IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 63, NO.
11, NOVEMBER 2015
Experimental Investigation of 3-D Human
Body Localization Using Wearable
Ultra-Wideband Antennas
Richa Bharadwaj, Student Member, IEEE, Clive Parini, Member, IEEE,
and Akram Alomainy, Senior Member, IEEE
AbstractThis paper presents experimental investigations and
analysis of ultra-wideband localization of body-worn antennas in
an indoor environment. Three-dimensional (3-D) tracking of wear-
able antennas placed on the upper human body is achieved using
a compact and cost-effective antenna and time of arrival (TOA)
localization techniques. Various features such as rms delay spread,
kurtosis, and signal amplitude have been used to classify the
channel type between each antenna location and base station. Non-
line-of-sight identification and mitigation techniques have been
used to enhance the accuracy of the localization technique applied.
Geometric dilution of precision of the base-station configuration
and numerical analysis on the radiation patterns for different
locations of the body-worn antenna has been studied in terms of
variation in localization accuracy. High-accuracy 3-D localization
results in the range of 12 cm are obtained with the antennas
placed on the limbs and 13 cm for the torso region. The accu-
racy achieved is 3 times better than commercially available UWB
localization and tracking systems. The significantly high-accuracy
results obtained for all antenna locations demonstrate the reliabil-
ity and suitability of UWB technology for various indoor tracking
and localization applications in the fields of healthcare, sports, and
well-being.
Index TermsBody-worn antenna, indoor localization, time of
arrival (TOA), ultra-wideband (UWB) technology.
I. I NTRODUCTION
U LTRA-wideband (UWB) is an upcoming and revolu-
tionary wireless technology for high-resolution indoor
localization and tracking applications in the field of healthcare,
medical, sports, and day to day life [1], [2]. Due to the wide-
bandwidth nature ranging from 3.1 to 10.6 GHz, UWB signals
transmit ultra-short pulses making it suitable to distinguish
between multipath components in an indoor environment in
comparison to the narrow band signals. UWB is also considered
to have fewer effects on humans because of its low-transmission
power [3], [4]. Impulse radio (IR) UWB technology has a num-
ber of advantages such as immunity to multipath interference,
Manuscript received June 19, 2015; revised September 07, 2015; accepted
September 07, 2015. Date of publication September 14, 2015; date of current
version October 28, 2015. This work was supported in part by the Queen Mary
University of London Principles Doctoral Scholarship and Engineering and
Physical Sciences Research Council (EPSRC), U.K.
The authors are with the School of Electronic Engineering and Computer
Science, Queen Mary University of London, London E14NS, U.K. (e-mail:
r.bharadwaj@qmul.ac.uk; c.g.parini@qmul.ac.uk; a.alomainy@qmul.ac.uk).
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/TAP.2015.2478455
0018-926X 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
5035
low cost, high data rate, portable, ease of implementation, and
low-energy consumption making it suitable for body-centric
and wearable networks applications [1][4].
Commercial UWB localization systems such as Ubisense,
Multispectral Solutions Inc. have an accuracy of 1015 cm with
an operating range of around 50 m [5]. Certain short-range
industrial, sports, and medical applications require significantly
higher accuracy than commercial UWB systems. There are
various body-centric applications that need more precise local-
ization accuracy such as tracking of movements of limbs and
torso while performing exercises and games, motion tracking
for gait analysis, physiotherapy or rehabilitation in the health-
care domain, following the movements of dancers and actors in
the entertainment industry etc.
Xsens offers 3-D motion tracking products based upon
miniature (MEMS) inertial sensor technology [6]. The motion
capture product uses a position aiding system (MVN motion
grid) that is based on UWB technology enabling 58 cm
positioning accuracy in an area of 20 20 m2 . Some work
related to human body tracking in the indoor environment has
been presented in [3], [4], and [7]. Results presented by Zetik
et al. [8], Low et al. [9], and Meier et al. [10] indicate that
UWB technology has the potential to achieve high centime-
ter and even millimeter accuracy levels for short-range indoor
environment localization [8][11]. Submillimetre range accu-
racy is possible using carrier-based UWB systems but it is a
complex, costly, and venue-specific set-up [5]. A number of
issues and challenges prevail for accurate localization of human
subjects in an indoor environment before the system can be
deployed for commercial applications. These include the effect
of the user presence on signal quality and behavior, portabil-
ity, and cost-effectiveness of the system, multiuser interference,
multipath effects, and mitigation of non-line-of-sight (NLOS)
propagation.
In our previous works, simulation of a human body model for
body-worn antenna localization using CST microwave studio
[12] and preliminary results of localization of human subject
in an indoor environment [13] were carried out. In [14], we
studied the localization of body-worn antennas on a human sub-
