the times that the assumed occupant visits respective grid nodes. Each grid size is ∆x×∆y,
counts of the times that the assumed occupant visits respective grid nodes. Each grid size is  x y,
and ∆x =
and Δx ∆y == 0.1
= Δy 0.1 m.
m. The
The binary
binary image
image of of accessible
accessible map map is is shown
shown as Figure 6.
as Figure 6. Let
Let N
N denote
denote the
the
number
number ofof visitable
visitable grid
grid nodes,
nodes, and
and hhii denotes
denotes the
the “heat”
“heat” score score of of the
the i-th grid node.
i-th grid node. Therefore
Therefore the
the
accessibility
accessibility map
map can
can be
be defined as aa set
defined as set of values {{hh11,,...,
of values . . . ,hhNN},}0
, 0<hihi <
255 .
255.

Figure
Figure 6.
6. The
The grid-based
grid-based grey
grey image
image of
of accessible
accessible map.
map.

2.2. Algorithm Design

2.2. Algorithm Design
The A-Star algorithm uses the method of best-first search and finds an optimal path from the
The A-Star algorithm uses the method of best-first search and finds an optimal path from the initial
initial node to the destination node [18]. It uses a distance and a cost heuristic function (usually
node to the destination node [18]. It uses a distance and a cost heuristic function (usually denoted
denoted by f ( n) ) to determine the order in which the search visits nodes in the tree. The
by f (n)) to determine the order in which the search visits nodes in the tree. The distance-plus-cost
distance-plus-cost
heuristic is a sum of heuristic is a sum(i)
two functions: of the
twopath-cost
functions: (i) the path-cost
function, which isfunction, which
the cost from is starting
the the cost from
node
the starting node to the current node (denoted by g ( n ) ) and (ii) an admissible “heuristic
to the current node (denoted by g(n)) and (ii) an admissible “heuristic estimate” of distance to the goal estimate”
of distance
(denoted n)):goal (denoted by h( n) ):
bytoh(the
f (n) = g(n) + h(n) (1)
f (n)  g ( n)  h(n) (1)
The A-Star algorithm creates a tree of nodes and maintains two lists, an OPEN list and a CLOSED
The A-Star algorithm creates a tree of nodes and maintains two lists, an OPEN list and a CLOSED
list. The OPEN list is a priority queue and keeps track of those nodes that need to be examined,
list. The OPEN list is a priority queue and keeps track of those nodes that need to be examined, while
while the CLOSED list keeps track of nodes that have already been examined. Each node n maintains
the CLOSED list keeps track of nodes that have already been examined. Each node n maintains f ( n) ;
f (n); intuitively, this is the estimate of the best solution that goes through n. If the initial node and
intuitively, this is the into
node n are combined estimate of thethen
one node, best the
solution that goesis:through n. If the initial node and node n
cost function
are combined into one node, then the cost function is:
ff ((nn)) ≈
 hh((nn)) (2)
(2)

Figure 77shows
showsthe theprinciple of next
principle position
of next estimation
position basedbased
estimation on theon A-star
the algorithm for tracking
A-star algorithm for
in the given
tracking in thesmart
given home.
smartThe start
home. node
The is node
start denoted as the original
is denoted position
as the original at the entrance
position of the
at the entrance
room.
of At time
the room. Att,time
the person is located
t, the person at the current
is located node. The
at the current blue
node. Thecircle
bluearea is area
circle the current trigger
is the current
PIR sensing
trigger area, the
PIR sensing “Goal
area, Node”Node”
the “Goal is the is
node
the with
node maximum
with maximum heat score of theofcurrent
heat score trigger
the current PIR
trigger
sensing
PIR area.area.
sensing The eight candidate
The eight N_1 ∼N N_8
nodesnodes
candidate _1 ~ for
N _ time
8 fort + 1 aret selected
time 1 are selected
around around
the current
the
node. The target node is determined by the following equation:
current node. The target node is determined by the following equation:
AA_map
_ map((xxnn,,yynn))
h( xnh,(yxnn), =
yn )wdirw·dirwh_d
wh ·_ dq , n, 
n 1=~18 ∼ 8 (3)
(3)
22 2 2
( x - x
( xn n− xGoal )  ( y - y
Goal ) + ( ynn −Goal )
yGoal )

