Content uploaded by Ben Earl Barrowes
Author content
All content in this area was uploaded by Ben Earl Barrowes on Feb 08, 2021
Content may be subject to copyright.
IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING, VOL. 4, NO. 4, DECEMBER 2011 791
Through-Wall Bio-Radiolocation With UWB Impulse
Radar: Observation, Simulation and Signal Extraction
Lanbo Liu, Zijian Liu, Student Member, IEEE, and Benjamin E. Barrowes,Memb
er, IEEE
Abstract—In this paper the cardio-respiratory signatures of
human beings were studied using both an ultra-wide band (UWB)
impulse radar system in a laboratory through-wall experiment
and a numerical simulation using the ¿nite difference time domain
(FDTD) method. Signals from both the physical experiment and
numerical simulation are processed with the Hilbert-Huang Trans-
form (HHT), a novel signal processing approach for nonlinear
and non-stationary data analysis. The results show that by using
the HHT, human respiration characteristics can be successfully
identi¿ed and differentiated for different subjects and a variety
of respiratory statuses. However, reliable detection of cardiologic
signatures requires a radar system with higher central frequency.
Our results demonstrate that this combination of UWB impulse
radar and HHT data processing has potential for through-wall
life detection and possibly other applications.
Index Terms—Biomedical signal processing, ¿nite difference
methods, ground penetrating radar, ultra wideband antennas.
I. INTRODUCTION
THROUGH-WALL life detection, with the purpose of
identifying vital signs of life-beings located behind an
obstacle, has signi¿cant meanings in military operations, rescue
efforts under extreme situations and other related ¿elds. This
technique relies on the detection of movements of human
skin and internal organs due to respiration, cardiac rhythm,
speech, or the motion of limbs and other body parts. Since
the cardio-respiratory movement exists constantly even if a
person stays otherwise quiet and motionless, detection of the
cardio-respiratory signatures stands to be the primary task in
through-wall live detection.
In the radio frequency (RF) band, detection and diagnostic
monitoring of human life signals, even behind opaque obstacles
such as concrete walls or debris, by means of using radar could
be termed as “bio-radiolocation” [1]. This technique relies on
the modulation of a reÀected radar signal by the movements
Manuscript received October 01, 2010; revised January 22, 2011 and March
14, 2011; accepted April 24, 2011. Date of publication June 20, 2011; date of
current version December 14, 2011. This work was supported in part by the U.S.
Department of Defense under Grant W913E5-07-C-008.
L. Liu is with the Department of Civil and Environmental Engineering, Uni-
versity of Connecticut, Storrs, CT 06269 USA (e-mail: lanbo@engr.uconn.edu).
Z. Liu is with the Biomedical Engineering Program, University of Con-
necticut, Storrs, CT 06269 (e-mail: zijian.liu@huskymail.uconn.edu).
B. E. Barrowes is with the US Army Corps of Engineers Cold Regions Re-
search and Engineering Laboratory, Hanover, NH 03755 USA (e-mail: ben-
jamin.e.barrowes@usace.army.mil).
Color versions of one or more of the ¿gures in this paper are available online
at http://ieeexplore.ieee.org.
Digital Object Identi¿er 10.1109/JSTARS.2011.2157461
of human beings. In summary, the signature of a human body
movement can be generated by [1]:
a) breathing with a frequency band between 0.2 and 0.5 Hz
and the thorax and chest movement amplitude of 0.5–1.5
cm;
b) heartbeat with a frequency band between 0.8 and 2.5 Hz
and the chest motion amplitude of 2–3 mm;
c) articulation or movement of the vocal apparatus (lips,
tongue, larynx);
d) movements of other body parts.
Radiolocation has been widely applied to detect respiratory
signals [1]–[8]. Its application can be achieved by using either
continuous wave (CW, [1], [2]) or the ultra-wide band (UWB)
technique [3]–[8]. This paper focuses on using the UWB time
domain impulse radar, which has been developed as a practical
tool for science and engineering applications in many ¿elds.
