sensors

Article
Surface Correlation-Based Fingerprinting Method
Using LTE Signal for Localization in Urban Canyon
Jung Ho Lee 1 , Beomju Shin 1 , Donghyun Shin 1 , Jinwoo Park 2 , Yong Sang Ryu 1 , Deok Ha Woo 1
and Taikjin Lee 1, *
1 Sensor System Research Center, Korea Institute of Science and Technology, 5, Hwarang-ro 14-gil,
Seongbuk-gu, Seoul 02972, Korea
2 Department of Electrical and Computer Engineering, Korea University, Anam-dong, Seungbuk-gu,
Seoul 136-713, Korea
* Correspondence: taikjin@kist.re.kr; Tel.: +82-2-958-5717




Received: 17 June 2019; Accepted: 28 July 2019; Published: 29 July 2019


Abstract: The Global Satellite Navigation System (GNSS) used in various location-based services is
accurate and stable in outdoor environments. However, it cannot be utilized in an indoor environment
because of low signal availability and degradation of accuracy due to the multipath distortion of
satellite signals in urban areas. On the contrary, LTE signals are available almost everywhere in
urban areas and are quite stable without much variation throughout the year. This is because of
the fixed location of base stations and the well-maintained policy of mobile communication service
providers. Its varied stability and reliability make LTE signals a more viable method for localization.
However, there are some complexities in utilizing LTE signals including signal interference distortion
phenomena during propagation multipath fading, and various types of noise. In this paper, we
propose a surface correlation-based fingerprinting method to utilize LTE signals for localization in
urban areas. The surface correlation converts timely measured signal strength into spatial pattern
using the walking distance from a Pedestrian Dead-Reckoning (PDR). The surface correlation is
carried out by comparing the spatial signal strength pattern of a pedestrian‘s movement trajectory
with a fingerprinting database to estimate the location. A reference trajectory of the moving pedestrian
is chosen to have a greater correlation among the multiple trajectory candidates generated from a
link-based fingerprinting database. By comparing spatial signal strength patterns, the proposed
method can improve robustness in localization overcoming the accuracy degradation problem due to
RF multipath and noise that are dominant in the conventional RSS measurement-based LTE localization
scheme. The test results in urban areas demonstrate that the proposed surface correlation-based
fingerprinting method has improved performance compared to the other conventional methods, thus
proving to be a useful complementary method to the GNSS in urban areas.

Keywords: localization; fingerprinting; LTE; correlation; pedestrian; PDR

1. Introduction
Recently, the rapid development of wireless communication technology has increased the
dependency of human life on location-based services. Reliable localization technology is widely
applied to various fields such as gaming, social networking service, industrial safety, as well as current
vehicle navigation systems. The most commonly used technology for localization is the Global Satellite
Navigation System (GNSS). It is obvious that GNSS is the most efficient and accurate in the outdoor
environment. However, in complex urban areas or indoor environments, the accuracy of the GNSS is
acutely degraded due to signal unavailability and the multipath fading of the satellite signals [1].

Sensors 2019, 19, 3325; doi:10.3390/s19153325 www.mdpi.com/journal/sensors

Sensors 2019, 19, 3325 2 of 19

In order to mitigate the limitation of GNSS usage in urban areas, several wireless communication
technologies for localization have been developed. In recent years, the Long Term Evolution (LTE)
that has become widely used in the metropolitan areas is attracting attention as a means of localizing
pedestrians or vehicles in these challenging areas [2–4]. The assisted-GNSS (A-GNSS) scheme
uses assistance data provided by LTE to improve the initial operation of Global Positioning System
(GPS)-based devices [5,6]. However, this scheme does not use LTE signals for localization. The enhanced
Cell-ID (e-CID) technology was proposed to improve localization accuracy by adding Round Trip
Time-based distance information to the existing Cell-ID method. However, its performance is limited
depending on the size of the cell [7,8]. The Observed Time Difference of Arrival (OTDoA) technology
defined in the 3GPP-LTE Release 9 standard is based on the Time Difference of Arrival (TDoA) method
using the positioning reference signal of LTE. This technology estimates a location based on the
hyperbolic trilateration of the TDoA measurement, Reference Signal Time Difference (RSTD) [9,10].
Since the OTDoA technology requires at least three Base Stations (BS) for trilateration, its usage is
limited in areas with low BS density. Despite the above-mentioned LTE-based attempts to improve
performance, such the technologies are still hard-pressed to satisfy the performance standards, such as
accuracy, stability, and availability, required by various LBS services.
In Non-Line-of-Sight (NLoS) environments, such as urban areas or indoors, a fingerprinting
technology can be more effective than timing-based localization technology. The LTE Positioning
Protocol requires that Received Signal Strength (RSS) information be provided as well as CID, Timing
Advanced (TA), angle of arrival and channel status information [11,12]. Gundlegard et al. [13]
proposed the use of Observed Time Differences (OTD) with RSS to improve the performance of a
fingerprinting-based localization technology. Zhu et al. [14] developed a maximum likelihood-based
fingerprinting method using RSS, RSTD, and TA of LTE system. It demonstrated that the proposed
fingerprinting method is generally superior to the e-CID and OTDOA based localization method.
Turkka et al. [15] developed a grid-based fingerprinting method using the RSS information of LTE and
WLAN. The performance of LTE-based fingerprinting method was improved with the RSS information
of WLAN.
However, it has been reported that the fingerprinting-based localization technologies based on
direct comparison of the measured RSS with fingerprinting database suffer from the lack of accuracy,
stability, and robustness [16–18]. Received RF signals experience many distortion phenomena during
propagation and vary consistently due to other RF signal‘ interference, multipath fading, and noise.
The RF signal distortion and variation becomes more prominent in the areas where structures that
prohibit direct propagation are densely located and diverse wireless applications are actively used.
This is typical in the LTE signal generated in urban areas. In conventional fingerprinting technology
that utilizes instantaneous RSS data, these distortion phenomena tend to increase ambiguity. Therefore,
compensation methods for the fingerprinting-based localization technology to reduce such the RF
signal‘s distortion and instability problems must be devised.
To solve this problem, there have been researches integrating the Pedestrian Dead-Reckoning
(PDR) with RF-based localization technology. The PDR can accurately estimate a location using a
foot-mounted Inertial Measurement Unit (IMU) for a short period of time. By integrating a map
matching algorithm, it is possible to operate accurately for a longer period of time [19,20]. However, in
the case of a smartphone, it is difficult to expect a high-performance of the PDR due to the various
activities of a pedestrian. Masiero et al. [21] developed a particle filter based integration method of
inertial measurements, RSS of Wi-Fi, and map information using a smartphone. It is highly reliable
and accurate to estimate the location of a pedestrian in an indoor environment. But it is still difficult to
guarantee stable performance in the outdoor where the density of LTE BS and signal strength are very
low due to the use of only instantaneous signal strength data.
In this paper, we propose a new fingerprinting method based on the concept of Surface Correlation
(SC) in the LTE environment of urban areas. The SC estimates a location by comparing the spatial RSS
pattern data provided by the PDR algorithm with a fingerprinting database, rather than relying just
Sensors 2019, 19, 3325 3 of 19

on instantaneous RSS measurements as in the conventional fingerprinting method. This spatial RSS
Sensors 2019, 19, x FOR PEER REVIEW
pattern not only improves the accuracy and stability of localization performance by reducing 3ambiguity of 18