where A _ map
where A_map ( x(nx,ny, ny)n )denotes
denotes
thethe “heat”
“heat” score
score forcandidate
for the the candidate node nnode n of position
of position ( xn , yn )
( xn , yn ) obtained
from the grid-based
obtained accessibilityaccessibility
from the grid-based map. The goal
map. The( xgoal
node yGoal ) (isxGoal
Goal , node the ,node
yGoalwith
) is the maximum
node withheat the
score of the triggered PIR sensing area. The weight coefficient wh_d represents the effect of heat score
maximum heat score of the triggered PIR sensing area. The weight coefficient wh _ d represents the
and distance between the target node and goal node. The direction weight wdir is determined by the
effect of heat score and distance between the target node and goal node. The direction weight wdir
 Sensors 2018, 18, 332 7 of 12
Sensors 2018,2018,
Sensors 18, x 18,
FOR PEERPEER
x FOR REVIEW
REVIEW 7 of 12
7 of 12

is relative
determined
is determined by the
position by relative
between
the relative position
the eight
positionbetween
candidate
betweenthe eight
nodes thewith candidate
eightthe current
candidate nodes withwith
estimated
nodes the current
direction.
the current estimated
For the time
estimated
direction.
no PIRFor
ofdirection. the
sensor time of
triggered,no PIR
the sensor
estimated triggered,
direction theis estimated
from the direction
current node is
For the time of no PIR sensor triggered, the estimated direction is from the current node from
to the the
point current
with node
maximum to to
the point
“heat”
the withwith
score
point inmaximum
themaximum “heat”
next triggering score
“heat” PIR in the
sensor
score next
area.
in the triggering
nextFortriggering
the timePIRofPIR
sensororarea.
onesensor
twoarea.For sensors
PIR the the
For timeare oftriggered,
time one or or
of one
twothe PIR
two sensors
estimated
PIR are are
sensors triggered,
direction is from the
triggered, the theestimated
current node
estimateddirection
todirectionis from
the point iswith the
from current
maximum
the currentnode
“heat”
nodetoscore
the point
to in
thethe withwith
current
point
maximum
triggering
maximum “heat”
PIR score
sensor
“heat” in the
area
score in current
during thetriggering
the current first PIR PIR
half part
triggering sensor ofarea
timesensor during
PIRarea
sensor the first
triggered.
during half
During
the first partpart
half time
the timeof PIR
latter half
of PIR
sensor
part triggered.
of
sensor the timeDuring
triggered. PIRthe
theDuring latter
sensor
the is half partpart
triggered,
latter half ofthe
the timetime
ofdirection
the theisthe
PIR
fromsensor
PIR the is triggered,
current
sensor node the
is triggered, direction
to the
the point is is
with
direction
from the current
maximum
from node
“heat”
the current to the
score
node in point
tothe
thenext with
point maximum
triggering
with PIR“heat”
maximum sensor score
area.
“heat” in the
in next
Suppose
score triggering
thethat
next the subject
triggering PIRwho
sensor
PIR area.area.
is tracked
sensor
Suppose
moves
Supposethat the subject
at a that
specific
the subjectwhowho
velocity, is tracked
then is the moves
time
tracked at atracked
of the
moves specific
at velocity,
route
a specific can then the time
be denoted
velocity, then theas ofa the
time of tracked
grid number
the route
tracked in route
the
can be denoted
grid-based
can as aasgrid
accessibility
be denoted number
a grid map.number in the in grid-based
the grid-based accessibility map.
accessibility map.

Figure 7. Diagram
Figure
Figure 7. 7.
Diagramof of
Diagramtarget
of node
target
target estimation
node
node byby
estimation
estimation the
by A-star
the
the algorithm.
A-star
A-star algorithm.
algorithm.