Compared with CW radar, UWB impulse radar radiates a very
short time duration pulse and thus intrinsically possesses a
wide spectral band, resulting in high resolution and low energy
consumption at the same time [3]. Prior to the studies on de-
tecting cardio-respiration signatures, some studies [4], [5] have
shown that UWB impulse radar system is capable of detecting
human beings by identifying the impedance contrast between
the human skin and the air. Recent research [3] also shows
the ef¿ciency of UWB impulse radar systems for capturing
human breathing motion even when the subject is otherwise
motionless. Based on these studies, Zaikov et al. [6] reported
an experiment using UWB radar for the detection of trapped
people. These key advantages led a number of researchers
[3], [6]–[8] to turn their attention to the application of UWB
impulse radar in through-wall life detection by searching the
cardio-respiration signals from a live person.
To meet the objective of detecting human vital signal in an ad-
verse environment a set of given technical requirements such as
bandwidth, power budget, dynamic range, stability etc. can be
subscribed uniformly to different radar systems for functionality
assessment [8]. For the two commonly used UWB systems, i.e.,
thetimedo
main impulse radar and the step frequency contin-
uous wave (SFCW), the impulse radar is moderately fast for data
acquisition but might have a problem with its linear dynamic
range. A way around this bottleneck might be division of the
total operational down-range in sub-ranges and operation within
these sub-ranges. On the other hand, for the stepped-frequency
continuous wave technique, while it possesses very good total
power budget and dynamic range, but suffers from a relatively
slow data acquisition, much longer time is needed than the im-
pulse system measurement time, even with a fast frequency
sweeper [8].
1939-1404/$26.00 © 2011 IEEE
792 IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING, VOL. 4, NO. 4, DECEMBER 2011
Fig. 1. The sketch of the laboratory setup for through-wall bio-radiolocation
experiment.
Nevertheless, using UWB impulse radar in human life detec-
tion is challenged by life signature identi¿cation based merely
on the subtle signal from breathing and/or heartbeat, since the
signal might be severely degraded or even completely masked
by noises and scatterings. The Hilbert-Huang transform (HHT)
[9] is a novel digital signal processing technique based on the
combination of empirical mode decomposition (EMD) and the
Hilbert spectral analysis (HSA). The HHT is designed specif-
ically for extracting subtle signal features in extremely low
signal noise ratio (SNR) environment and analyzing nonlinear,
dispersive, and non-stationary data [9]. HHT should be an
ideal approach to overcome the aforementioned dif¿culties in
through-wall bio-radiolocation. After applying the HHT to the
data generated from a radar system with pseudo-random noise
source (not a time domain impulse system), Narayanan [10]
presented results with several potential breath-signal peaks in
the frequency range of 0.4–0.8 Hz, but failed to convincingly
localize the signal due solely to breathing.
This paper is organized as follows. First, the setup and pro-
cedure of laboratory experiments using UWB impulse radar is
described in an attempt to detecting human cardio-respiration
signals. Next, the numerical simulation using the ¿nite differ-
ence time domain (FDTD) technique is presented for assisting
experiment data interpretation. Finally, both experimental and
numerical bio-radiolocation signals are analyzed with the fast
Fourier transform (FFT) and the HHT for time-frequency anal-
ysis to extract the life signals.
II. THROUGH-WALL UWB IMPULSE RADAR
BIO-RADIOLOCATION EXPERIMENT
A through-wall bio-radiolocation experiment was set up in
the environmental engineering laboratory at the University of
Connecticut. Two human subjects (LL and LZ) participated in
the experiment. As Fig. 1 demonstrated, in the test the human
subject and the UWB impulse radar were placed on opposite
sides of a concrete block wall with a thickness of 20 cm formed
by cinderblocks (no reinforcing steel bars). On one side the
UWB impulse radar antenna was coupled closely to the wall at
the same height of the subject’s chest, while on the other side,
TAB L E I
NUMBER OF RECORDING TRACES FOR EACH SUBJECT AND CONDITION
the subject was standing upright a distance of 1 m from the wall.
The subject was asked to stand quietly and as stable as possible
to reduce whole body motion.