but also increases the availability of the LTE-based localization technology. The contributions of this
than relying just on instantaneous RSS measurements as in the conventional fingerprinting method.
studyThis
are spatial
following:
RSS pattern not only improves the accuracy and stability of localization performance by
reducing ambiguity but also increases the availability of the LTE-based localization technology. The
• We develop an SC-based fingerprinting technology that computes a location through comparison
contributions of this study are following:
of spatial RSS patterns, rather than instantaneous RSS data as applied in the conventional
• We developmethod,
fingerprinting an SC-based fingerprinting
between technology
the database that computes a location through comparison
and measurements.
of spatial RSS patterns, rather than instantaneous RSS data as applied in the conventional
• For comparison between the measurements and database, we develop methods for generating
fingerprinting method, between the database and measurements.
spatial RSS patterns from them according to the movement of a pedestrian.
• For comparison between the measurements and database, we develop methods for generating
• We obtain accurate
spatial RSS and
patterns stable
from them location
accordinginformation by of
to the movement using few LTE base stations even in
a pedestrian.
error-prone urban areas.
• We obtain accurate and stable location information by using few LTE base stations even in error-
prone urban areas.
The paper is organized as follows: Section 2 provides an overview of the state-of-the-art surface
Theapproach
correlation paper is organized as follows:
to the proposed Section 2 provides
fingerprinting an overview
method. of the
Section state-of-the-art
3 describes surfaceof its
the details
correlation approach to the proposed fingerprinting method. Section 3 describes the details of its
application on the fingerprinting method. The performance of the proposed fingerprinting method is
application on the fingerprinting method. The performance of the proposed fingerprinting method
discussed in Section 4. Finally, Section 5 presents conclusions drawn in this work.
is discussed in Section 4. Finally, Section 5 presents conclusions drawn in this work.
2. Algorithm Overview
2. Algorithm Overview
Generally, a fingerprinting
Generally, a fingerprinting method
methodinvolves
involves an offlinephase
an offline phaseandandanan online
online phase.
phase. Inoffline
In the the offline
phase, a survey
phase, is performed
a survey is performed to to
construct
constructaafingerprinting database
fingerprinting database asas displayed
displayed in Figure
in Figure 1. This
1. This is is
achieved by storing
achieved the RSS
by storing theof signals
RSS received
of signals from surrounding
received signalsignal
from surrounding sources at a specified
sources Reference
at a specified
PointReference
(RP). In the online
Point (RP).phase,
In thea online
terminal device
phase, measuresdevice
a terminal the RSS received
measures thefrom
RSS surrounding
received fromsignal
surrounding signal sources and estimates location by searching the RP with similar
sources and estimates location by searching the RP with similar fingerprint. As shown in Figure fingerprint. As 1,
shown in Figure 1, the RSS measurement in the conventional fingerprinting
the RSS measurement in the conventional fingerprinting method is through instantaneous RSS data, method is through
whichinstantaneous
can increaseRSS data, which can increase ambiguity due to noise and distortion by the surrounding
ambiguity due to noise and distortion by the surrounding environment. While the
environment. While the effects of noise and distortion are low in areas with high RSS, it is impossible
effects of noise and distortion are low in areas with high RSS, it is impossible for stable and accurate
for stable and accurate localization due to the increased ambiguity in areas with very low RSS such
localization due to the increased ambiguity in areas with very low RSS such as deep urban areas.
as deep urban areas.

Figure
Figure 1. Overviewofofconventional
1. Overview conventional and
and proposed
proposedfingerprinting methods.
fingerprinting methods.
Sensors 2019, 19, 3325 4 of 19

Sensors 2019, 19, x FOR PEER REVIEW 4 of 18

The SC method compares RSS patterns accumulated during movement with the fingerprinting
databaseThetoSC method
estimate thecompares RSSSC
location. The patterns
methodaccumulated
can provide high during movement(availability)
performance with the fingerprinting
regardless
database to estimate the location. The SC method can provide high
of the high and low areas of the LTE signal strength because it compares the spatial RSS variation performance (availability)
regardless
pattern, not of the high anddata,
instantaneous low areas
with of
thethe LTE signal
reference strength
pattern frombecause it compares
the database. thepaper,
In this spatialwe RSS
variation pattern, not instantaneous data, with the reference pattern from
define the RSS pattern accumulated during movement, i.e. the spatial RSS pattern, as the Surface. the database. In this paper,
We we define
have the RSS
already shown pattern
that accumulated
the performance during movement,
of the conventional i.e. the spatial RSS method
fingerprinting pattern, improves
as the Surface.
by
comparing a one-dimensional fingerprinting database with spatial RSS patterns [22,23]. Animproves
We have already shown that the performance of the conventional fingerprinting method urban
by comparing
area a one-dimensional
is a two-dimensional space withfingerprinting
multiple roads database with spatial
combined, RSS
not just patternsone-dimensional
a simple [22,23]. An urban
area is
space. a two-dimensional
Therefore, the SC method space with multiple
requires roads combined,
the generation of reference not
RSSjust a simple
patterns thatone-dimensional
match with a
pedestrian‘s route from the fingerprinting database for comparison between the measured match
space. Therefore, the SC method requires the generation of reference RSS patterns that spatialwith
RSS a
pedestrian`s
pattern (Surface)route
and thefrom the fingerprinting
database. To do this, itdatabase
is required fortocomparison
construct a between
link-centeredthe measured spatial
database rather
RSS pattern (Surface) and the database. To do this, it is required to
than an RP-oriented database as seen in the conventional fingerprinting method. The SC method construct a link-centered database
rather of
consists than
fouran RP-oriented
processes database as(1)seen
for localization: in the
domain conventional
conversion of RSSfingerprinting
measurements;method. The SC
(2) generation
ofmethod
reference consists
trajectoryof four
fromprocesses for localization:
a fingerprinting database;(1) (3)domain conversion
correlation; of RSS measurements; (2)
(4) localization.
generation of reference trajectory from a fingerprinting database; (3) correlation; (4) localization.
3. Surface Correlation-based Fingerprinting Method
3. Surface Correlation-based Fingerprinting Method
3.1. Fingerprinting Database
3.1. Fingerprinting Database
Generally, the fingerprinting database consists of the coordinates of RPs and RSS data from
Generally,
surrounding BS at the
the fingerprinting
RP location. Asdatabase
mentioned consists
above, of thetheSC coordinates
method comparesof RPsspatial
and RSSRSS data from
patterns
surrounding
generated during BSpedestrian
at the RPmovement
location. Aswithmentioned
the database; above,
thus,theit isSC methodtocompares
necessary generate aspatial
referenceRSS
patternsmatching
trajectory generatedwith during
the pedestrian
pedestrian’smovement
route fromwith the database;
the database. thus, ititisutilizes
Therefore, necessary to generate
a database of
a reference
‘link-node’ trajectory
structures to matching
generate thewith the pedestrian’s
reference trajectories route from the database.
dynamically. The link isTherefore,
generated itbased
utilizes
on a
database and
crossroads of ‘link-node’
consists of structures
a set of RPstoasgenerate
shown inthe reference
Figure 2. Thetrajectories
crosspointdynamically.
connecting each Thelink
linkis is
generated
defined based The
as a node. on crossroads
fingerprintinganddatabase
consists canof abesetexpressed
of RPs asasshown follows:in Figure 2. The crosspoint
connecting each link is defined as a node. The fingerprinting database can
be expressed as follows:
RPlk = [(x, y), (PCI1 : RSS1 ), · · · , (PCIn : RSSn ) ]

(1)
𝑅𝑃 = [ 𝑥, 𝑦 , 𝑃𝐶𝐼 : 𝑅𝑆𝑆 , ⋯ , 𝑃𝐶𝐼 : 𝑅𝑆𝑆 ] (1)
where, l is the kth RP on the lth link, (x,y) is the coordinate of RP, PCI is Physical Layer Cell ID (PCI)
RP𝑅𝑃
where, k is the kth RP on the lth link, (x,y) is the coordinate of RP,n 𝑃𝐶𝐼 is Physical Layer Cell ID
nth BS,
of(PCI) of nth RSS
and is the
BS, nand received
RSS n is thesignal strength
received signalfrom PCIn , from
strength respectively. The reference
PCIn, respectively. Thetrajectory,
reference
which matches the route of athe
pedestrian, l
of RPk .
trajectory, which matches route of aispedestrian,
generated isthrough a combination
generated through a combination of 𝑅𝑃 .

Figure The
2. 2.
Figure link-node
The structure
link-node forfor
structure fingerprinting database.
fingerprinting database.

3.2. Domain Conversion

RSS measurement is distributed in the time domain since it is measured at a certain time interval
on a terminal device. Thus, it can be expressed as:
Sensors 2019, 19, 3325 5 of 19

3.2. Domain Conversion

RSS measurement is distributed in the time domain since it is measured at a certain time interval
on a terminal device. Thus, it can be expressed as:

RSS1t · · · RSSnt 
 
Sensors 2019, 19, x FOR PEER REVIEW
 t1 1 1 5 of 18
.. .. ..
 
Meas = 
 .. 
(2)
 . . . . 

1 n
tm 𝑡 𝑅𝑆𝑆tm ·⋯
RSS · · 𝑅𝑆𝑆RSStm
 

Meas = ⋮ ⋮ ⋱ ⋮ (2)
where, RSSntm is the received signal strength from 𝑡 nth𝑅𝑆𝑆BS at t⋯ m . The
𝑅𝑆𝑆conventional fingerprinting database
consists of the RP coordinate and the RSS values at that point as Equation (1). The conventional
where, 𝑅𝑆𝑆 is the received signal the strength from RSS nth BSdataat(RSS𝑡 . 1:nThe conventional fingerprinting
fingerprinting methods compare instantaneous tm ) at time tm with the RSS vector
database
on each RP, so it is possible to compute a location. Given that theasRSS
consists of the RP coordinate and the RSS values at that point Equation (1).measured
pattern The conventional
on time
:
fingerprinting methods compare the instantaneous RSS data (𝑅𝑆𝑆
domain cannot be compared with the RSS pattern on the database distributed in the spatial ) at time 𝑡 with the domain,
RSS vector
the
on each RP, so it is possible to compute a location. Given that the RSS pattern
measured RSS data should be converted into a spatial domain. To do this, the proposed fingerprinting measured on time domain
cannot
methodbeutilizes
compared the with the RSS
pedestrian pattern
step lengthonestimation
the database distributed
algorithm already in the spatial in
proposed domain, the
an earlier
measured RSS
study [24], data should
where be converted
step length is modeled into aasspatial
a lineardomain. To do this,
combination of the
the proposed fingerprinting
acceleration amplitude,
method
walking frequency, and sway angle at every step. The SC method estimates the walking earlier
utilizes the pedestrian step length estimation algorithm already proposed in an study
distance d by
[24], where step length is modeled as a linear combination of the
accumulating the step length and accumulates the measured RSS according to the walking distance as acceleration amplitude, walking
frequency, and 3.
given in Figure sway
Thus,angle at every
the Surface canstep. The SC as:
be expressed method estimates the walking distance d by
accumulating the step length and accumulates the measured RSS according to the walking distance as
given in Figure 3. Thus, the Surface  Scan RSS1d · · · RSS1d 
 
be expressed as:
 
1   d1
  1 1 
 .   .
𝑅𝑆𝑆... ⋯. . . 𝑅𝑆𝑆 ...
 