2.3. PIR
2.3.
2.3. Sensor
PIR
PIR Model
Sensor
Sensor Model
Model
InIn
the ideal
Inthe
the
idealPIR
idealPIRsensing
PIRsensingmodel,
sensing model,
model, a PIR sensor
a PIR
a PIR reports
sensor
sensor “detected”,
reports
reports which
“detected”,
“detected”, means
which
which there
means
means is is
there
there a isperson
a person
a person
within
withinthe
within sensor
thethe
sensor detection
sensor
detection region
detection (thecircle
region
region (the circle
(the region
circle
region with
region
with the
with
the radius
the
radius of R
theofR
Rradius the
PIR PIR PIR
sensor).
ofsensor).
the For For
sensor).
For a a
a practical
practical
PIR PIR PIR
practical
sensor, sensor,
there isthere
sensor,
a timeis delay
therea is
time delay
a time
when awhen
delay
PIRwhena PIR
sensor sensor
a PIR detects
sensor
detects the so
the detects
target, target,
thewe so we
target, so adefine
define we apractical
define
more more
a more
practical
model model
practical
of the of the
model
PIR of PIR sensors
the PIR
sensors as shown
sensors
as shown inasFigure
shownin 8.
Figure 8. 8.
in Figure

t t
t2 t24 4

R1 R1

R2 R2
t1 t1
t0 t0

Figure 8. 8.
Figure
Figure PIR8.
PIRSensor
PIR Model.
Sensor
Sensor Model.
Model.

AA person
Aperson
personis isalways
isalways detected
always within
detected
detected ananinner
within
within aninner disk
inner ofofradius
disk
disk of R1Rwhen
radius
radius R1 when moving moving in the sensor’s
in the sensor’s
1 when moving in the sensor’s
detection area,
detection while
area, the
while PIR
the sensor
PIR physical
sensor sensing
physical area
sensing isisthe
area is outer
thethe disk
outer ofofradius
disk of RR
radius2. Time t0–t4t0–t4
R2. Time
detection area, while the PIR sensor physical sensing area outer disk radius 2 . Time t0 –t4
represents the
represents time
the of
time human
of humanin moving.
in Time
moving. t
Time 0 tot0t1to
and
t t
1 and2 tot2t4to
the
t person
4 the
represents the time of human in moving. Time t0 to t1 and t2 to t4 the person has already been in person has already
has been
already in
been the
inthe
the
PIR
PIRsensor
PIRsensor area,
sensor area,but
area, but the
but
thePIR
the sensor
PIR
PIR reports
sensor
sensor “no
reports
reports “no person”.
“no person”.
person”. Only
Only from
Only time
from
from t
time
time 1 to t
t1 tto , does the PIR sensor
1 to t2, does the PIR sensor
2
t2 , does the PIR sensor
data reflect
data a “true
reflect a “measurement.
“true “measurement.
data reflect a “true “measurement.

3. Experiments andand
3. Experiments Results
Results
Sensors 2018, 18, 332 8 of 12

Sensors 2018, 18, x FOR PEER REVIEW 8 of 12

3. Experiments and Results
3.1. PIR Sensor Parameters
3.1. PIR Sensor Parameters
The x-y plane area of the given smart home environment is 12 m × 7.2 m2. The thickness of wall
The x-y plane area of the given smart home environment is 12 m × 7.2 m2 . The thickness of
is 0.2 m, the door of the kitchen and bathroom space is 0.2 × 1.0 m2, and other door spaces are
wall is 0.2 m, the door of the kitchen and bathroom space is 0.2 × 1.0 m2 , and2 other door spaces
0.2 × 1.0 m2. The total possible coverage area for the PIR sensors is 75.12 m . The number and
are 0.2 × 1.0 m2 . The total possible coverage area for the PIR sensors is 75.12 m2 . The number and
deployment of PIR sensors are defined according to our previous work [19]. PIR sensor parameters
deployment of PIR sensors are defined according to our previous work [19]. PIR sensor parameters
are listed in Table 1. For the free space in the smart home environment, the degree of accessibility for
are listed in Table 1. For the free space in the smart home environment, the degree of accessibility
a given area is obtained by training. The whole apartment floorplan is gridded by 8640 points. The
for a given area is obtained by training. The whole apartment floorplan is gridded by 8640 points.
distance between two points is 0.1 m. The blue areas represent the obstacles. The PIR sensor
The distance between two points is 0.1 m. The blue areas represent the obstacles. The PIR sensor
deployment, room layout in 2D view and PIR sensing areas are shown in Figure 9.
deployment, room layout in 2D view and PIR sensing areas are shown in Figure 9.
Table 1. PIR sensor parameters.
Table 1. PIR sensor parameters.
Parameter Value
No. ofParameter
PIR sensors 10
Value
Number of PIR sensing
No. of PIR sensors area 14
10
Number of PIRcoverage
Maximum sensing ratio
area cmax 14
0.4182
Maximum coverage ratio cmax 0.4182
Actualcoverage
Actual ratec actcact
coveragerate 0.3903
0.3903
Overlap
Overlap ococ
raterate 0.0714

Figure
Figure 9.
9. PIR
PIR deployment
deployment and
and sensing
sensing area.
area.