A Sensors & Software Noggins GPR with a 1-GHz antenna
was used as the UWB impulse source and receiving system
to conduct this through-wall bio-radiolocation experiment. The
data were acquired at a rate of one trace every 0.1 second, with
stacking 8 times to minimize random noises. Hereinafter the
time for recording radar traces is referred as the recording time
with the unit of second. The time window of each recording
trace is set up to be 16 nanoseconds with 161 sample points at
a time interval of 0.1 ns. With the time zero setting at sample
point 31, there are 130 sampling points (13 ns) left as the ef-
fective recording length. It is long enough for recording radar
signals reÀected back from an object within approximately a
2-meter radius.
After setting up the experiment, a set of 226 recording traces
(at 0.1 s interval (with a total of 22.6 seconds) was collected
to de¿ne the electromagnetic ambient background reference,
without the presence of the human subject. Then radar reÀec-
tion data were collected for both of the subjects in three condi-
tions: i) normal breathing, ii) breath holding, and iii) repeatedly
speaking the words “one, two, three”. The numbers of traces
for the reference condition, as well as each condition of the two
subjects, are summarized in Table I. In general, the number of
recording traces for one status is greater than that needed for
recording 4 periods of human respiration for assuring enough
data redundancy. If ef¿ciency is of greater importance, a dura-
tion of 10 sec, i.e., the time needed for 2 periods of human
respiration, can still give satisfactory analysis results.
As an example, Fig. 2(a) shows the individual traces for the
electromagnetic background reference and the case when a sub-
ject under normal breath condition in present. Hereinafter the
time of one recording trace is referred as the travel time with
the unit of nano-second (ns). From Fig. 2(a), it is clear that the
reÀection from the backside of the wall is at 2.7 ns in the travel
time (sample number 27), implies that the bulk dielectric con-
stant of the 20-cm thick wall is about 4. The presence of the
human subject causes amplitude and phase alternations in the
later time at about 8.6 ns in travel time (sample number 86) on
the recording trace, giving a distance between the wall and the
subject of exactly 1 m. All the collected pro¿les are shown in
Fig. 2(b) with the background removed and only the reÀection
from the human subject is emphasized. The reÀections from the
subject’s body, as well as its variation along the recording time
direction, can be clearly noticed after 8.6 ns in travel time (the
relative sample number of 46 in Fig. 2(b)). This observed fea-
ture in radar pro¿les underlies the premise for studying the sub-
ject’s cardio-respiratory signature. Section IV presents a more
LIU et al.: THROUGH-WALL BIO-RADIOLOCATION WITH UWB IMPULSE RADAR: OBSERVATION, SIMULATION AND SIGNAL EXTRACTION 793
Fig. 2. (a) An example of the individual recoding traces. The trace of 130
points represents a total time window of 13 ns, with all sampling points be-
fore the time zero removed. Both normal breath (the red curve) and the back-
ground (the black curve) are superimposed together. The reÀection from the
backside of the wall occurs at sampling number 27 (2.7 ns in travel time); the
phase distortion appears after sample number 86 (travel time of 8.6 ns). (b) All
data pro¿les collected from two human Subjects (LL (top row) and LZ (bottom
row)) after eliminating the background reference for three conditions (normal
breathing, breath-holding, and speaking). In each dataset, the horizontal axis
corresponds to the recording time consisting of certain number of recording
traces (the number of traces for each case is summarized in Table I) collected
at an interval of 0.1 s. The ¿rst 40 sampling points of the recording trace (0–4
ns in travel time) was excluded in all panels to emphasize the human subject
reÀection.
detailed time-frequency data analysis to extract the cardio-res-
piratory signals.
III. FDTD NUMERICAL SIMULATION
One important tool was used to comprehend the physical ex-
periment described in the last section is the ¿nite differences
time domain (FDTD) method [11]–[14]. The staggered grid, 2D
FDTD method [11] in conjunction with the perfectly matched
layer (PML) as the absorption boundary condition (ABC) to
truncate the computation domain for suppressing unwanted ar-
ti¿cial reÀections [12] is an ef¿cient, robust algorithm and has
been widely used in electromagnetic wave propagation simu-
lations [13], [14]. With the FDTD method the UWB radar re-
sponse from a human’s cardio-respiratory movement was simu-
lated using proper material physical properties and tempo-spa-
tial discretization. First, based on a slice of a magnetic resonance
image (MRI) of a generic human chest, a quasi-static cardio-res-
piration model was constructed with 45 instants (or statuses).