ject sitting on a chair in an uncluttered indoor environment and
the encouraging localization results (13 cm accuracy) of this
exploratory investigation motivated a further detailed study.
This paper presents detailed study and analysis of UWB
human body localization using compact wearable antennas
5036
Fig. 1. (a) Tapered slot coplanar UWB antenna. (b) Schematic of the antenna
with dimensions.
placed on the upper body using analytical and numerical tech-
niques. To the best of the authors knowledge, very limited
work is presented in the open literature in the field of local-
ization of body-worn antennas using UWB technology. The
experimental evaluation is achieved through measurements that
are performed in realistic and cluttered indoor environments.
In addition, the effects of the indoor environment and pres-
ence of the human subject in the localization area on the UWB
indoor propagation channels are investigated and incorporated
in the localization methods. The main objective of this work is
to achieve high-accuracy localization of the body-worn anten-
nas using channel information from statistical analysis and time
of arrival (TOA) positioning techniques.
This paper is organized as follows. The localization mea-
surement set up and scenarios are presented in Section II.
Section III describes the TOA techniques and algorithms pro-
posed for accurate estimation of the position of the body-worn
antennas. The results are analyzed and discussed in Section IV
in terms of accuracy achieved, channel characterization, mul-
tipath components for different antenna locations on the body,
and base-station configuration set up. Section V concludes this
paper with key findings and contributions.
II. UWB L OCALIZATION M EASUREMENT S ET U P
The body-worn antennas localization was performed over the
frequency band of 310 GHz using compact and cost effec-
tive UWB tapered slot antennas (TSAs) [15]. The TSA (Fig. 1)
operates in the frequency range 2.211 GHz with excellent
impedance matching, constant gain, and radiation performance
across the whole band [15]. The total antenna size is 27 mm
16 mm with a gain of 2 to 2 dBi and efficiency of around 80%
in free space and 63% when placed on the body.
Measurements were performed in the Body-Centric Wireless
Sensor (BCWS) Lab, Queen Mary University of London [13] to
take into account the effects of the indoor environment on the
radio propagation channel. Real-human test subject of height
1.68 m and average build was chosen for the experimental
study. Eight body-worn antenna locations were chosen with six
at the joints of the arm and two on the torso. The distance
between the human body surface and the body-worn antenna
is around 35 mm, which was chosen, based on parametric
IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 63, NO. 11, NOVEMBER 2015
Fig. 2. Measurement set up used for localization of the body-worn antennas
in an area of 1.5 1.5 m2 . The base stations (red circle) and the body worn
antenna (green circle) are connected to the four-port vector network analyzer
(PNA-X, N5244A).
analysis to ensure stability of performance across the frequency
band when the movements occur. The antennas were placed
on the shoulder, elbow, and wrist for the limbs. For the torso
region, they were placed at the centre of the chest and waist.
The locations were chosen considering the suitability and ease
of placement of potential wearable antennas on a human subject
while in motion or performing regular daily activities. In addi-
tion, the upper body is generally at a level higher than common
furniture pieces and obstacles presented in an indoor environ-
ment. Therefore, this results in better link between the body-
worn antennas and base stations. Besides the presence of the
human subject in the indoor environment, there were various
objects/furniture present in the vicinity of the localization area.
There are some works present in the open literature that take
into consideration, identification, and material characterization
of such objects for propagation modeling [16][18].
The TSA antenna was used as the body-worn antenna
(mobile station, MS), which acts as the transmitter (Tx) and
the four base stations (BS), which are the receiver (Rx). The
antennas were connected to an Agilent four-port PNA-X (pro-
grammable vector network analyzer), model number N5244A
[19], by calibrated low-loss coaxial cables to measure the trans-
mission response (S21 ) between the transmitting (MS) and
receiving (BS) antennas. The PNA is set to 6400 data sam-
ples, which are sufficient to capture all the necessary impulse
response information required to provide appropriate statis-
tical data set considering Doppler shift caused by natural
unpredictable movements. The base stations are positioned
in a cuboid shape configuration [14] as shown in Fig. 2 to
obtain high-accuracy positioning in three dimensions (3-D).
The antenna at BS1, see Fig. 2, is considered as the zero coor-
dinate point, which acts as the reference to find the position of
the target placed inside the volume of the cuboid. The mobile
station (MS) is moved in 49 different positions with spacing
of 15 cm in an area sized 1.5 1.5 m2 for each wearable
antenna location. Cartesian coordinates and directional infor-
mation related to the position of the target are obtained through
TOA data fusion and peak detection algorithms.
BHARADWAJ et al.: EXPERIMENTAL INVESTIGATION OF 3-D HUMAN BODY LOCALIZATION 5037