3.2.
3.2. Hardware
Hardware Platform
Platform
The
The testbed
testbed contains
contains aa motion
motion capture
capture system
system that
that provides
provides the the ground
ground truth
truth of
of human
human location,
location,
shown
shown in Figure 10. A human subject wearing a cap with reflective markers on itasked
in Figure 10. A human subject wearing a cap with reflective markers on it was to move
was asked to
around the mock
move around the apartment, enabling
mock apartment, the head
enabling the location to be tracked
head location by theby
to be tracked OptiTrack system.
the OptiTrack This
system.
is theisOptiTrack
This system
the OptiTrack manufactured
system by Natural
manufactured Point Inc
by Natural Point(Corvallis, OR, USA).
Inc (Corvallis, OR, [16]
USA).and it has
[16] and12 it
V100:R2 cameras. Special reflective markers are placed on the human body
has 12 V100:R2 cameras. Special reflective markers are placed on the human body to track humanto track human subjects.
We used We
subjects. EKMC1601111
used EKMC1601111units (Panasonic ElectricElectric
units (Panasonic Works,Works,
Tokyo, Japan)
Tokyo, as the
Japan) as PIR sensor
the PIR sensorcells. A
cells.
hand-made cylindrical lens hood was used to narrow the detection area of each
A hand-made cylindrical lens hood was used to narrow the detection area of each sensor. The PIR sensor. The PIR data
was
data read everyevery
was read 0.05 s0.05
from each each
s from sensor (sampling
sensor rate of
(sampling rate20ofHz). The radius
20 Hz). of the
The radius ofnarrowed
the narrowed PIR
sensing
PIR sensingareaarea
is less than
is less thethe
than PIRPIR
sensing inner
sensing innerradius
radiusR1R. This can offset the effect of PIR sensor
1 . This can offset the effect of PIR sensor
response
response time. A ZigBee module is used for transmitting the PIR sensor
time. A ZigBee module is used for transmitting the PIR sensor data.
data. Based
Based onon these
these data,
data,
the server PC estimates the human’s location in the apartment and derives the
the server PC estimates the human’s location in the apartment and derives the movement trajectory movement trajectoryof
of the human.
the human.
Sensors 2018, 18, 332 9 of 12
Sensors 2018, 18, x FOR PEER REVIEW 9 of 12