This 2D model contains two cycles of breath (lung movement)
and 11 cycles of heartbeat (heart deformation). Taking the time
interval between two adjacent statuses to be 0.2 second, then the
Fig. 3. Example of 9 snapshots of a 45-status cycle of cardio-respiration model
representing 2 respirations and 11 heartbeats based on a MRI image of the cross-
section of a human chest.
TAB L E I I
RELATIVE DIELECTRIC PERMITTIVITY USED IN THE THROUGH -WALL MODEL
total recording time duration is 9 seconds. Consequently, this
model gives a breath frequency of 0.22 Hz and a heartbeat fre-
quency of 1.22 Hz. The motion amplitude of breath and heart-
beat are set to be 5–15 mm and 2–3 mm respectively. Fig. 3
shows every ¿fth snapshot out of the total of 45 statuses for
conducting the FDTD simulation.
Next, this dynamic cardio-respiration model was embedded
into the 2-D computational domain to mimic the through-wall
setup as shown in Fig. 4. The geometry is very similar to the
layout of the laboratory experiment described in the last section.
The total number of grid is 2000 1000, with a grid size of
1 mm by 1 mm to occupy a 2 m 1 m 2D area. Using the
generic values of dielectric constants for air and dry wall, as
well as the published values for human tissues [15]–[17], the
complete list of the dielectric constants for different materials at
1GHzisshowninTableII.IntheFDTDsimulation,anUWB
impulse with a central frequency of 1 GHz was transmitted from
the source which was placed on one side of the wall, while a
human chest model (Fig. 3) was placed on the other side. A
receiving point was placed 10 cm away from the source. The
total recording length was set to be 16 ns, identical to that of
the physical experiment, with a total of 8000 time steps at the
sampling interval of 0.002 ns.
Because the time elapsed from the instant the radar wave
begins radiating to the time a reÀection arrives back from the
human body is about eight orders of magnitude shorter than the
time scale of human’s cardio-respiration period (1–5 seconds),
it is a reasonable approximation to treat each of the 45 statuses
as a quasi-static snapshot as if the chest movement is ‘frozen’.
Therefore, the 45 simulations were repeated to form a pro¿le in
recording time that contains 256 traces at an interval of 0.2 sec-
onds, as shown in Fig. 5. After background removal it is clear
that the chest displacement caused by cardio-respiration mainly
794 IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING, VOL. 4, NO. 4, DECEMBER 2011
Fig. 4. (a) Layout of FDTD computational domain, where grid size is chosen
to be 1 1 mm. The location of dynamic cardio-respiration modal, wall and
UWB impulse radar (location of Tx/Rx) are set up to be as similar as the real
experiment in Section II. Perfect matched layer plays a function of absorbing
outgoing waves and reduce reÀection. (b) A snapshot of radar wave ¿eld at 6 ns
after ¿ring the source impulse. The trace at the bottom is the wave-¿eld pro¿le
along the central array denoted by the white crosses.
Fig. 5. The simulation result based on the synthetic model. The left panel (a)
shows the original record in terms of the sampler number, with the sampling
interval of 0.02 ns, and the trace number with a recording interval of 0.2 sec.
Apparently, the direct coupling dominates the record. (b) After background re-
moval, the reÀections from the human body can be clearly see at the radar wave
travel time after 9.8 ns, corresponding to the sample number of 490.
appears after 9.8 ns (sample number 490). The simulation re-
sults will guide us as a reference for searching the cardio-respi-
ration signals in our laboratory experiments data.
Fig. 6. Time domain syntheses and their frequency spectra for the result of the
FDTD synthetic simulation. (a) The signal of human respiration is the dominant
part for reÀections (9.8–10.3 ns). (b) For the record after 12 ns, the spectrum of
the heartbeat at 1.2 Hz is about the same height as that of respiration at 0.2 Hz.