III. TOA T ECHNIQUES A ND A LGORITHMS

A. Algorithm Description and Parameters
The high-time resolution of the UWB signals makes time of
arrival (TOA)-based approach most suitable for ranging and
localization [20][21] for wireless sensor networks. For the
TOA-based technique, the distance information is extracted
from the propagation delay between the Tx and Rx. The TOA
between the mobile and base stations is estimated by chan-
nel impulse response (CIR), statistical information derived
from the propagation environment and peak detection tech-
niques. Firstly, the S21 parameters between the MS and each
receiver are measured on the analyzer and Inverse Fast Fourier
Transform (IFFT) is then applied to obtain the CIR of the
measured channels. The CIR is given by [22]
K

h (, t) = ak (t)( k )ejk (t) (1)
k=1

where is the Dirac delta function, K is the number of resolv-

able multipath components, k are the delays of the multipath
components, ak are the path amplitude values, and k are the
path phase values.
Fig. 3 shows a flowchart of the proposed localization algo-
rithm, which is based on the CIR and TOA localization estima-
tion technique. First steps include computation of the CIR and
received signal waveforms using IFFT and convolution tech-
niques. Further channel classification is performed using var-
ious features that can distinguish between line-of-sight (LOS)
and NLOS situations. Based on these results, TOA estimation
is performed using different peak detection techniques for LOS
scenarios and threshold-based algorithms for NLOS scenar-
ios. Afterward, 3-D localization results are computed based on Fig. 3. Proposed localization scheme for body-centric UWB localization in
realistic indoor environment with multipath and NLOS situations.
the range estimates using least-square data fusion techniques.
Various steps and techniques used to obtain high-accuracy
localization results are briefly described in Fig. 3.
In a cluttered indoor environment, the radio channel propa-
gation is affected by the presence of objects, furniture, walls,
people, glass windows, etc. leading to phenomenon such as
refraction, reflection, and interference. The complex propaga-
tion phenomenon gives rise to number of error sources that
may degrade the accuracy of the TOA estimation between
each BS and MS, such as thermal noise, multipath propaga-
tion, direct path (DP) blockage between the Tx and Rx and DP
excess delay [23], [24]. Different channel conditions such as
LOS, partial NLOS (PNLOS), and NLOS have been observed
from the measured data for each antenna location on the body
(Fig. 4) [25]. From the measured data, it can be observed
that NLOS situations significantly degrade the ranging perfor-
Fig. 4. Different kind of CIR observed: line of sight direct path and detectable
mance, which is due to the large positive bias (high-multipath non-line of sight path.
situations and delay in propagation) occurrence in such chan-
nels. Fig. 4 depicts CIRs obtained in the LOS and NLOS
condition supporting the observations. Hence, it is necessary square (rms) delay spread, and kurtosis are chosen to iden-
to identify and mitigate the presence of NLOS effects to obtain tify different channel situations for each MS- and BS-link [25],
accurate localization results. [26]. The tradeoff between estimation accuracy, complexity,
Considering that LOS and NLOS channels have very differ- and sensitivity to the channel parameter choice for different
ent characteristics, radio channels, and received signal wave- body-worn antenna locations is also discussed. In some situa-
forms information such as maximum amplitude, root mean tions, where there are more BS antennas, the BS with high level
5038
Fig. 5. Graphical comparison of amplitude of the received signals for the dif-
ferent antenna locations on the body with respect to the base-station locations.
of NLOS can be neglected and localization can be performed
with the LOS BS only, leading to higher accuracy results. In
the indoor localization measurements performed in the BCWS
Lab, only four BS antennas are used to obtain high-accuracy
3-D localization of the wearable antennas. In such a scenario,
it is difficult to neglect the NLOS BSs and hence it is impor-
tant to diminish the NLOS effect. Later, a description is given
on how to mitigate NLOS scenarios and obtain accurate range
estimates for the concerned link.
B. NLOS Identification
1) Amplitude of Received Signal: Sine modulated Gaussian
pulse is used as the UWB source pulse and the frequency range
is from 3 to 10 GHz. In NLOS conditions, signals are con-
siderably more attenuated, weak, and have smaller energy and
amplitude due to propagation through obstacles or obstructions.
Fig. 5 shows the strength of the received signal versus posi-
tion of the base station averaged over the 49 grid locations for
each mobile antenna location. The relation between the Rx sig-
nal and the body-worn antenna position is a complex integrated
effect of, the distance of the antenna from the BS, height of
the BSs, position of the wearable antenna and the extent of the
NLOS situation presented between the wearable antenna and
BS.
Results and analysis based on the received signal amplitude
(Fig. 5.) demonstrate that the peak amplitude magnitude on
average is 11 times less for NLOS situations (BS3 and BS4)
in comparison to LOS situations (BS1 and BS2). The lowest
amplitude is observed for the BS located on the chest as the
BS4 is completely in out of sight scenario from the body-worn
antenna. This is due to the fact that BS4 is behind the human
subject making it in NLOS situation. BS4 in general shows min-
imum signal amplitude as it is in NLOS situation with most of
the wearable antennas and is at a lower height (0.6 m) leading
to reduction in signal strength and high multipath propagation.
Higher values of amplitude are observed for BS2 and espe-
cially for the antennas placed on the left elbow, shoulder, and
the chest, which is due to the fact that they are in LOS with
the BS.
2) RMS Delay Spread: The radio channel is commonly
characterized by the first and second central moment of the
power delay profile (PDP), i.e., the mean excess delay and the
IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 63, NO. 11, NOVEMBER 2015
Fig. 6. RMS delay spread for various receiver positions for the eight body worn
antenna locations.
rms delay spread, and by the maximum excess delay [26]. The
rms delay spread is defined as

2 2
k (k m ) .|h (k ; d)|
rms =  2 (2)
k |h (k ; d)|
where ak are the multipath amplitudes and k are the multipath
delays relative to the first arriving multipath component and d is
the separating distance between Tx and Rx. These parameters
help to easily distinguish between LOS and NLOS link as the
rms values for NLOS scenarios is much larger than the LOS
situations.
As shown in Fig. 6, the highest multipath is generally
observed for the situation when the antennas are placed on
the torso region (i.e., waist and chest) especially when there
is NLOS situation between the antenna and the base station. In
the conducted measurements, BS3 and BS4 are at the back of
the human subject; hence, the body itself acts like an obstruc-
tion causing delay and high multipath propagation. An increase
of 3.5 times in the rms delay spread is observed when the
BS-MS link is in NLOS situation in comparison to LOS situ-
ation. Highest multipath (11 ns) is observed when the antenna
is placed in the centre of the chest for BS4.
Lowest multipath (1 ns) is observed for BS1-right wrist
link and for BS2- left elbow and shoulder link, which is due
the direct path propagation between the BS and the wearable
antenna, the distance between the antennas is least and also the
antennas are facing each other bringing the antennas in LOS
situation. For the antennas placed on the right limb, higher rms
delay spread is observed for BS4 as it is in NLOS scenario with
respect to the antenna position, whereas for the left limb BS3 is
in NLOS situation, hence higher rms delay will be obtained for
BS3-MS link.
Fig. 7 shows two different cases (right elbow and chest) of the
wearable antennas locations and the rms delay spread surface
distribution over the area of localization. It can be observed for
Fig. 7(a) that BS4 values are in range of 610 ns, whereas BS1
and BS2 values are in the range of 14 ns. For BS3, distinctly
variable ranges of values are observed which is dependent on
the position of the wearable antenna. The blue region is the area
where the body-worn antenna is in NLOS situation and as it
goes further toward the other extreme end, the antenna gets in
direct view of BS3, hence tending toward LOS situation and
low rms values in the range of 24 ns are obtained. Fig. 7(b)
BHARADWAJ et al.: EXPERIMENTAL INVESTIGATION OF 3-D HUMAN BODY LOCALIZATION 5039

Fig. 8. Kurtosis index values for different body-worn antenna locations.