S =  .. 𝑆=  ..𝑑

 (3)
 𝑆 =  ⋮ = ⋮ ⋮ n ⋱· · · RSS ⋮ n
   
Sm dm RSS

 (3)
𝑆 𝑑 𝑅𝑆𝑆 dm ⋯ 𝑅𝑆𝑆 dm
where,mmisisthe
where, thenumber steps,SSisisthe
numberofofsteps, Surfaceatatwalking
theSurface distance𝑑dm and
walkingdistance RSSnd is
and 𝑅𝑆𝑆 is the
the received
received
m
signal strength from nth BS at walking distance d . The proposed fingerprinting
signal strength from nth BS at walking distance 𝑑 . The proposed fingerprinting method estimates
m method estimates
locationby
location by correlating
correlatingthe
the RSS
RSS pattern
pattern onon the
the reference
reference trajectory
trajectory with
with the
the measured Surface. The
measured Surface. The
referenceRSS
reference RSSpattern
patterngenerated
generatedfrom
fromthethe fingerprinting
fingerprinting database
database also
also forms
forms a Surface
a Surface asEquation
as in in Equation
(3)
(3) by calculating the Euclidean distance between the adjacent
by calculating the Euclidean distance between the adjacent RPs. RPs.

(a) (b)
Figure
Figure 3.
3. Domain
Domainconversion:
conversion:(a)
(a)Received
ReceivedSignal
SignalStrength
Strength(RSS)
(RSS)pattern
patternon
ontime
timedomain;
domain;(b)
(b)RSS
RSS
pattern
patternon
onthe
thespatial
spatialdomain.
domain.

3.3. Surface Correlation

The Surface Correlation is the process of determining a location by calculating the similarity
between the measured and reference RSS patterns. The similarity is expressed as a Surface Similarity
Distance (SSD) and calculated as follows:

⎧ ∑ 𝑆, −𝑆 ,
⎫
𝜌 = (4)
⎨ 𝑚 ⎬
⎩ ⎭
 Sensors 2019, 19, 3325 6 of 19

3.3. Surface Correlation

The Surface Correlation is the process of determining a location by calculating the similarity
between the measured and reference RSS patterns. The similarity is expressed as a Surface Similarity
Distance (SSD) and calculated as follows:
r 2 
 Pm  Meas DB

S − S
 
L−w w 
 
k = 1
 
XX j,k i+ j−1,k 
ρi =
 
(4)
 
m

 

 
i=1 j=1 





 

Sensors
where2019,
i is 19,
thex FOR PEER
epoch of REVIEW
RP on a generated reference trajectory and m is the number of BS. ρi6isofthe 18

for ith
SSD 2019,
Sensors 19, xcalculation, SMeas
FOR PEER REVIEW is the measured Surface, and S
DB is the RSS pattern on the reference 6 of 18
Surface during
trajectory. L isthethetime
number of walking
of RPs distance D, that is,
in the reference the number
trajectory and of w estimated
is the number steps.
ofTo calculate
samples on the
the
similarity
Surface
Surfaceduringdistance,
during the time
the the
time number
of of
walking
walking of RPs
distance selected
distance D, that in the
is, the
D, that reference
number
is, the number trajectory
of estimated
of estimatedshould
steps. be equal
To calculate
steps. to the
the
To calculate
number
similarity of
the similaritysamples
distance, in the
the
distance, themeasured
numbernumberof RPs ofSurface during
selected
RPs in the
selected the
in period
reference
the oftrajectory
reference walking
trajectory distance
should
should D,be
be which
equal
equaltois not
tothe
the
usually
number possible.
of samples This
in is
the because
measured the step
Surface length
during of athepedestrian
period of and the
walking
number of samples in the measured Surface during the period of walking distance D, which is not interval
distancebetween
D, whichthe RPs
is notis
not equal.
usually
usually To solve
possible.
possible. This
Thisthis problem,
isisbecause
because thewe
the stepgenerate
step length ofa virtual Surface
aa pedestrian
pedestrian equal
and
and to the number
theinterval
the interval between
between ofthe
RPs
the within
RPs
RPs is
is not
distance
not equal.
equal. D
To To through
solvesolve
thisthislinear spacing
problem,
problem, as shown
we generate
we generate in Figure
a virtual
a virtual Surface 4. The
Surface
equalequalvirtual
to the to Surface
the number
number is utilized
of RPs of for
RPs distance
within withinthe
calculation
distance
D through of
D throughSSDspacing
linear aslinear
shown as in Figure
spacing
shown asin5a. The ith
shown
Figure in RP
4. Figure
The coordinate
4. The
virtual with the
virtual
Surface is minimum
Surface
utilized 𝜌 calculation
is the
for is considered
utilized for theof
as
SSDtheascurrent
calculation of SSD
shown location
in as shown
Figure as5a.
shown
inThe in
ithFigure
Figure RP 5b.ith RPwith
5a.coordinate
The coordinate with theρminimum
the minimum is 𝜌
considered is
asconsidered
the current
i
aslocation
the current location
as shown as shown
in Figure 5b. in Figure 5b.

Figure 4. Generation of virtual Surface for the correlation process.

Figure 4. Generation of virtual Surface for the correlation process.
Figure 4. Generation of virtual Surface for the correlation process.

(a) (b)
Figure
Figure 5. (a)process:
5. Correlation
Correlation process: (a)
(a)Correlation
Correlationbetween
betweenmeasured
measuredSurface (b)reference
Surfaceand
and referenceRSS
RSSpattern;
pattern;
(b) Surface
(b) Surface similarity distance.
similarityprocess:
distance.
Figure 5. Correlation (a) Correlation between measured Surface and reference RSS pattern;
(b) Surface similarity distance.
3.4. Generation of Reference Trajectory
3.4. Generation of Reference Trajectory
For accurate
3.4. Generation
For localization
of Reference
accurate using
Trajectory
localization thethe
using proposed surface
proposed correlation
surface method,
correlation it is essential
method, to generate
it is essential to
a reference
generate trajectory that matches the pedestrian’s route. This is because it is necessary to compare
For accurate localization using the proposed surface correlation method, it is essential toto
a reference trajectory that matches the pedestrian's route. This is because it is necessary
generate athe
compare measured
reference RSS pattern
trajectory with thethe
that matches reference RSS pattern
pedestrian's route. that
This matches
is because theitmovement
is necessary route
to
to make accurate localization possible. The reference trajectory is generated
compare the measured RSS pattern with the reference RSS pattern that matches the movement route by combining links
toadjacent to the link
make accurate where apossible.
localization pedestrian
Theisreference
currentlytrajectory
located; therefore,
is generated it isbyvery important
combining linksto
determine the link where the pedestrian is currently located. It is possible with
adjacent to the link where a pedestrian is currently located; therefore, it is very important to substantial accuracy
in an environment
determine with good
the link where GNSS conditions,
the pedestrian butlocated.
is currently quite difficult in a complex
It is possible urban area.
with substantial In this
accuracy
 Sensors
Sensors 19,19,
2019,
2019, 3325 PEER REVIEW
x FOR 7 of7 of
1819