Sensors 2018, 18, x FOR PEER REVIEW 9 of 12

Figure10.
Figure 10.Testbed
Testbedwith
with Opti-track
Opti-track System
System and
andPIR
PIRsensors.
sensors.
Figure 10. Testbed with Opti-track System and PIR sensors.
3.3.3.3. Human
Human Tracking
Tracking
3.3. Human Tracking
ToTo evaluate
evaluate our method, we pre-defined two trajectories. For route1, the human subject entered
our method, we pre-defined two trajectories. For route1, the human subject entered
the apartment triggering the PIR at the front door and the system started performing localization and
To evaluate
the apartment our method,
triggering the PIR weatpre-defined
the front door two trajectories.
and the system For route1,
startedthe human subject
performing enteredand
localization
tracking. He is asked to have a rest at the couch, than go to the bedroom to change clothes, and finally
the apartment
tracking. triggering
He is asked to havethe PIR
a rest at at
thethe
front
couch,doorthanand gotheto
system started performing
the bedroom localization
to change clothes, andand finally
he goes to the other bedroom to relax. For the route2, the person is coming home after shopping, and
tracking. He is asked to have a rest at the couch, than go to the
he goes to the other bedroom to relax. For the route2, the person is coming home after shopping,bedroom to change clothes, and finally
he first goes to the kitchen for preparing the meal, then has a dinner at the table at the living room,
he he
and goes to goes
first the other
to thebedroom
kitchento forrelax. For thethe
preparing route2,
meal,the person
then has aisdinner
comingathome afteratshopping,
the table the livingand room,
and at last goes to the bathroom and washes his hands.
he first
and at last goes to the kitchen for preparing the meal, then has a dinner at the table at the living room,
For goes to the the
the route1, bathroom
beginning andofwashes
the route his ishands.
the “Start Node’, the Goal Node is the point with
andFor at last goes to the bathroom and washes his ishands.
maximum heat score in the PIR sensor area (s3). the
the route1, the beginning of the route “Start Node’,
According the Goal
to Equation (3), Node
the cost is function
the pointh(n) with
For the route1, the beginning of the route is the “Start Node’, the Goal Node is the point with
maximum heat score in the PIR sensor area (s3). According to Equation
value of the eight candidate estimated points are calculated. The next moment position is the point (3), the cost function h(n) value
maximum heat score in the PIR sensor area (s3). According to Equation (3), the cost function h(n)
of the
witheight candidateh(n)
the maximum estimated
value. T3 points
denotesare calculated.
the triggering Thetime
nextofmoment position
the PIR sensor is the
(s3). Duringpointthewith
firstthe
value of the eight candidate estimated points are calculated. The next moment position is the point
half timeh(n)
maximum value.
of T3, “StartT3Node”
denotes the triggering
is still the begining time of the
of the PIR“Current
route, sensor (s3). Node”During
is thethe first
last half time
estimated
with the maximum h(n) value. T3 denotes the triggering time of the PIR sensor (s3). During the first
of T3,
point “Start
of theNode”
route, isandstill the begining
“Goal Node” is of stillthetheroute,
point“Current
with maximum Node”heat is the lastinestimated
score the PIR sensorpointareaof the
half time of T3, “Start Node” is still the begining of the route, “Current Node” is the last estimated
route,
(s3).andFor “Goal
the latterNode”
half is
timestillofthe
T3 point
and t1, with
“Start maximum heat score
Node” changes in the
to the point PIR sensor
with area (s3). For
the maximum heatthe
point of the route, and “Goal Node” is still the point with maximum heat score in the PIR sensor area
score
latter in time
the PIR sensor area (s3). “Goal Node” is still the point with the maximum heat score in in
the
(s3). For the latter half time of T3 and t1, “Start Node” changes to the point with the maximum heatthe
half of T3 and t1, “Start Node” changes to the point with the maximum heat score
PIRPIR sensor
sensor area
area (s4). In our lab, suppose that the moving speed of the subject is constant, the number
score in the PIR(s3). “Goal
sensor areaNode”
(s3). “Goalis still
Node”the point
is stillwith the maximum
the point heat scoreheat
with the maximum in the
scorePIRin sensor
the
of (s4).
area grid In in our
the accessibility map can denote asspeed
time. In Figure 11, t0~t10 and T3~T10 are the ofsubject
PIR sensor area (s4). In our lab, suppose that the moving speed of the subject is constant, the number in
lab, suppose that the moving of the subject is constant, the number grid
themoving grid number fordenote
different PIR sensing areas, t0~t10
the sum of them forare
thetheroute1 is 211. For the
ofaccessibility map can
grid in the accessibility map can as time.
denote In Figure
as time.11, In Figureand T3~T10
11, t0~t10 and T3~T10 subject
are the moving
subjectgrid
A-start
number tracking algorithm, the iteration time can be set as 211. For the route2, the iteration time is 143.
movingfor different
grid number PIRforsensing
different areas, the sumareas,
PIR sensing of themthefor
sum theofroute1
them for is 211. For theisA-start
the route1 211. For tracking
the
algorithm,
A-start trackingthe iteration timethe
algorithm, caniteration
be set as 211.
time For
can bethe route2,
set as 211. Forthethe
iteration
route2,time is 143. time is 143.
the iteration
triggering
triggering

triggering

triggering

s10 s10
s14 s14
s3 s3

s4 s4

triggering
triggering

triggering

triggering

t0 T3 t1 T4 T14 t10 T10

………
t0 T3 t1 T4 T14 t10 T10
………
……… The end
Start
Towards

Towards

Towards
Towards

Towards

Towards
Towards

s10 s10

………
s10 s10
s14 s14
s3 s3

s4 s4

The end
s3 s3

s4 s4

Start
Towards

Towards

Towards
Towards

Towards

Towards
Towards

Figure 11. State transition for the route1.