Fig. 7. RMS delay spread distribution for all the four base stations for different
locations when the body-worn antenna is placed on the (a) elbow and (b) chest.
The base-station locations are represented as red circles.

shows the rms distribution for an antenna placed on the chest.

For BS3 and BS4, it can be seen that the rms values are very
high in the range of 812 ns and for the other two BSs low
rms values are observed. Overall, it is observed that the body-
worn antenna placement w.r.t. BS position plays an important
role in the values obtained and the type of propagation channel
that exists between the MS and BS can be classified as LOS and
NLOS based on the rms variations.
3) Kurtosis: The NLOS scenarios have dense multipath
components and also a number of weaker components before
the occurrence of the maximum peak. Kurtosis index is pro-
posed as a simple and efficient parameter to precisely differenti-
ate between the indoor channel conditions and thereby identify
the best and worst MS-BS link in the indoor environments. The
kurtosis is a statistical parameter that indicates the fourth-order
moment of the received signal amplitude [26]. Kurtosis is
mathematically defined as follows:
 are in LOS situation showing higher values of kurtosis index
4
1 i (xi x)
(x) = 4
N ues for the antenna located on the waist is shown in Fig. 9(b).
where 4 is the standard deviation of the variable x and x is the
mean value of x. N is the number of samples of x. The kurto-
sis index is supposed to be high in case of LOS conditions
Fig. 9. Kurtosis index values and their mean values calculated over the 49 grid
locations for (a) right elbow and (b) waist antenna locations.
while it has low values for signals received under NLOS condi-
tions. Kurtosis captures both the statistics of individual channel
realizations, and the relative energy of the signal-to-noise ratio
(SNR) [27]. Fig. 8 shows the average kurtosis index values for
the different wearable antenna locations used for measurements
for each BS. It is observed that the kurtosis level for waist and
chest shows lowest level of kurtosis index in the range of 10 to
20 showing higher spread of multipath components in the CIR.
The BS3 and BS4 are in NLOS situation and BS1 and BS2
in the range of 4060. A detailed description of the index val-
(3)
Overall, it is observed that the values of kurtosis in the range of
5060 show complete LOS situation or direct path propagation
between BS and the wearable antenna. For lowest values of kur-
tosis index in the range of 1020, complete NLOS situations are
5040 IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 63, NO. 11, NOVEMBER 2015
observed. The benefit of using kurtosis as a feature of differen-
tiation between different channel scenarios is that it can classify
the situations more distinctly, i.e., differentiates between partial
LOS/NLOS and LOS/NLOS situations.
For the case of the right arm (Fig. 8), as expected higher
NLOS situation is observed for BS4, hence the average kur-
tosis values of the wrist, elbow, and shoulder are in the range of
1020. For BS3, the kurtosis values seem to increase from the
antenna placed at the wrist to the one placed on the shoulder.
The reason for this is the reduction in distance between the BS3
and the MS position when the body-worn antenna is placed near
the shoulder bringing it in more direct path situation in com-
parison to when the antenna is placed on the wrist. A detailed
description of kurtosis values for the right elbow is shown in
Fig. 9(a), displaying values for each body-worn antenna posi-
tion on the grid with respect to the base stations. For BS3 (right
elbow), there is an increase in magnitude from the wearable
antenna location 1 to antenna location 49 on the grid, which
depicts the relation between distance of the antenna from the
base station, position of the antenna location w.r.t. to the BS
bringing it from NLOS (e.g., location 1) to LOS (e.g., location
49) situation. BS1 and BS2 show higher values of kurtosis for
the right arm with BS1 showing highest values for each antenna
position as it is in direct LOS situation with the antenna posi-
tions. For similar reasons, BS2 shows higher kurtosis values for
antennas placed on the left arm. Lowest values of the kurtosis
index are observed for BS3 as it is in NLOS situation with the
antenna locations on the left arm. It is observed that the Kurtosis
values for waist and chest shows lowest level of kurtosis index
(Fig. 8) in the range of 10-20 showing higher spread of mul-
tipath components in the channel impulse response. Fig. 9(b)
shows the different kurtosis index values when the antenna is
located on the waist. The BS3 and BS4 are in NLOS situation
and BS1 and BS2 are in LOS situation showing higher val-
ues (4060) of the Kurtosis index with respect to the wearable
antenna.
All the three features aforementioned clearly assist in dif-
ferentiating between NLOS and LOS channels in a simple and
efficient way. It can be observed that kurtosis and rms delay
spread are inversely related to each other, as higher the rms
delay spread (more multipath) lower will be the kurtosis index
values. Kurtosis is an excellent feature to distinguish between
different channel conditions with a probability of more than
95% accurate classification confidence.
C. NLOS Mitigation
In urban or indoor environments, the accuracy in estimation
of TOA depends on the multipath components, NLOS situa-
tions, low SNRs, sampling precision of the analyzer, and the
environment in which measurements are taking place. Due to
the large bandwidth of a UWB signal, multipath components
are usually resolvable but can be challenging in dense multi-
path environments where the nondominant direct path estimate
corresponds to the TOA [25]. Threshold-based algorithms such
as search back technique and leading edge detection methods
[28], [29] are used to reduce the error obtained in range esti-
mation for the NLOS situations. The search back method first
finds the strongest path (SP), and then looks for a peak arriving
Fig. 10. CDF of the range error for the recorded positions for LOS and NLOS
situation using different time of arrival estimation techniques.
before the strongest path, which has greater power than a detec-
tion threshold level. A few iterations are required to obtain the
peak value nearest to the expected one or within the localiza-
tion range based on the selected threshold level. For situation of
where the expected peak value of the target is quite low, leading
edge detection algorithms can be used to distinguish between
the noise and the information regarding the range values in the
CIR measurements.
For example, Fig. 10 shows the cumulative distributive func-
tion (cdf) of the ranging error obtained for LOS scenario and
NLOS situations using leading edge and search back TOA
estimation techniques. It can be observed that for the LOS
situations the ranging error is much less than that of NLOS sit-
uations. Using leading edge detection techniques, and channel
information regarding the localization environment, the NLOS
ranging error can be reduced to a level, which is compara-
ble with the LOS ranging error, hence mitigating the NLOS
effect and enhancing the performance of the UWB localiza-
tion system. After applying TOA-based estimation techniques
for NLOS situations, higher accuracy results are obtained. In
LOS conditions, a ranging error below 3 cm occurs in more
than 90% of the measurements. On the other hand, in NLOS
conditions ranging error below 3 cm occurs in less than 20% of
the measurements and increases to 70% after applying NLOS
mitigation techniques.
D. 3-D Localization Algorithm
The range-based schemes like TOA and time difference of
arrival (TDOA) provide excellent level of accuracy due to the
high time resolution (large bandwidth) of UWB signals. The
TOA data fusion method [30], [31] is used to find the position
of the body worn antenna, which is based on combining esti-
mates of the TOA of the MS signal that arrives at four different
BSs. Let (xi , yi , zi ) represent the position of the ith base station
and ri is the range value obtained from the TOA measurement.
When the TOA estimate (depending upon LOS and NLOS sce-
nario) is multiplied by the speed of light c, ri between the Tx
and Rx is obtained. The following four equations are solved
jointly by using least square solution to estimate the position of
the target (xm , ym , zm ) via trilateration:
2 2 2
ri2 = (xi xm ) + (yi ym ) + (zi zm ) i = 1, 2, 3, 4 (4)
BHARADWAJ et al.: EXPERIMENTAL INVESTIGATION OF 3-D HUMAN BODY LOCALIZATION 5041