generates reference trajectories using the selected links. A reference trajectory is fixed by correlation
the measured RSS pattern with the reference RSS pattern that matches the movement route to make
the measured RSS pattern and the RSS patterns on the generated reference trajectories. In other
accurate localization possible. The reference trajectory is generated by combining links adjacent to the
words, SSD is the criterion not only for determining the similarity between the RSS patterns, but also
link where a pedestrian is currently located; therefore, it is very important to determine the link where
for selecting the actual trajectory where the current pedestrian is located among the generated
the pedestrian is currently located. It is possible with substantial accuracy in an environment with
multiple reference trajectories. The detail of this process is given in Figure 6.
good GNSS conditions, but quite difficult in a complex urban area. In this study, we determine the link
Step 1: Generate
where reference
a pedestrian trajectories
is currently as follows:
located without GNSS. There are two steps to determine the required
• link: (1) Generation of reference trajectory
Find and select n candidate links including in measured
an initial phase; (2)the
PCI from Generation of reference
fingerprinting trajectory
database.
• after
Forthe initial
each phase.
of the selected candidate links, generate bi-directional 2n reference.
• If the estimated walking distance D is larger than the shortest reference trajectory L before
3.4.1. Generation of Reference Trajectory in Initial Phase
checking Minimum ρ < β, generate new reference trajectories by combining links connected with
Since
the therereference
current is no location information
trajectory. Only thein the initial
newly phase ofreference
generated localization, the current
trajectories link should
matching with
be determined
estimated turn using only
event PCI
(left, information
right, and
or straight) RSSthe
from measurement from a terminal(PDR)
Pedestrian Dead-Reckoning device.areAt first,
used
theforproposed fingerprinting
the correlation process. method selects candidate links using only PCI information, and then
generates reference trajectories using the selected links. A reference trajectory is fixed by correlation
Step 2: Calculate the SSD on each reference trajectory using Equation (4).
the measured RSS pattern and the RSS patterns on the generated reference trajectories. In other words,
SSD3:isWhen
Step the criterion not only
the pedestrian for determining
moves over a certainthe similarity
distance between
(α) and the SSDthevalue
RSS patterns, ρ
but also for
is stable (Minimum
< selecting
β), the reference
the actual trajectory
trajectory withwhere the current
the minimum SSDpedestrian is located
is fixed and among
the location the generated
is estimated multiple
on the fixed
reference
reference trajectories. The detail of this process is given in Figure 6.
trajectory.

Figure 6. Flow chart for the generation of reference trajectory at the initial phase.
Figure 6. Flow chart for the generation of reference trajectory at the initial phase.

Step 1: Generate reference trajectories as follows:

3.4.2. Generation of Reference Trajectory after Initial Phase
• After
Findthe
andinitial n candidate
selectreference links including
trajectory is fixed,measured PCI from
a new reference the fingerprinting
trajectory database. by
is again generated
• For each
combining of the selected
the current candidate
reference links,
trajectory andgenerate bi-directional
the adjacent 2n reference.
links. A reference trajectory is fixed for
• moving
the If the estimated
direction walking route of D
and the distance is larger
the than the
pedestrian by shortest reference
correlating trajectory
the newly L before
generated checking
reference
Minimum ρ < β, generate new reference trajectories by combining links connected with the
trajectories. Specifically, the walking direction is estimated by selecting a reference trajectory between current
reference trajectory. Only the newly generated reference trajectories matching
the initial bidirectional links, not in the absolute direction from PDR. Let us assume that the with estimated
pedestrian moves as shown in Figure 7a; measured Surface during the movement is shown in Figure
Sensors 2019, 19, 3325 8 of 19

turn event (left, right, or straight) from the Pedestrian Dead-Reckoning (PDR) are used for the
correlation process.

Step 2: Calculate the SSD on each reference trajectory using Equation (4).
Step 3: When the pedestrian moves over a certain distance (α) and the SSD value is stable
(Minimum ρ < β), the reference trajectory with the minimum SSD is fixed and the location is estimated
on the fixed reference trajectory.

3.4.2. Generation of Reference Trajectory after Initial Phase

After the initial reference trajectory is fixed, a new reference trajectory is again generated by
combining the current reference trajectory and the adjacent links. A reference trajectory is fixed for the
moving direction and the route of the pedestrian by correlating the newly generated reference trajectories.
Specifically, the walking direction is estimated by selecting a reference trajectory between the initial
Sensors 2019, 19, x FOR PEER REVIEW 8 of 18
bidirectional links, not in the absolute direction from PDR. Let us assume that the pedestrian moves
as shown in Figure 7a; measured Surface during the movement is shown in Figure 7b. The proposed
7b. The proposed fingerprinting method generates two reference trajectories by combining connected
fingerprinting method generates two reference trajectories by combining connected links with the
links with the current reference trajectory as shown in Figure 8a and b. Figure 8c and d show the
current reference trajectory as shown in Figure 8a and b. Figure 8c and d show the respective reference
respective reference RSS patterns on the generated reference trajectory.
RSS patterns on the generated reference trajectory.

(a) (b)
Figure 7. Movement
Figure scenario
7. Movement of aofpedestrian:
scenario (a) (a)
a pedestrian: Route; (b) (b)
Route; Measured RSSRSS
Measured during movement.
during movement.

In
Inthe
the initial phase,
phase,the
theproposed
proposedmethod
methodfixesfixes a reference
a reference trajectory
trajectory based
based on minimum
on the the minimum
SSD.
SSD.
AfterAfter the initial
the initial phase,phase,
it fixesitafixes a reference
reference trajectory
trajectory based onbased on the
the ratio of ratio
SSDs of
as SSDs
shownasinshown in
Equation
Equation
(5) below:(5) below:
ρ1
R= ,𝜌 i f ρ1 > ρ2 (5)
𝑅= ρ2 , 𝑖𝑓 𝜌 > 𝜌 (5)
𝜌
where ρ2 is the minimum SSD between the reference trajectories and ρ1 is the second smallest value of
where 𝜌 is the minimum SSD between the reference trajectories and 𝜌 is the second smallest value
the SSD. The values of SSDs change after the pedestrian has passed a crossroad as shown in Figure 9a.
of the SSD. The values of SSDs change after the pedestrian has passed a crossroad as shown in Figure
The reason why the ratio R has a value 1 before the crossroad is that the reference trajectory 1 and the
9(a). The reason why the ratio R has a value 1 before the crossroad is that the reference trajectory 1 and
reference trajectory 2 have a common current reference trajectory. As shown in Figure 9b, after the
the reference trajectory 2 have a common current reference trajectory. As shown in Figure 9b, after the
crossroad, R increases because of the difference between the measured RSS pattern and the reference
crossroad, R increases because of the difference between the measured RSS pattern and the reference
RSS patterns. Therefore, it is possible to fix a reference trajectory that matches the pedestrian’s walking
RSS patterns. Therefore, it is possible to fix a reference trajectory that matches the pedestrian’s walking
route using only a threshold for the ratio of SSDs. After fixing the reference trajectory, new reference
route using only a threshold for the ratio of SSDs. After fixing the reference trajectory, new reference
trajectories are generated by combining links with the fixed reference trajectory again.
trajectories are generated by combining links with the fixed reference trajectory again.
 the reference trajectory 2 have a common current reference trajectory. As shown in Figure 9b, after the
crossroad, R increases because of the difference between the measured RSS pattern and the reference
RSS patterns. Therefore, it is possible to fix a reference trajectory that matches the pedestrian’s walking
route using only a threshold for the ratio of SSDs. After fixing the reference trajectory, new reference
Sensors 2019, 19,are
trajectories 3325
generated by combining links with the fixed reference trajectory again. 9 of 19

Sensors 2019, 19, x FOR PEER REVIEW 9 of 18

Sensors 2019, 19, x FOR PEER REVIEW 9 of 18
(a) (b)

(c) (d)
(c) (d)
Figure 8. Generation of reference RSS pattern: (a) Regenerated reference trajectory 1; (b) Regenerated
Figure8.trajectory
Figure
reference 8.Generation
Generation ofof
2; (c) RSSreference
reference RSS
pattern RSS
of pattern:
pattern:
reference (a) Regenerated
(a)trajectory
Regenerated RSSreference
1; (d)reference trajectory
trajectory
pattern of 1; (b) 1;
reference (b) Regenerated
Regenerated
trajectory 2.
reference trajectory 2; (c) RSS pattern of reference trajectory 1; (d) RSS pattern of reference
reference trajectory 2; (c) RSS pattern of reference trajectory 1; (d) RSS pattern of reference trajectory trajectory
2. 2.

(a) (b)
(a) (b)
Figure
Figure9.9.Fix
Fixaareference
reference trajectory using aaratio
trajectory using ratiobetween
betweenvalues
valuesofof SSD:
SSD: (a)(a) SSDs
SSDs of reference
of reference trajectories;
trajectories;
Figure
(b)
(b)Ratio 9.between
Fix a reference
Ratiobetween SSDs. trajectory using a ratio between values of SSD: (a) SSDs of reference trajectories;
SSDs.
(b) Ratio between SSDs.
3.5. Localization
3.5. Localization
As mentioned before, the SC method selects the RP with the minimum SSD on the fixed reference
As mentioned
trajectory before,
as the current the SC method
location. Figure selects the RPthe
10b shows with the minimum
result SSD onaccording
of localization the fixed reference
to the
trajectorysimilarity
calculated as the current
distance location. Figure
in Figures 9 and10b10a.
shows the result
As shown of localization
in Figure 10b, it canaccording
be seen thatto the
a
localization error occurs at the crossroad when the RP with the minimum similarity distance value is a
calculated similarity distance in Figures 9 and 10a. As shown in Figure 10b, it can be seen that
localization
selected. error occurs
A crossroad is a at the crossroad
relatively when the
open space; RP withthe
therefore, theRSS
minimum similarity
patterns betweendistance value is
the reference
selected. A
trajectories arecrossroad is a relatively
very similar. open ambiguity
This increases space; therefore, the RSS patterns
when compared with thebetween
measured the Surface,
reference
resulting in a localization error. The proposed method gives a priority to the reference trajectorySurface,
trajectories are very similar. This increases ambiguity when compared with the measured that
resulting
matches in aevent
a turn localization
from the error.
PDRThe proposed
to reduce methodatgives
ambiguity a priority
a crossroad, to themeans
which reference
that trajectory that
the location
 Sensors 2019, 19, 3325 10 of 19