Figure Statetransition
Figure 11. State transitionfor
forthe
theroute1.
route1.
Sensors 2018, 18, 332 10 of 12
Sensors 2018,
Sensors 2018, 18,
18, xx FOR
FOR PEER
PEER REVIEW
REVIEW 10 of
10 of 12
12

In Figure 12a,
In Figure
Figure 12a, the
12a, the localization
the localization and
localization and tracking
tracking results
results of of aa person
person following
following thethe pre-defined
pre-defined route1
route1
are
are shown. line is the ground truth obtained from the OptiTrack
are shown. The black line is the ground truth obtained from the OptiTrack system. The
shown. The black OptiTrack system.
system. The red
The red line
red line is
line is the
is the
the
estimated
estimated trajectory.
estimated trajectory.
trajectory.The The
The distance
distance
distance error
error between
between
error the
the ground
between ground
the groundtruth andtruth and
the and
truth the
estimated estimated lines
lines is calculated
the estimated lines is is
calculated
for for each
each node,
calculated for each node,in
as shown
node, asFigure
as shown12b.
shown in Figure
in Figure 12b. The
The maximum
12b. The maximum
maximum
distance error distance error
is 0.747
distance is
m, is
error and0.747 m,
them,
0.747 and the
minimum
and the
minimum
distance distance
error is 0.021error
m. Onis 0.021 m.
average, On
the average,
mean of the
the mean
distance of the
error distance
is 0.227 m.
minimum distance error is 0.021 m. On average, the mean of the distance error is 0.227 m. Another error is
Another 0.227 m. Another
experiment was
experimentusing
conducted
experiment was route2,
was conducted
conductedshown using route2,
in Figure
using route2, shown
13.shown
The mean in Figure
in Figure 13. The
The error
of the distance
13. meanisof
mean of them.
0.188
the distance error is
The maximum
distance error is
0.188 m.
distance The
errormaximum
is 0.445 distance
m, and theerror is 0.445
minimum m, and
distance the minimum
error is 0.109 distance
m. error
Besides,
0.188 m. The maximum distance error is 0.445 m, and the minimum distance error is 0.109 m. Besides, is 0.109
Table 2 m. Besides,
shows the
Table 22 shows
comparison
Table shows the comparison
between
the comparison between
our methodbetween and some our method
recent
our methodhuman andtracking
and some recent
some recent human
projects,
human tracking
in terms projects,
of average
tracking in
error,
projects, in
terms of
sensor of average
number per 2
error, sensor
m , and number
overall per m
accuracy.m2,, and
2 and overall
overall accuracy.
accuracy.
terms average error, sensor number per
1
1
0.9
True route1 0.9
True route1
Tracking path1 0.8
Tracking path1 0.8 Emax
Emax
0.7
0.7
0.6
0.6

Error(m)
Error(m)
0.5
0.5
0.4
0.4
0.3
0.3 Eavg
0.2 Eavg
0.2
0.1
0.1
0
00 50 100 150 200
0 50 100
Time 150 200
Time

(a)
(a) (b)
(b)
Figure 12.
Figure 12. The
The result
Theresult of
resultof position
ofposition tracking
positiontracking with
trackingwith
withTrue
True route1.
True
route1. (a)(a)
route1.
(a) The
The path
The estimated
path
path with
estimated
estimated TrueTrue
with
with True route1;
route1; (b)
route1;
(b)
The
(b) error
TheThe curve
error
error for
curve
curve True
for for
True route1.
True route1.
route1.