TABLE I
AVERAGE L OCALIZATION E RROR

IV. L OCALIZATION ACCURACY A NALYSIS

A. Wearable Antenna Localization Accuracy
Table I lists the accuracy achieved in 3-D for the eight
body-worn antenna locations chosen on the upper body when
compared with actual coordinates that have been calculated
geometrically. Highest accuracy results have been observed for
Fig. 11. Comparison of actual (blue), estimated (pink) positions of the wear-
the body-worn antennas placed on the elbows (right and left able antennas (a) right elbow and (b) waist.
arm). Accuracy results below a centimeter have been observed
in more than 50% of the antenna locations with an average
accuracy of 12 cm. Fig. 11(a) shows the localization results for ax1 ay1 az1 1
ax2 ay2 az2 1
the right elbow when compared to the actual positions. This can where H =
ax3

be attributed to the fact that the position of the elbow falls in the ay3 az3 1
centre of the volume of the cuboid in z direction, hence along ax4 ay4 az4 1
with x- and y-axis best accuracy results are also obtained in the
(xi xm ) (yi ym ) (zi zm )
z direction. Lower accuracy is observed for the antennas on the and axi = , ayi = , azi =
r ri ri
chest and waist in the range of 23 cm. Fig. 11(b) shows the i


localization results for the antennas placed on the waist, which 1 1
HDOP = (H T H) + (H T H) (6)
shows lower accuracy results. Although NLOS mitigation has 1,1 2,2