3.5. Localization
As mentioned before, the SC method selects the RP with the minimum SSD on the fixed reference
trajectory as the current location. Figure 10b shows the result of localization according to the calculated
similarity distance in Figures 9 and 10a. As shown in Figure 10b, it can be seen that a localization
error occurs at the crossroad when the RP with the minimum similarity distance value is selected.
A crossroad is a relatively open space; therefore, the RSS patterns between the reference trajectories
are very similar. This increases ambiguity when compared with the measured Surface, resulting in a
localization error. The proposed method gives a priority to the reference trajectory that matches a turn
event from the PDR to reduce ambiguity at a crossroad, which means that the location is estimated
from the reference trajectory that corresponds to a turn event from the PDR. However, various motions
of the pedestrian can cause a false turn event reading of the PDR. To prevent this, the proposed method
continuously calculates SSDs for all the generated reference trajectories until one of them is fixed. If the
fixed reference trajectory is different from the turn event matched trajectory, localization is processed
on the fixed reference trajectory. In this paper, we set a threshold to fix a reference trajectory as 1.16 to
Sensors 2019, 19, x FOR PEER REVIEW 10 of 18
reduce the error of the PDR turn event at a crossroad.

(a) (b)

Figure
Figure10.
10.The
Theproblem
problem of localizationusing
of localization usingthe
the minimum
minimum SSDSSD
at a at a cross
cross point:point: (a)ofSSD
(a) SSD eachof each
reference
reference trajectory
trajectory at a crossroad;
at a crossroad; (b) Localization
(b) Localization error anderror and compensation
compensation using Pedestrian
using Pedestrian Dead-
Dead-Reckoning
(PDR) turn
Reckoning event.
(PDR) turn event.

InIn addition
addition toto the
the left
left andand right
right turn
turn ofof a pedestrian,
a pedestrian, the
the event
event ofof turning
turning back
back onon a link
a link should
should
also be considered. The SC method generates a reference trajectory through
also be considered. The SC method generates a reference trajectory through a combination of links. a combination of links.
Therefore, when the pedestrian is moving as shown in Figure 11a, the SC
Therefore, when the pedestrian is moving as shown in Figure 11a, the SC method computes the method computes the location
using ausing
location reference trajectory
a reference comprising
trajectory a combination
comprising of L1, L2,
a combination of and candidate
L1, L2, L3 before
and candidate L3the turnthe
before back
event. It is possible to generate a reference trajectory by combining L1, L2, L2’ and
turn back event. It is possible to generate a reference trajectory by combining L1, L2, L2’ and L1’. But,L1’. But, in this case,
initthis
is impossible to perform
case, it is impossible to an accurate
perform comparison
an accurate comparison Surfaces because
between between the reference
Surfaces because trajectory
the reference
containscontains
trajectory RPs thatRPs the pedestrian did not actually
that the pedestrian did not pass. It ispass.
actually very Itdifficult
is veryto determine
difficult an exact RP
to determine an at
the time of turn event; thus, it is also impossible to divide L2 and L2’ to generate
exact RP at the time of turn event; thus, it is also impossible to divide L2 and L2’ to generate a a reference trajectory.
reference trajectory.

(a) (b)
location using a reference trajectory comprising a combination of L1, L2, and candidate L3 before the
turn back event. It is possible to generate a reference trajectory by combining L1, L2, L2’ and L1’. But,
in this case, it is impossible to perform an accurate comparison between Surfaces because the reference
trajectory contains RPs that the pedestrian did not actually pass. It is very difficult to determine an
exact RP at the time of turn event; thus, it is also impossible to divide L2 and L2’ to generate a
Sensors 2019, 19, 3325 11 of 19
reference trajectory.

(a) (b)

(c) (d)
Figure 11. The localization of the Surface Correlation (SC) method in case of a turn back event: (a) Links
for the generation of a reference trajectory; (b) Test route for a turn back event; (c) SSD; (d) Localization
error by comparison with GPS.

For these reasons, the SC method performs an initialization process that resets both the measured
Surface and the current reference trajectory when a turn back event occurs as shown in Figure 11b.
However, the resetting of the measured Surface may cause ambiguity as in conventional fingerprinting
methods as Figure 11c and d. The length of the Surface is a very important factor in identifying
the reliability of the location information through SC, along with the SSD. To mitigate the unstable
performance due to the ambiguity after the turn event, the SC method fixes the reverse component
of a current link as the reference trajectory and an RP as a start point when the turning event occurs.
Then, it determines the location using length of the Surface until its length (lt++∆t ) is greater than the
0
length of the previous Surface (l−
t0 ). Let tepoch denote the fixed RP where the turn event has occurred and
denote the total number of RPs on the reference trajectory. If lt++∆t /l− t0 ≥ 1, the SC method determines
0
the location as an RPa* with the minimum SSD as presented below:

a∗ = arg min ρ[1,tepoch ] (6)

1≤a≤tepoch ≤mr

4. Experiments and Results

4.1. Test Environment

To verify the proposed fingerprinting method, we performed tests in Barcelona’s old districts
with a common commercial smartphone as shown in Figure 12. Before the test, we constructed a
fingerprinting database consisting of 20 links and 13 nodes through several surveys of the test area.
4.1. Test Environment
To verify the proposed fingerprinting method, we performed tests in Barcelona’s old districts
with a common commercial smartphone as shown in Figure 12. Before the test, we constructed a
Sensors 2019, 19, 3325 12 of 19
fingerprinting database consisting of 20 links and 13 nodes through several surveys of the test area.

Figure 12. Test Site.

Figure 12. Test Site.
As shown in Figure 13a, four BSs are available on average on each link. Over 10 BSs were available
As shown in Figure 13a, four BSs are available on average on each link. Over 10 BSs were
in a square or a relatively wide alley, but only one BS was available in narrow alleys. In addition,
available in a square or a relatively wide alley, but only one BS was available in narrow alleys. In
measured RSS detected as very low in most RPs, as shown in Figure 13b. As shown in Figure 14a
addition, measured RSS detected as very low in most RPs, as shown in Figure 13b. As shown in
and b, tests were performed in two scenarios as follows: (1) From an open space into a narrow alley;
Figure 14a and b, tests were performed in two scenarios as follows: (1) From an open space into a
(2) From the narrow alley to an alley. In order to analyze the performance in each scenario, localization
narrow alley; (2) From the narrow alley to an alley. In order to analyze the performance in each
results from PDR were corrected by applying a map-matching algorithm and it was considered as a
scenario, localization results from PDR were corrected by applying a map-matching algorithm and it
true location. Collected data from the smartphone were as follows:
was considered as a true location. Collected data from the smartphone were as follows:
• Inertial Measurement Unit (IMU): 40 Hz
• Inertial Measurement Unit (IMU): 40 Hz
• PCI & RSS of LTE: 0.5 Hz
• PCI & RSS of LTE: 0.5 Hz
•• GNSS:
Sensors 1xHz
2019, 19,1
GNSS: Hz
FOR PEER REVIEW 12 of 18

(a) (b)
Figure 13. The
Figure The condition
conditionofofthe Long-Term
the Long-TermEvolution (LTE)
Evolution environment
(LTE) in thein
environment testthe
site: (a)site:
test The (a)
number
The
of available
number BSs at each
of available BSslink; (b) link;
at each The maximum RSS at each
(b) The maximum RSSRP.
at each RP.
 (a) (b)

Figure 13. The condition of the Long-Term Evolution (LTE) environment in the test site: (a) The
Sensorsnumber
2019, 19,of
3325
available BSs at each link; (b) The maximum RSS at each RP. 13 of 19

(a) (b)
Figure 14. Test
Test scenarios: (a)
(a) Open
Open space
space into
into narrow
narrow alley; (b) Narrow alley to alley.