1
1
True route2
True route2
Tracking path2
Tracking path2 0.8
0.8

0.6
Error(m)

0.6
Error(m)

Emax
Emax
0.4
0.4

0.2 Eavg
0.2 Eavg

0
00 20 40 60 80 100 120 140
0 20 40 60 Time80 100 120 140
Time
(a)
(a) (b)
(b)
Figure 13.
Figure 13. The result
result of position tracking
tracking with True
True route2.
route2. (a)
(a) The
The path
path estimated
estimated with
with True
True route2;
route2; (b)
(b)
Figure 13. The
The resultofofposition
position trackingwith
with True route2. (a) The path estimated with True route2;
The error
TheThe
error curve
curve for True
for for
True route2.
route2.
(b) error curve True route2.
Table 2.
Table 2. Recent
Recent project
project of
of human
human tracking.
tracking.
Table 2. Recent project of human tracking.
Method
Method Average Error
Average Error (m)
(m) Num_Uint
Num_Uint (/m
(/m22)) Text
Text Bed
Bed
Method [19] 0.85
Average0.85
Error (m) 0.017
Num_Uint 2 96%
Text Bed
[19] 0.017 (/m ) 96%
[20]
[20] 0.20
0.20 0.750
0.750 96.21%
96.21%
[19] 0.85 0.017 96%
[21]
[20] [21] 0.18
0.18
0.20 0.673
0.673
0.750 86.21%
86.21%
96.21%
Our
Our method
[21] method 0.21
0.18
0.21 0.116
0.673
0.116 87.36%
86.21%
87.36%
Our method 0.21 0.116 87.36%
4. Conclusions
4. Conclusions
4. Conclusions
In this
this work,
work, wewe studied
studied the
the problem
problem of of human
human localization
localization inin an
an indoor
indoor environment.
environment. We We have
have
In
proposed
proposed a method
In thisa work,
method that
wethat combines
studied a grid-based
the problem
combines accessible
of human
a grid-based map
localization
accessible and an A-star
in ananindoor
map and algorithm
A-starenvironment. for
algorithm forWe human
have
human
tracking.
proposed First,
a the
method grid-based
that accessibility
combines a map,
grid-based which reflected
accessible map human
and an visiting
A-star preferences
algorithm
tracking. First, the grid-based accessibility map, which reflected human visiting preferences and the for and the
human
physical layout of the area is built. Then PIR sensors which are deployed according to the
physical layout of the area is built. Then PIR sensors which are deployed according to the grid-based grid-based
accessibility map
accessibility map provide
provide anan external
external rough
rough position
position of
of the
the human.
human. Finally,
Finally, the
the A-star
A-star algorithm
algorithm fuses
fuses
Sensors 2018, 18, 332 11 of 12

tracking. First, the grid-based accessibility map, which reflected human visiting preferences and the
physical layout of the area is built. Then PIR sensors which are deployed according to the grid-based
accessibility map provide an external rough position of the human. Finally, the A-star algorithm fuses
the PIR sensor data and grid-based accessible map information to estimate the trajectory and reduce
the errors. In our research, PIR sensor data are collected and sent through a Zigbee communication
unit to a PC. The experiment results demonstrate the performance of our method. Our method can
be used for robots who accompany the elderly who are living alone, help the robots understand the
environment better and supply better service by localization tracking.
In the future, we will further improve our algorithm by considering more factors. For example,
we need to consider the speed and direction of human movement in smart homes. More practical
PIR sensors should also be considered, especially with regard to the response delay. We will also
investigate tracking multiple residents living together in a smart home.