been performed and the error is drastically reduced, there are
still chances of high-range error for NLOS situations depending
on the body-worn antenna and BS position. Highest range error
is observed for the edges of the 7 7 grid where high NLOS
situations are observed. For the shoulder and wrist locations, an
average accuracy of 1.52.5 cm has been observed.
B. Geometric Dilution of Precision Analysis
In order to obtain accurate position of the target, a good
geometrical distribution of the base stations is necessary. The
accuracy achieved for the BS configuration used for localization
of the antennas can be analyzed theoretically using geometric
dilution of precision (GDOP) [32], [33], which shows the effec-
tiveness of the placement of the base stations. GDOP can be
defined as
 whole localization area. It can be seen that the best localization
1
GDOP = tr(H T H) (5)
1
V DOP = (H T H) . (7)
3,3
The BS configuration used can be scaled depending upon the
localization area of interest and the application intended. The
GDOP is computed for the different positions of the antenna
locations for each BS configuration in an area of 1.5 1.5 m2 .
Horizontal dilution of precision (HDOP) and vertical dilution of
precision (VDOP) are also computed for more detailed analysis
for all the three dimensions. The average GDOP, HDOP, and
VDOP values for cuboid shape configuration used is 2.15, 1.22,
and 1.64, respectively. It is observed that the VDOP values are
slightly higher in comparison to HDOP values. This is because
the distance of the BSs in the z direction is only 0.8 m where
as in the xy plane the distance is 1.5 m which gives better
coverage area in the xy plane.
Fig. 12 shows a slice graph for different GDOP values for the
results will be obtained for the body-worn antennas, which are
5042
Fig. 12. Sliced surface graph for GDOP values for different heights of the
antenna locations in the volume of 1.5 1.5 0.8 m3 .
Fig. 13. Azimuth radiation patterns for 6-GHz frequency: (a) chest centre and
(b) right wrist. The schematic of the human body model (top view) facing left
is also shown.
at a height of 40 cm as it lies in the centre of the volume. In
the experiment performed, the elbow is at a height of 40 cm
from BS1, hence it gives the best positioning results. Low DOP
values are obtained for all the body-worn antenna locations in
the range of values (12), which is considered as an excellent
accuracy result [33].
C. Body-Worn Antenna Radiation Pattern Numerical Analysis
Numerical analysis was carried out using CST microwave
simulations to obtain the radiation patterns when the antenna
is placed on different locations of the human body model in
free space for the 310-GHz bandwidth [12]. For the human
body model, a homogeneous tissue that comprises bone, fat,
muscle, and skin was considered. Weights assigned to calculate
weighted average of the dielectric permittivity and conductivity
are: 10% skin, 30% fat, 40% muscle, and 20% bone (average
of bone cancellous, bone cortical and bone marrow), which
resulted in an weighted averaged relative permittivity of 25.87
conductivity of 3.14 S/m at 6.5 GHz [34]. The height of the
human body model was 1.72 m and had average build. Tapered
slot UWB antenna was placed on five different locations of the
human body (wrist, elbow, shoulder, chest, and waist) and radi-
ation patterns were simulated for each case at three different
frequencies of the UWB (3, 6, and 10 GHz).
A few azimuth radiation patterns at different locations on the
body are shown in Figs. 13 and 14 for the xy plane where the
human model is shown in top-view position facing toward left.
IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 63, NO. 11, NOVEMBER 2015
Fig. 14. Azimuth radiation patterns for 3 GHz frequency: (a) shoulder;
(b) elbow; and (c) wrist. The schematic of the human body model (top view)
facing left is also shown.
It can be observed that the directivity of the pattern changes
with the position of the body-worn antenna. When the radiation
patterns are compared to the case where the body is not present
[15], the directivity increases for all the cases as the front to
back ratio of the radiation pattern is significantly increased due
to high reflections from the body at these frequencies caused by
increase in conductivity.
Overall the radiation patterns are quite similar but some vari-
ations are observed for different frequency values and position
of the body-worn antenna. For 3-GHz radiation pattern from
shoulder to wrist (Fig. 14), it can be observed that the radia-
tion pattern coverage area is increasing, e.g., for the wrist, the
blockage (shadowing) by the body is reduced to the mass of the
body presented by the wrist, hence depends on how much of
the energy is being absorbed by the body. This variation will
influence the localization accuracy as the wrist and elbow give
better localization results in comparison to the scenario when
the antenna is placed on the chest. Even for each body-worn
position when the frequency is increased from (3 6
10 GHz), the coverage area of the radiation pattern reduces.
With the increase of frequency, the electric length gets reduced
and hence, the antenna becomes slightly electrically large at
higher frequencies and more edge reflections are created and
hence more directive patterns.
V. C ONCLUSION
Human body 3-D localization using compact and cost-
effective wearable antennas based on UWB technology placed
at eight different locations on the upper body is presented in this
BHARADWAJ et al.: EXPERIMENTAL INVESTIGATION OF 3-D HUMAN BODY LOCALIZATION
paper. Simple and effective techniques for identifying and mit-
igating NLOS situations have been applied and analyzed and
a UWB localization scheme for human body tracking has been
proposed. GDOP analysis has been performed for validating the
compact base-station configuration used with results showing
GDOP values in the range of 12 which are considered as high-
accuracy values. Numerical analysis has been carried out on
the effect of body-worn antenna location. It has been observed
that the radiation patterns are more directive in the presence of