4.2. Performance Analysis

4.2. Performance Analysis according
according to
to the
the Length
Length of
of Surface
Surface
The length
The lengthofofthe Surface
the is the
Surface is most important
the most factor factor
important determining the performance
determining of the proposed
the performance of the
fingerprinting
proposed fingerprinting method. As its length increases, the performance of the proposedbecomes
method. As its length increases, the performance of the proposed method method
more robust
becomes against
more robustsignal noise
against and distortion,
signal noise and and the ambiguity
distortion, and thecan be reduced
ambiguity canwhile performing
be reduced a
while
comparison with the RSS pattern. On the other hand, the following problems may occur
performing a comparison with the RSS pattern. On the other hand, the following problems may occuras the length
of the Surface
as the increases:
length of the Surface increases:
•
• Drift error of walking distance from PDR The walking distance is calculated by accumulating the
Thelengths
step walking distance
from is calculated
the PDR, by accumulating
and the Surface the based
is constructed step lengths
on this from the PDR,
distance. Drift and
errorthe
of
Surface
the is constructed
walking based aon
distance induces this distance.
localization error Drift
whenerror of thethe
comparing walking
measureddistance
Surfaceinduces
with thea
localization
reference RSS error when comparing the measured Surface with the reference RSS pattern.
pattern.
•
• Sensitivity degradation in in fixing
fixing reference
referencetrajectory
trajectoryatatcrossroad
crossroadAs the length of Surface increases,
Asproportion
the the lengthof ofthe
Surface increases,
RSS pattern thecommon
to the proportion of the trajectory
reference RSS pattern to the than
is higher common
that ofreference
the RSS
trajectory is higher than that of the RSS pattern after a crossroad. As a result, the
pattern after a crossroad. As a result, the similarity of RSS patterns between reference trajectoriessimilarity of
RSS patterns
increases and thebetween reference
ratio of correlationtrajectories
coefficientincreases
becomesand the ratio
smaller. of correlation
This serves coefficient
as a delay in fixing
becomes smaller. This serves as a delay in fixing the reference trajectory, which
the reference trajectory, which may lead to a problem of fixing the previous reference trajectory may lead to a
problemthe
beyond of fixing the previous reference trajectory beyond the next crossroad.
next crossroad.
We selected the optimal length by comparing the localization errors for different lengths of
We selected the optimal length by comparing the localization errors for different lengths of Surface.
Surface. In order to correctly analyze the accuracy according to the length of the Surface, localization
In order to correctly analyze the accuracy according to the length of the Surface, localization was
was performed only for the reference path in Figure 8b while the pedestrian moved as shown in
performed only for the reference path in Figure 8b while the pedestrian moved as shown in Figure 7a.
As shown in Figure 15, accuracy improves as the length of the Surface increases up to 72 m.
 Sensors 2019, 19, x FOR PEER REVIEW 13 of 18

Figure
Sensors 7a.19,As
2019, shown
3325 in Figure 15, accuracy improves as the length of the Surface increases up
14 to 72
of 19
m.

Figure 15.The
Figure15. TheCDF
CDFofoflocalization
localizationerror
erroraccording
accordingtotothe
thelength theSurface.
lengthofofthe Surface.

The
Theaccuracy
accuracydecreases
decreasesasas
thethe
length
lengthincreases more
increases than
more 72 m.
than 72 Therefore, we selected
m. Therefore, 72 m72
we selected asm theas
maximum
the maximumlengthlength
of the of
Surface in this study.
the Surface in thisInstudy.
addition, a reference
In addition, trajectory trajectory
a reference was set to be
was combined
set to be
with up to three links considering the length of the Surface and that of the shortest link.
combined with up to three links considering the length of the Surface and that of the shortest This limitation
link.
also
Thishas the effectalso
limitation of bounding the increase
has the effect of computation
of bounding complexity
the increase in the comparison
of computation complexity between
in the
measured
comparison Surface and measured
between reference patterns.
Surface and reference patterns.

4.3. Test Result from Open Space into Narrow Alley

4.3. Test Result from Open Space into Narrow Alley
In this scenario, a test was performed in an open space with good GNSS conditions and LTE
In this scenario, a test was performed in an open space with good GNSS conditions and LTE
environment as shown in Figure 13. To verify the performance of the proposed fingerprinting method,
environment as shown in Figure 13. To verify the performance of the proposed fingerprinting method,
we compared its result with the GPS, the conventional fingerprinting method (kNN), and a Particle
we compared its result with the GPS, the conventional fingerprinting method (kNN), and a Particle
Filter (PF) as shown in Figure 16. We set the k as 5 to estimate the location from the kNN fingerprinting
Filter (PF) as shown in Figure 16. We set the k as 5 to estimate the location from the kNN
method. We utilized the heading and stride information from PDR on a resampling process of the PF
fingerprinting method. We utilized the heading and stride information from PDR on a resampling
for the particle propagation. Additionally, 150 particles were used to compute the location. As shown
process of the PF for the particle propagation. Additionally, 150 particles were used to compute the
in Figure 16a–d, the kNN-based fingerprinting and PF show a relatively stable performance in a region
location. As shown in Figure 16a–d, the kNN-based fingerprinting and PF show a relatively stable
with good LTE environment. However, in a region with poor LTE signal environment, the performance
performance in a region with good LTE environment. However, in a region with poor LTE signal
is very unstable due to increased ambiguity. The GPS showed a relatively stable performance when
environment, the performance is very unstable due to increased ambiguity. The GPS showed a
entering a narrow alley after convergence of performance in open space. However, it can be seen that a
relatively stable performance when entering a narrow alley after convergence of performance in open
multipath effect on the satellite signal in the alley deviates it substantially from the actual route. The SC
space. However, it can be seen that a multipath effect on the satellite signal in the alley deviates it
method shows a very stable performance on the overall test route. However, at the initial stage of the
substantially from the actual route. The SC method shows a very stable performance on the overall
test, we confirmed that the location was matched to a different link than the actual link, just like the
test route. However, at the initial stage of the test, we confirmed that the location was matched to a
kNN as in Figure 17a. As shown in Figure 17b, this is because the RSS pattern for the two reference
different link than the actual link, just like the kNN as in Figure 17a. As shown in Figure 17b, this is
trajectories is very similar near the starting point and sufficient RSS pattern is not accumulated for
because the RSS pattern for the two reference trajectories is very similar near the starting point and
ambiguity reduction. It can be seen that the localization result is very stable after 35 m point where
sufficient RSS pattern is not accumulated for ambiguity reduction. It can be seen that the localization
the RSS pattern is accumulated sufficiently and ambiguity decreases. Except for the initial phase,
result is very stable after 35 m point where the RSS pattern is accumulated sufficiently and ambiguity
we can confirm that the SC-based fingerprinting method is very stable and accurate in most areas as
decreases. Except for the initial phase, we can confirm that the SC-based fingerprinting method is
demonstrated in Figure 18a,b. The Root Mean Square Error (RMSE) of the GPS is 10.86 m and of the
very stable and accurate in most areas as demonstrated in Figure 18a and b. The Root Mean Square
conventional fingerprinting method is 41.22 m, respectively. The RMSE of the PF is 28.59 m while the
Error (RMSE) of the GPS is 10.86 m and of the conventional fingerprinting method is 41.22 m,
surface correlation method has an RMSE of 4.67 m except for the initial error.
respectively. The RMSE of the PF is 28.59 m while the surface correlation method has an RMSE of
4.67 m except for the initial error.
Sensors 2019, 19, x FOR PEER REVIEW 14 of 18
Sensors2019,
Sensors 19,x3325
2019,19, FOR PEER REVIEW 1415ofof1819

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

(c) (d)
(c) (d)
Figure 16. Scenario 1 test results: (a) GPS; (b) Conventional fingerprinting (kNN); (c) Particle Filter
Figure
Figure
with 16.
PDR; Scenario
SC. 11test
Scenario
16.(d) testresults:
results:(a)
(a)GPS;
GPS;(b)
(b)Conventional
Conventionalfingerprinting
fingerprinting(kNN);
(kNN);(c)
(c)Particle
ParticleFilter
Filter
with
withPDR;
PDR;(d)(d)SC.
SC.

(a) (b)
(a) (b)
Figure
Figure17.
17.Initial localization
Initial localizationerror:
error:(a)(a)
Result of of
Result initial localization;
initial (b)(b)
localization; RSS
RSSpattern
patternofofreference
reference
Figure 17. from
trajectories
trajectoriesInitial localization
fromsixsix
Base error:
BaseStations (BSs).
Stations (a) Result of initial localization; (b) RSS pattern of reference
(BSs).
trajectories from six Base Stations (BSs).
 Sensors 2019, 19, x FOR PEER REVIEW 15 of 18
Sensors
Sensors 19,19,
2019,
2019, x 3325
FOR PEER REVIEW 15 16
of of
1819