A quick spectral analysis of the synthetic record (Fig. 6) in-
dicates that, as expected, the dominant frequency for the res-
piration is 0.22 Hz (Fig. 6(a)), and the heartbeat is at 1.22 Hz
(Fig. 6(b)), in exact agreement with the prescribed 45-status
cardio-respiration model shown in Fig. 3. The spectral peaks
between the peaks of breathing and heartbeat appear to be as-
sociated with the harmonics of breathing, since they do occur
at the frequencies of an integer times of the basic breathing fre-
quency (0.22 Hz).
IV. TIME-FREQUENCY ANALYSIS WITH FFT AND HHT
As described in the last two sections, a total of six pro¿les
were collected on these two subjects, LL and LZ, for three sta-
tuses, plus one record of the background without subject pres-
ence. Parallel to the physical experiment, the numerical simula-
tion has also been conducted to get 2 synthetic data sets, without
and with the human chest model included. After background re-
moval by subtracting the radar record without subject presence
from the record with the subject, a total of 6 processed radar
pro¿les formed by individual traces for 6 cases (normal breath,
breath holding, and speaking from 2 subjects) was generated
LIU et al.: THROUGH-WALL BIO-RADIOLOCATION WITH UWB IMPULSE RADAR: OBSERVATION, SIMULATION AND SIGNAL EXTRACTION 795
Fig. 7. The dataset of the A-scan traces superimposed together for Subject
LL in conditions of normal breath (241 traces), breath-holding (170 traces),
and speaking (254 traces), The top row shows the early time cases for a time
window from 0–6 ns (the corresponding sample number 1–60). The bottom row
shows the same case for a time window from 7–13 ns (with the sample number
70–130).
from the physical experiment plus the one from the FDTD sim-
ulation. Three approaches were used for data analysis: simple
examination in time domain; frequency analysis with FFT and
time-frequency analysis with HHT.
A. Cardio-Respiration Signature in Time Domain
The ¿rst approach is the most intuitive way to present the fea-
tures of the cardio-respiratory signature in time domain. With
the simple superposition of the recorded traces with one another
for all the radar pro¿les for the 3 conditions of the two subjects
as shown in Figs. 7 and 8. The top rows of Figs. 7 and 8 show
the superimposed traces of the early time corresponding to the
direct transmitter-receiver coupling and the reÀection from the
wall; while the bottom rows show the reÀection from the sub-
jects.
Since the essential physical environment should have no
time-dependent variation, the traces of early time are basically
repeating of one another in time window for both Fig. 7 and
Fig. 8. In contrast, for the later time windows as shown in the
bottom rows of these two ¿gures, apparent variation in the radar
reÀections can be found for these two subjects that make the
superimposed traces appear to be ‘thicker’. Obviously, in con-
trast to the response to the ‘static’ wall, the changes of behavior
in the recording traces are associated with the presence of the
living human subject. For Subject LL (Fig. 7) it can be seen that
the reÀections starting at 8.6 ns in travel time (sample number
86 in the bottom rows) have less variations along the time axis,
when compared with the counterpart records for Subject LZ
(the bottom row of Fig. 8, the reÀections starting from about
8.3 ns in travel time (sample number 83 in the bottom rows).
Two possible explanations for these observed differences are:
First, the chest movement of Subject LL due to respiration
is weaker than Subject LZ. Second, Subject LZ may have a
larger whole-body movement during the experiment. Further
examination for the ‘normal breath’ status of Subject LL (the
Fig. 8. The dataset of the A-scan traces superimposed together for Subject
LZ in conditions of normal breath (376 traces), breath-holding (287 traces),
and speaking (409 traces), The top row shows the early time cases for a time
window from 0–6 ns (the corresponding sample number 1–60). The bottom row
shows the same case for a time window from 7–13 ns (with the sample number
70–130).
lower left plot labeled as LLNBRH in Fig. 7) revealed that
though the time-axis variation of the entire reÀection portion
is relatively low, when compared with Subject LZ, there is
a relatively large variation in a short time window between
11–12 ns in travel time (sample number 110–120) that deserves
further investigation.