the human body and the area of the patterns decreases when
the antenna is placed on the thick muscular region (torso) in
comparison to the upper limbs. Hence, the localization accu-
racy will be higher for the limbs in comparison to the torso.
High-accuracy 3-D localization results in the range of 12 cm
are obtained with the antennas placed on the limbs and 13 cm
for the torso region as there will be more interference in the sig-
nal propagation by the torso. Best results are obtained when the
antenna is placed on the elbow (0.51.5 cm) for the BS config-
uration used, hence this location can be best suited for UWB
indoor tracking of the human subject. UWB positioning of the
human body will be further investigated using realistic indoor
environments for medical and healthcare applications using
UWB chipsets, hence making the system more cost-effective
and highly portable.
ACKNOWLEDGMENT
The authors would like to thank Mr. K. Yang from Queen
Mary University of London for his valuable support and assis-
tance with the experimental campaigns.
R EFERENCES
[1] S. Gezici and H. V. Poor, Position estimation via ultra-wide-band
signals, Proc. IEEE, vol. 97, no. 2, pp. 386403, Feb. 2009.
[2] Z. Sahinoglu, S. Gezici, and I. Guvenc, Ultra-Wideband Positioning
Systems: Theoretical Limits, Ranging Algorithms, and Protocols.
Cambridge, U. K.: Cambridge Univ. Press, 2008.
[3] W. Mekonnen, E. Slottke, H. Luecken, C. Steiner, and A. Wittneben,
Constrained maximum likelihood positioning for UWB based human
motion tracking, in Proc. Int. Conf. Indoor Positioning Indoor Navigat.
(IPIN), Zurich, Switzerland, 2010, pp. 110.
[4] D. Zhang, F. Xia, Z. Yang, L. Yao, and W. Zhao, Localization tech-
nologies for indoor human tracking, in Proc. IEEE Int. Conf. Future Inf.
Technol. (FutureTech10), Pusan, Korea, 2010, pp. 16.
[5] C. Zhang, M. J. Kuhn, B. C. Merkl, A. E. Fathy, and M. R. Mahfouz,
Real-time noncoherent UWB positioning radar with millimeter range
accuracy: Theory and experiment, IEEE Trans. Microw. Theory Tech.,
vol. 58, no. 1, pp. 920, Jan. 2010.
[6] Xsens Motion Grid. (2012). MVN MotionGrid Position Aiding for Xsens
MVN Available: http://www.xsens.com/en/general/motiongrid
[7] A. J. Ali, W. G. Scanlon, and S. L. Cotton, Pedestrian effects in
indoor UWB off-body communication channels, in Proc. Loughborough
Antennas Propag. Conf. (LAPC10), 2010, pp. 5760.
[8] R. Zetik, J. Sachs, and R. Thom, UWB localizationActive and pas-
sive approach, in Proc. 21st IEEE Instrum. Meas. Technol. Conf., 2004,
vol. 2, pp. 10051009.
[9] Z. N. Low, J. H. Cheong, C. L. Law, W. T. Ng, and Y. J. Lee, Pulse
detection algorithm for line-of-sight (LOS) UWB ranging applications,
IEEE Antennas Wireless Propag. Lett., vol. 4, pp. 6367, 2005.
[10] C. Meier, A. Terzis, and S. Lindenmeier, A robust 3D highpreci-
sion radio location system, in IEEE MTT-S Int. Microw. Symp., 2007,
pp. 397400.
[11] D. Clarke and A. Park, Active-RFID system accuracy and its implica-
tions for clinical applications, in Proc. IEEE Symp. Comput. Based Med.
Sys., Salt Lake City, UT, USA, 2006, pp. 2122.
5043
[12] R. Bharadwaj, A. Alomainy, and C. Parini, Numerical investigation of
body-worn ultra wideband antenna localisation techniques for motion
capture applications, in Proc. Prog. Electromagn. Res. Symp. (PIERS),
2013, pp. 11831187.
[13] R. Bharadwaj, A. Alomainy, and C. Parini, Indoor tracking of human
movements using UWB technology for motion capture, in Proc. 8th Eur.
Conf. Antennas Propag. (EUCAP), 2014, pp. 20972099.
[14] R. Bharadwaj, S. Swaisaenyakorn, C. G. Parini, J. Batchelor, and
A. Alomainy, Localization of wearable ultrawideband antennas for
motion capture applications, IEEE Antennas Wireless Propag. Lett.,
vol. 13, pp. 507510, Mar. 2014.
[15] A. Alomainy, A. Sani, A. Rahman, J. G. Santas, and Y. Hao, Transient
characteristics of wearable antennas and radio propagation channels
for ultrawideband body-centric wireless communications, IEEE Trans.
Antennas Propag., vol. 57, no. 4, pp. 875884, Apr. 2009.
[16] A. Safaai-Jazi, S. M. Riad, A. Muqaibel, and A. Bayram, Ultrawideband
propagation measurements and channel modeling: Through-the-
wall propagation and material characterization, DARPA NETEX,
Arlington, VA, USA, Program Rep. Nov. 2002 [Online]. Available:
http://www.antd.nist.gov/wctg/manet/docs/models.pdf
[17] J. Meng, Q. Y. Zhang, and N. T. Zhang, Impact of IR-UWB wave-
form distortion on NLOS localization system, in Proc. IEEE Conf.
Ultra-Wideband (ICUWB09), Sep. 911, 2009, pp. 123128.
[18] R. Zetik et al., Cooperative localization and object recognition
in autonomous UWB sensor networks, in INTECH, Mar. 2013,
pp. 179240.
[19] Agilent Technologies. (2013). Agilent 2-Port and 4-Port PNA-X Network
Analyzer Available: http://cp.literature.agilent.com/litweb/pdf/N5245-
90008.pdf
[20] I. Guvenc, C. Chia-Chin, and F. Watanabe, Joint TOA estimation and
localization technique for UWB sensor network applications, in Proc.
IEEE 65th Veh. Technol. Conf. VTC 07-Spring, 2007, pp. 15741578.
[21] C. Falsi, D. Dardari, L. Mucchi, and M. Z. Win, Time of arrival esti-
mation for UWB localizers in realistic environments, EURASIP J. Adv.
Signal Process., vol. 2006, pp. 113, 2006.
[22] D. Humphrey and M. Hedley, Prior models for indoor super-resolution
time of arrival estimation, in Proc. IEEE 69th Veh. Technol. Conf. VTC
Spring, 2009, pp. 15.
[23] B. Denis, J. K. Keignart, and N. Daniele, Impact of NLOS propaga-
tion upon ranging precision in UWB systems, in Proc. IEEE Conf. Ultra
Wideband Syst. Technol., 2003, pp. 379383.
[24] I. Guvenc and C. Chia-Chin, A survey on TOA based wireless localiza-
tion and NLOS mitigation techniques, IEEE Commun. Surv. Tutorials,
vol. 11, no. 3, pp. 107124, 3rd Quarter 2009.
[25] N. A. Alsindi, B. Alavi, and K. Pahlavan, Measurement and modeling
of ultrawideband TOA-based ranging in indoor multipath environments,
IEEE Trans. Veh. Technol., vol. 58, no. 3, pp. 10461058, Mar. 2009.
[26] S. Marano, W. M. Gifford, H. Wymeersch, and M. Z. Win, NLOS iden-
tification and mitigation for localization based on UWB experimental
data, IEEE J. Sel. Areas Commun., vol. 28, no. 7, pp. 10261035, Sep.
2010.