(a) (b)
(a) (b)
Figure 18. Localization error on test scenario 1: (a) Localization error according to the walking
Figure
Figure18. Localization
Localization
18.(b)
distance; CDF of theerror ontest
on testscenario
localization scenario
error. 1: (a)
1: (a) Localization
Localization errorerror according
according to the to the walking
walking distance;
(b) CDF(b)
distance; of the
CDFlocalization error. error.
of the localization
4.4. Test Result from Narrow Alley to Alley
4.4.Test
4.4. TestResult
Resultfrom
fromNarrow
NarrowAlleyAlleytotoAlley
Alley
In order to verify the performance in an area where the LTE environment is poor, the test was
InInorder
performed
ordertotoverify
verifythe
with the same
theperformance
performanceininananarea
route as shown
area wherethe
earlier inwhere
theLTE
Figure 14b.
LTEenvironment
environmentisispoor,
This test area had
poor,the
only one
thetest
testwas
was
available
performed withthe
performed thesame
sameroute
route asshown
shown earlier
earlier in in Figure
Figure 14b. This test area had
had only one
oneavailable BS
BS as seenwith in Figure 13a, andasalso had poor GPS signal14b. This
conditions test as
areashown only
in Figure available
19a. The
BSas seen in Figure
as seen initial
in Figure 13a, and
13a, andalso had poor GPS signal conditions as shown in Figure 19a. The estimated
estimated location fromalso
GPShad waspoor GPS signal
at origin because conditions
it did notasreceive
shownsatellite
in Figure signal19a.atThe
the
initial
estimated location from GPS was at origin because it did not receive satellite signal at the beginning ofthe
the
beginninginitial
of thelocation
test. In thefrom GPS
case waskNN
of the at origin
and the because it did notwas
PF, localization receive
almostsatellite signalinatareas
impossible
test.
beginningIn the case of the kNN and the PF, localization was almost impossible in areas where the LTE
where theofLTE theenvironment
test. In the case wasofvery
the kNN
poor withand the a fewPF,BSs
localization
while thewas almostaimpossible
SC shows stable and in areas
accurate
environment
where the LTE was very poorwas
environment withvery
a few BSswith
poor while a the SC
few BSs shows
while athe
stable
SC and accurate
shows a stable performance
and accurate as
performance as seen in Figures 19b through 19d. In test scenario 2, there was only one available BS;
seen in Figure 19b–d.
performance inIn test scenario 2, there19d. was only one available BS; therefore, 24 more steps were
therefore, 24as seensteps
more Figures 19b through
were needed to stabilizeIn thetest scenario
performance 2, than
there inwas
testonly one
scenario available
1 as shown BS;in
needed
therefore, to
24stabilize
more the
steps performance
were needed than
to in test
stabilize scenario
the 1 as
performance shown in
than Figure
in test 20. The
scenario RMSE
1 as of the GPS
shown in
Figure 20. The RMSE of the GPS is 18.52 m and the kNN is 53.04 m, respectively. The PF shows RMSE
is 18.52
Figure m and
20.mThe the
RMSE kNN is 53.04 m, respectively. The PF shows RMSE at 31.73 m while the SC-based
at 31.73 while theof the GPS fingerprinting
SC-based is 18.52 m and the kNN shows
method is 53.04itm, at respectively.
7.85 m except TheforPFtheshows
initialRMSE
phase
fingerprinting
aterror.
31.73Them while method shows itfingerprinting
the SC-based at 7.85 m except for theshows
method initial it
phase error.
at 7.85 The testfor resultsinitial
for thephase
above
test results for the above two scenarios are summarized in m except
Table 1, and thethey can confirm
two
error. scenarios
The proposedare summarized
test resultsmethod
for the abovein Table 1, and they can confirm that our proposed method shows better
that our showstwo scenarios
better or similarare summarized
performanceinwith TableGPS 1, andeventheyin can
a poorconfirm
LTE
or similar
that performance
our proposed method withshows
GPS evenbetterin ora poor LTE performance
similar environment.with Especially,even the availability
poor of the
environment. Especially, the availability of the RF coverage is definitelyGPS improved in byautilizing LTEthe
RF coverage
environment. is definitely improved by utilizing the accumulated RSS pattern visible in the results of
accumulated Especially,
RSS pattern thevisible
availability
in theofresults
the RFofcoveragescenariois 1definitely
and scenario improved2 with by different
utilizing the LTE
scenario
accumulated 1 and RSSscenario 2 with
pattern visible different LTE
in thethat environmental
results of scenario conditions.
1 and scenario This implies that
2 with different it is possible
LTEa
environmental conditions. This implies it is possible to maintain a high performance with only
to maintain
environmental a high performance with only a small number of BSs using the proposed fingerprinting
small number conditions.
of BSs usingThis the implies
proposed that it is possible method
fingerprinting to maintain evena inhigh performance
urban areas. with only a
method
small evenof
number inBSs
urban areas.
using the proposed fingerprinting method even in urban areas.

(a) (b)
(a) (b)
Figure 19. Cont.
Sensors 2019, 19, x FOR PEER REVIEW 16 of 18
Sensors 2019, 19, 3325 17 of 19
Sensors 2019, 19, x FOR PEER REVIEW 16 of 18

(c) (d)
(c) (d)
Figure 19. Scenario 2 test results: (a) GPS; (b) Conventional fingerprinting (kNN); (c) Particle Filter
Figure
Figure
with 19. Scenario
Scenario
19.(d)
PDR; SC. 22 test
test results:
results: (a)
(a) GPS;
GPS; (b)
(b) Conventional
Conventional fingerprinting
fingerprinting(kNN);
(kNN);(c)
(c) Particle
Particle Filter
Filter
with
with PDR;
PDR; (d)
(d) SC.
SC.

(a) (b)
(a) (b)
Figure
Figure20.
20. Localization error on
Localization error ontest
testscenario
scenario 2: Localization
2: (a) (a) Localization
errorerror according
according to the walking
to the walking distance;
Figure 20.
distance; (b) Localization
CDF of the error on test
localization scenario 2: (a) Localization error according to the walking
error.
(b) CDF of the localization error.
distance; (b) CDF of the localization error.
Comparisonofofthe
Table1.1.Comparison
Table theperformance.
performance.
Table 1. Comparison of the performance.
Scenario
Scenario GPS
GPS kNN
kNN PFPF SC SC
Scenario 10.15GPS
10.15 m (50%)
m (50%) 15.47
15.47mkNN
m(50%)
(50%) 11.93 PF
m (50%)
11.93m (50%) 3.423.42 mSC
m (50%)
(50%)
1 1 10.15 m(80%)
(50%)
1
13.99 m
13.99 m (80%) 46.53mmm
15.47
46.53 (50%)
(80%)
(80%) 30.711.93m
mm
30.7 (80%)(50%)
(80%) 7.197.19 3.42mm(80%)
m (80%)(50%)
13.99
11.69m
11.69 m(50%)
m (80%)
(50%) 46.53
44.99
44.99 mmm (80%)
(50%)
(50%) 30.7
25.98 mm
25.98 m (50%)
(80%) 8.598.59
(50%) 7.19mm(50%)
m (50%)(80%)
2 2 11.69
15.73 m
m (50%)
(80%) 44.99
62.63 m m (50%)
(80%) 25.98
38.26 m m
(80%)(50%) 13.28 8.59
m m (50%)
(80%)
2 15.73 m (80%) 62.63 m (80%) 38.26 m (80%) 13.28 m (80%)
15.73 m (80%) 62.63 m (80%) 38.26 m (80%) 13.28 m (80%)
5.5.Conclusions
Conclusions
5. Conclusions
Accuratelocalization
Accurate localization inin urban
urban areas
areas stillstill
is aischallenging
a challenging tasktask
eveneven withwith GNSS.GNSS.
In thisInpaper,
this paper,
we
proposed Accurate
we proposed localization
a surface
a surface in urban
correlation
correlation areas
basedbased still is a challenging
fingerprinting
fingerprinting method task
method even
using using with
LTE signals GNSS.
LTE signals In
in urban this paper,
in urban
areas. we
areas.
The
proposed
The surface
surface a surface
correlation correlation
correlation
method method based
comparescomparesfingerprinting
the spatial
the spatial method
RSS RSS
pattern using
pattern
of the LTE
of the signals in
pedestrian
pedestrian urban areas.
movement
movement with The
with
a
fingerprinting database. For this comparison of patterns, it generates a reference trajectory from thea
surface
a correlation
fingerprinting method
database. compares
For this the
comparisonspatial of RSS pattern
patterns, it of the
generates pedestrian
a movement
reference trajectorywith
from
fingerprinting
the fingerprinting
fingerprinting database.
database
database Formatches
that this
that comparison
matches theofwalking
the walking patterns,
path ofpathaitpedestrian.
generates aAccuracy
of a pedestrian. reference trajectory
Accuracy
and andfrom
stability the
stability
of the
fingerprinting
of the localization
localization database
performance that
canmatches
performance becan bethe
improved walking
improved through paththe
through of acomparison
pedestrian.
the comparison Accuracy
between
between and
the thestability
RSS of the
RSSpatterns.
patterns.
localization
Moreover,the
Moreover, performance
theproposed
proposedmethodcan be
methodprovidesimproved
providesspatial through
spatialdiscriminationthe comparison
discriminationfor fora asignal between
signalofofvery the
verylow RSS patterns.
lowstrength,
strength,soso
Moreover,
thethescope
scopeofofthe proposed
applying
applying method is
localization
localization provides
isvery
verywide.spatial
wide. The
The discrimination
testresults
test resultsshow forthat
show athat
signal
the of very low strength,
theperformance,
performance, measured
measured so
inthe scope
interms
terms ofaccuracy,
ofof applyingstability,
accuracy, localization
stability, andisrobustness,
and very wide.of The
thetest results method
proposed
proposed show
method that the performance,
isissimilar
similar or
orbetter measured
betterthan
than that
thatof
in terms of accuracy, stability, and robustness, of the proposed method
of GPS even in the narrow alleys of an urban area. Additionally, it is expected that its combination is similar or better than that
of GPS even in the narrow alleys of an urban area. Additionally, it is expected that its combination
Sensors 2019, 19, 3325 18 of 19

GPS even in the narrow alleys of an urban area. Additionally, it is expected that its combination with
GNSS will make it more seamless and make accurate localization possible in urban areas. The surface
correlation method is applicable not only to LTE, but also to most RF signals that can be measured in
signal strength such as Wi-Fi, Bluetooth, and 5G; hence, it can be applied not only to urban areas but
also to various environments, such as underground or indoors.