Acknowledgments: The authors thank the National Natural Science Foundation of China (Grant Nos. 51607029)
for supporting this research and the Fundamental Research Funds for the Central Universities of China under
Grant no. N150404003 and N160404009.
Author Contributions: Dan Yang and Bin Xu conceived and designed the experiments; Dan Yang wrote the
paper; Bin Xu and Kaiyou Rao performed the experiments and analyzed the data; Weihua Sheng contributed
materials and experimental tools.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Wikipedia. Elderly Care. Available online: https://en.wikipedia.org/wiki/Elderly_care (accessed on
12 November 2017).
2. Wickramasinghe, A.; Torres, R.L.S.; Ranasinghe, D.C. Recognition of falls using dense sensing in an ambient
assisted living environment. Pervasive Mob. Comput. 2017, 34, 14–24. [CrossRef]
3. Dardari, D.; Closas, P.; Djurić, P.M. Indoor Tracking: Theory, Methods, and Technologies. IEEE Trans.
Veh. Technol. 2015, 64, 1263–1278. [CrossRef]
4. Want, R.; Hopper, A.; Falcao, V.; Gibbons, J. The active badge location system. ACM Trans. Inf. Syst. (TOIS)
1992, 10, 91–102. [CrossRef]
5. Priyantha, N.B.; Chakraborty, A.; Balakrishnan, H. The cricket location-support system. In Proceedings
of the 6th Annual International Conference on Mobile Computing and Networking, Boston, MA, USA,
6–11 August 2000; pp. 32–43.
6. Jiang, Z.; Xi, W.; Li, X.; Tang, S.; Zhao, J.; Han, J.; Zhao, K.; Wang, Z.; Xiao, B. Communicating is crowdsourcing:
Wi-Fi indoor localization with CSI-based speed estimation. J. Comput. Sci. Technol. 2014, 29, 589–604. [CrossRef]
7. Baldini, A.; Ciabattoni, L.; Felicetti, R.; Ferracuti, F.; Longhi, S.; Monteriù, A.; Freddi, A. Room occupancy
detection: Combining RSS analysis and fuzzy logic. In Proceedings of the 6th IEEE International Conference
on Consumer Electronics-Berlin (ICCE-Berlin), Berlin, Germany, 5–7 September 2016; pp. 69–72.
8. Yu, C.-R.; Wu, C.-L.; Lu, C.-H.; Fu, L.-C. Human localization via multi-cameras and floor sensors in smart
home. In Proceedings of the IEEE International Conference on Systems, Man and Cybernetics (SMC’06),
Taipei, Taiwan, 8–11 October 2006; Volume 5, pp. 3822–3827.
9. Bai, Y.; Lu, X. Research on UWB indoor positioning based on TDOA technique. In Proceedings of the
9th International Conference on Electronic Measurement and Instruments (ICEMI’09), Beijing, China,
16–19 August 2009; pp. 1167–1170.
10. Gozick, B.; Subbu, K.P.; Dantu, R.; Maeshiro, T. Magnetic maps for indoor navigation. IEEE Trans. Instrum. Meas.
2011, 60, 3883–3891. [CrossRef]
11. Yun, X.; Bachmann, E.R.; Moore, H.; Calusdian, J. Self-contained position tracking of human movement
using small inertial/magnetic sensor modules. In Proceedings of the 2007 IEEE International Conference on
Robotics and Automation, Roma, Italy, 10–14 April 2007.
12. Luo, R.C.; Chen, O. Wireless and pyroelectric sensory fusion system for indoor human/robot localization
and monitoring. IEEE/ASME Trans. Mechatron. 2013, 18, 845–853. [CrossRef]
Sensors 2018, 18, 332 12 of 12

13. Yang, B.; Luo, J.; Liu, Q. A novel low-cost and small-size human tracking system with pyroelectric infrared
sensor mesh network. Infrared Phys. Technol. 2014, 63, 147–156. [CrossRef]
14. Al-Naimi, I.; Wong, C.B.; Moore, P.; Chen, X. Advanced approach for indoor identification and tracking
using smart floor and pyroelectric infrared sensors. In Proceedings of the 5th International Conference on
Information and Communication Systems (ICICS), Irbid, Jordan, 1–3 April 2014; pp. 1–6.
15. Wang, H.; Zhou, J.; Zheng, G.; Liang, Y. HAS: Hierarchical A-Star algorithm for big map navigation in special
areas. In Proceedings of the ICDH 2014 International Conference on Digital Home, Guangzhou, China,
28–30 November 2014.
16. Song, B.; Choi, H.; Lee, H.S. Surveillance Tracking System Using Passive Infrared Motion Sensors in Wireless
Sensor Network. In Proceedings of the International Conference on Information Networking, Busan, Korea,
23–25 January 2008; pp. 1–5.
17. Xiaomu, L.; Tong, L.; Baihua, S.; Qinqun, C.; Liwen, G.; Xiaoyan, L. Human Indoor Localization Based on
Ceiling Mounted PIR Sensor Nodes. In Proceedings of the 13th IEEE Annual Consumer Communications &
Networking Conference(CCNC), Las Vegas, NV, USA, 9–12 January 2016; pp. 868–874.
18. Yang, D.; Sheng, W.; Zeng, R. Indoor human localization using PIR sensors and accessibility map. In Proceedings
of the 5th Annual IEEE International Conference on Cyber Technology in Automation, Control and Intelligent
Systems (CYBER), Shenyang, China, 8–12 June 2015; pp. 577–581.
19. The VN-100 IMU/AHRS. Available online: http://www.vectornav.com/products/vn100-rugged (accessed on
12 November 2017).
20. Kim, H.H.; Ha, K.N.; Lee, S.; Lee, K.C. Resident location recognition algorithm using a Bayesian classifier in
the PIR sensor-based indoor location-aware system. IEEE Trans. Syst. Man Cybern. Part C 2009, 39, 240–245.
21. Shen, B.; Wang, G. Object localization with wireless binary pyroelectric infrared sensors. In Proceedings of
the 2013 Chinese Intelligent Automation Conference, Yangzhou, China, 23–25 August 2013.

© 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).