[27] L. Mucchi and P. Marcocci, A new parameter for UWB indoor chan-
nel profile identification, IEEE Trans. Wireless Commun., vol. 8, no. 4,
pp. 15971602, Apr. 2009.
[28] I. Guvenc and Z. Sahinoglu, Threshold-based TOA estimation for
impulse radio UWB systems, in Proc. IEEE Int. Conf. Ultra-Wideband
(ICU), 2005, pp. 420425.
[29] M. J. Kuhn, J. Turnmire, M. R. Mahfouz, and A. E. Fathy, Adaptive
leading-edge detection in UWB indoor localization, in Proc. IEEE
Radio Wireless Symp. (RWS), New Orleans, LA, USA, Jan. 1014, 2010,
pp. 268271.
[30] K. W. Cheung, H. C. So, W.-K. Ma, and Y. T. Chan, Least squares
algorithms for time-of-arrival-based mobile location, IEEE Trans. Signal
Process., vol. 52, no. 4, pp. 11211130, Apr. 2004.
[31] A. H. Sayed, A. Tarighat, and N. Khajehnouri, Network-based wireless
location: Challenges faced in developing techniques for accurate wireless
location information, IEEE Signal Process. Mag., vol. 22, no. 4, pp. 24
40, Jul. 2005.
[32] B. Li, A. G. Dempster, and J. Wang, 3D DOPs for positioning applica-
tions using range measurements, Wireless Sensor Netw., vol. 3, no. 10,
pp. 334340, 2011.
[33] R. B. Langley, Dilution of precision, GPS World, vol. 10, no. 5, pp. 52
59, May 1999.
[34] Italian National Research Council-Institute For Applied Physics. (1997
2015). Calculation of the Dielectric Properties of Body Tissues in the
Frequency Range 10 Hz100 GHz, IFAC-CNR, Florence (Italy) [Online].
Available: http://niremf.ifac.cnr.it/tissprop
5044 IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 63, NO. 11, NOVEMBER 2015
Richa Bharadwaj (S12) received the Bachelors
of Engineering degree (with Hons.) in electron-
ics and communication from Panjab University,
Chandigarh, India, in 2008, the M.S. degree in micro
and nanotechnologies for integrated systems from
Politecnico di Torino, Turin, Italy; INPG Grenoble,
Grenoble, France, and EPFL Lausanne, Lausanne,
Switzerland, in 2010, and the Ph.D. degree in elec-
tronic engineering (specialised in ultra-wideband
technology) from the School of Electronics and
Computer Science (Antennas and Electromagnetics
Research Group), Queen Mary University of London, London, U.K., in 2015.
She has authored a number of research papers in leading interna-
tional journals and peer-reviewed conferences. Her research interests include
ultra-wideband communication, 3-D localization, wireless sensor networks,
radio propagation characterization, miniaturised antenna design, nanosensors,
MEMS, and molecular electronics.
Clive Parini (M96) the B.Sc. and Ph.D. degrees
from Queen Mary University, London, U.K., in 1973
and 1976, respectively.
He then joined ERA Technology Ltd., U.K., work-
ing on the design of microwave feeds and offset
reflector antennas. In 1977, he returned to Queen
Mary University and is currently a Professor of
Antenna Engineering and heads the Antenna and
Electromagnetics Research Group. He has authored
over 300 papers on different research topics including
array mutual coupling, array beam forming, antenna
metrology, microstrip antennas, millimetre wave compact antenna test ranges,
millimetre wave integrated antennas, metamaterials, and on-body communica-
tions. In 2008, he co-authored a book entitled Principles of Planar Near-Field
Antenna Measurements.
Prof. Parini is a Fellow of the IET, an elected Fellow of the Royal Academy
of Engineering (2009), and a Member and past Chairman of the IET Antennas
and Propagation Professional Network Executive Team. He is a member of the
Editorial Board and past Honorary Editor for the IET Journal Microwaves,
Antennas, and Propagation. He was the recipient of the IEE Measurements
Prize for work on near-field reflector metrology, in 1990, and the BAE
SYSTEMS Chairmans Bronze Award for Innovation for work on microwave
near-field metrology, in 2002.
Akram Alomainy (M03SM13) received the
M.Eng. degree in communication engineering and the
Ph.D. degree in electrical and electronic engineer-
ing (specialized in antennas and radio propagation)
from Queen Mary University of London (QMUL),
London, U.K., in 2003 and 2007, respectively.
He joined the School of Electronic Engineering
and Computer Science, QMUL, in 2007, where he
is an Associate Professor (Senior Lecturer) in the
Antennas and Electromagnetics Research Group. He
is a Member of the Institute of Bioengineering and
Centre for Intelligent Sensing, QMUL. He is the Lead of Wearable Creativity
Research, Queen Mary University of London and has been invited to partic-
ipate at the Wearable Technology Show 2015, Innovate U.K. 2015, and also
in the recent Wearable Challenge organized by Innovate U.K. IC Tomorrow
as a leading challenge partner to support SMEs and industrial innovation. He
has authored and co-authored a book, five book chapters, and more than 150
technical papers (2700+ citations and H-index 24) in leading journals and
peer-reviewed conferences. His research interests include small and compact
antennas for wireless body area networks, radio propagation characterization
and modeling, antenna interactions with human body, computational electro-
magnetic, advanced antenna enhancement techniques for mobile and personal
wireless communications, and advanced algorithm for smart and intelligent
antenna, and cognitive radio systems.
Dr. Alomainy is a member of the IET, Fellow of the Higher Education
Academy (UK), and also a College Member for Engineering and Physical
Sciences Research (EPSRC, U.K.) and its ICT prioritisation panels. He is also a
Reviewer for many funding agencies around the world including Expert Swiss
National Science Foundation (SNSF) Research, the Engineering and Physical
Sciences Research Council (EPSRC), U. K. and the Medical Research Council
(MRC), U.K. He is an Elected Member of U.K. URSI (International Union of
Radio Science) panel to represent the U.K. interests of URSI Commission B
(September 1, 2014 until August 31, 2017). He was selected to deliver a TEDx
talk about the science of electromagnetic and also participated in many pub-
lic engagement initiatives and festivals. He was the recipient of the Isambard
Brunel Kingdom Award, in 2011, for being an outstanding young science and
engineering communicator.