Author Contributions: This work was undertaken with collaboration among all authors. J.L., B.S., D.S., and T.L
designed the proposed method. J.L. and T.L. performed the tests and analyzed the test results. Y.R. and D.W.
drafted the manuscript. J.P., and T.L. corrected the whole manuscript. All authors contributed and improved
the manuscript.
Funding: This research was supported by the KIST Institutional Program (Project No. 2E29490).
Acknowledgments: This work was partly supported by Institute for Information & communications Technology
Promotion (IITP) grant funded by the Korea government (MSIT) (No. 2019-0-01401, Multi-source-based 3D
Emergency LOCalization using machine learning techniques (ME LOC)). This work was partially supported by the
KIST Institutional Program (No. 2V06880) and also supported by the KIST Institutional Program (No. 2E29490).
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Ben-Moshe, B.; Elkin, E.; Levi, H.; Weissman, A. Improving Accuracy of GNSS Devices in Urban Canyons.
In Proceedings of the 23rd Canadian Conference on Computational Geometry, Toronto, ON, Canada,
10–12 August 2011; pp. 511–515.
2. Del Peral-Rosado, J.A.; López-Salcedo, J.A.; Seco-Granados, G.; Zanier, F.; Crisci, M. Analysis of Positioning
Capabilities of 3GPP LTE. In Proceedings of the 25th International Technical Meeting of the Satellite Division
of the Institute of Navigation 2012 (ION GNSS 2012), Nashville, TN, USA, 17–21 September 2012; pp. 650–659.
3. Seco-Granados, G.; Navarro-Gallardo, M.; Mıguez-Sanchez, J.; Lopez-Salcedo, J.A.; Del Peral-Rosado, J.A.;
Garcıa-Molina, J.A.; Zanier, F.; Crisci, M.; I Castillo, R.E. Performance Analysis of Hybrid GNSS and
LTE Localization in Urban Scenarios. In Proceedings of the 8th ESA Workshop on Satellite Navigation
Technologies and European Workshop on GNSS Signals and Signal Processing, Noordwijk, The Netherlands,
14–16 December 2016; pp. 1–8.
4. Mahyuddin, M.F.M.; Isa, A.A.M.; Zin, M.S.I.M.; H, A.M.A.; Manap, Z.; Ismail, M.K. Overview of Positioning
Techniques for LTE Technology. JTEC 2017, 9, 43–50.
5. Mike, T.; Ewald, Z. LTE Location Based Technology Introduction; Rohde Schwarz Inc.: Seoul, Korea,
2013; Available online: https://www.rohde-schwarz.com/kr/applications/white-paper_230854-122561.html
(accessed on 29 May 2015).
6. Kangas, A.; Wigren, T. Location Coverage and Sensitivity with A-GPS. In Proceedings of the URSI International
Symposium on Electromagnetic Theory, Pisa, Italy, 23–27 May 2004; p. 13.
7. Kunczier, H.; Anegg, H. Enhanced Cell ID based Terminal Location for Urban Area Location based
Applications. In Proceedings of the First IEEE Consumer Communications and Networking Conference,
Las Vegas, NV, USA, 5–8 January 2004; pp. 595–599.
8. Wigren, T. Adaptive Enhanced Cell-ID Fingerprinting Localization by Clustering of Precise Position
Measurements. IEEE Trans. Veh. Technol. 2007, 56, 3199–3209. [CrossRef]
9. Liu, J.; Feng, S. RSTD Performance for Small Bandwidth of OTDOA Positioning in 3GPP LTE. In Proceedings
of the 2013 IEEE 78th Vehicular Technology Conference (VTC Fall), Las Vegas, NV, USA, 2–5 September 2013;
pp. 1–5.
10. Zhang, T.; Xiao, D.; Cui, J.; Luo, X. A Novel OTDOA Positioning Scheme in Heterogeneous LTE-Advanced
Systems. In Proceedings of the 2012 3rd IEEE International Conference on Network Infrastructure and
Digital Content, Beijing, China, 21–23 September 2012; pp. 106–110.
11. Evolved Universal Terrestrial Radio Access (E-UTRA). LTE Positioning Protocol (LPP). 3GPP TS 36.355.
Available online: https://www.3gpp.org/ftp/Specs/archive/36_series/36.355/ (accessed on 22 January 2015).
12. Evolved Universal Terrestrial Radio Access (E-UTRA). LTE Positioning Protocol A (LPPa). 3GPP TS 36.455.
Available online: https://www.3gpp.org/ftp/Specs/archive/36_series/36.455/ (accessed on 22 January 2015).
Sensors 2019, 19, 3325 19 of 19

13. Gundlegard, D.; Akram, A.; Fowler, S.; Ahmad, H. Cellular Positioning using Fingerprinting based on
Observed Time Differences. In Proceedings of the 2013 International Conference on Smart Communications
in Network Technologies (SaCoNeT), Paris, France, 17–19 June 2013; pp. 1–5.
14. Zhu, J.; Luo, X.; Chen, D. Maximum Likelihood Scheme for Fingerprinting Positioning in LTE System.
In Proceedings of the 2012 IEEE 14th International Conference on Communication Technology, Chengdu,
China, 9–11 November 2012; pp. 428–432.
15. Turkka, J.; Hiltunen, T.; Mondal, R.U.; Ristaniemi, T. Performance Evaluation of LTE Radio Fingerprinting
using Field Measurements. In Proceedings of the 2015 International Symposium on Wireless Communication
Systems (ISWCS), Brussels, Belgium, 25–28 August 2015; pp. 466–470.
16. Liu, H.; Darabi, H.; Banerjee, P.; Liu, J. Survey of Wireless Indoor Positioning Techniques and Systems.
IEEE Trans. Syst. Man Cybern. Part C 2007, 37, 1067–1080. [CrossRef]
17. Torres-Sospedra, J.; Moreira, A. Analysis of sources of large positioning errors in deterministic fingerprinting.
Sensors 2017, 17, 2736. [CrossRef] [PubMed]
18. Deng, Z.; Yu, Y.; Yuan, X.; Wan, N.; Yang, L. Situation and Development Tendency of Indoor Positioning.
China Commun. 2013, 10, 42–55. [CrossRef]
19. Ascher, C.; Kessler, C.; Wankerl, M.; Trommer, G.F. Dual IMU Indoor Navigation with Particle Filter based
Map-Matching on a Smartphone. In Proceedings of the 2010 International Conference on Indoor Positioning
and Indoor Navigation (IPIN), Zurich, Switzerland, 15–17 September 2010; pp. 1–5.
20. Klepal, M.; Beauregard, S. A Backtracking Particle Filter for Fusing Building Plans with PDR Displacement
Estimates. In Proceedings of the 2008 5th Workshop on Positioning, Navigation and Communication
(WPNC), Hannover, Germany, 27 March 2008; pp. 207–212.
21. Masiero, A.; Guarnieri, F.; Pirotti, F.; Vettore, A. A Particle Filter for Smartphone-Based Indoor Pedestrian
Navigation. Micromachines 2014, 5, 1012–1033. [CrossRef]
22. Shin, B.; Yu, B.; Bang, J.; Kee, C.; Lee, T. WiFi based Robust Positioning System in Large Scale and Weak
Signal Environment. In Proceedings of the 30th International Technical Meeting of the Satellite Division of
the Institute of Navigation, Portland, OR, USA, 25–29 September 2017; pp. 2894–2897.
23. Shin, B.; Jeon, S.; Lee, J.H.; Kee, C.; Lee, T. Precise Localization Technology of Mobile Phone on a Vehicle
in Tunnel using LTE Signal based Surface Correlation. In Proceedings of the 31st International Technical
Meeting of the Satellite Division of the Institute of Navigation (ION GNSS+ 2018), Miami, FL, USA,
24–28 September 2018; pp. 190–222.
24. Lee, J.H.; Shin, B.; Lee, S.; Park, J.; Kim, J.; Lee, T. A Step Length Estimation based on Motion Recognition and
Adaptive Gait Cognition using as Smartphone. In Proceedings of the 27th International Technical Meeting of
the Satellite Division of the Institute of Navigation, Tampa, FL, USA, 8–12 September 2014; pp. 243–249.

© 2019 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/).