B. The Spectra of the Cardio-Respiration Signals by Fast
Fourier Transform
As the second approach the power spectra were generated
by applying the fast Fourier transformation (FFT) along the
recording time direction of the observed data for a given travel
time interval (Fig. 9). First, Fig. 9(a) shows the examples of the
time domain vital signals of the two subjects along the recording
time direction of the observed data. Then Fig. 9(b) shows the av-
eraged power spectra of the recording traces in the travel time
interval of 9–11 ns.
It can be seen that the main peak of the respiration is about
0.35 Hz for both subjects LL and LZ; while the possible heart-
beat peak is around 1.0 Hz for LL and 1.1 Hz for LZ. Around
the breadth frequency band of 0.15–0.5 Hz there are quite a few
possible harmonics for subject LL. As for the heartbeat signal,
it is much weaker and relatively harder to be con¿rmed without
further more advanced analysis for both subjects.
The peaks at low frequencies for all statuses are possibly as-
sociated with involuntary and unconscious motion of the sub-
jects. This can also be seen in time domain as shown in all panels
of Fig. 9(a), expressed as the long-period undulation of the base-
line in each panel.
C. Time-Frequency Analysis by the Hilbert-Huang Transform
As the third approach, the Hilbert-Huang transform (HHT)
time-frequency analysis was carried out onto the cardio-respira-
tory radar data. The HHT, a novel approach pioneered by Huang
796 IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING, VOL. 4, NO. 4, DECEMBER 2011
Fig. 9. (a) The examples of the observed radar data collected for Subjects LL
and LZ with three statuses: normal breathing, breath holding, and speaking. (b)
The power spectra in each status of (a). The main peak of respiration is about
0.34 Hz for LL and 0.36 Hz for LZ. The heartbeat peak is likely around 1.0 Hz
for LL and 1.1 Hz for LZ.
[9], is an adaptive time-frequency analysis tool particularly suit-
able for analyzing the non-linear and non-stationary signals. The
analysis procedure consists of two techniques in conjunction:
the empirical mode decomposition (EMD) and the Hilbert spec-
tral analysis (HSA). The EMD treats the signal as a collection of
many coexisting simpler oscillatory modes called the intrinsic
mode functions (IMFs). Each IMF has to satisfy the following
conditions:
a) the number of extremes must equal to or differ at most one
from the number of zero-crossings, which is similar to the
traditional narrow band requirements for the stationary
Gaussian process;
b) At any point, the mean value of the upper and lower en-
velopes de¿ned by the local maxima and minima, respec-
tively, is zero. These upper and lower envelopes are de-
termined by some interpolation algorithm such as cubic
splines. This condition ensures phase function for getting
the bias-free instantaneous frequency.
The EMD method separates those IMFs from the original
signal one by one by means of an algorithm called shifting
process until the residue appears to be monotonic [9]. Thus, the
Fig. 10. Result of frequency analysis for the result of synthetic simulation by
using HHT. The motion of human respiration is clearly identi¿ed as 0.22 Hz
around 10 ns in the travel time direction. The feature of the heartbeat around
1.2 Hz is relative weak, but can still visible around the travel time of 12.5 ns.
original signal is written as the sum of all IMFs. This decompo-
sition can be simply achieved in time domain, and proved to be
powerful of extracting physically meaningful features from ex-
treme noisy background [9], [10]. Next, the IMFs of each radar
pro¿le can be transformed from the recording-time domain to
frequency domain by using the Hilbert transform based spec-
tral analysis (HSA), a technique based on instantaneous param-
eters obtained from classic Hilbert transform and proved to be
a better analysis tool than the conventional FFT for non-linear,
non-stationary signals [9]. The principle and procedure of HSA
has been comprehensively reviewed in [9]. Based on this obser-
vation, the HSA was used to determine the frequency spectra of
certain IMF which mainly contain respiration feature. The result
of HHT analysis for the synthetic simulation and the observed
radar signals are shown in Fig. 10 and Fig. 11 respectively. For
the result of synthetic model in Fig. 10, the respiration is clearly
identi¿ed around 0.22 Hz and 9.5–11 ns along travel time di-
rection. In addition, a slight energy concentration can be found
around1.2Hzand12nsintraveltime,whichislikelyassoci-
ated with heartbeat but far from being a convincing evidence
of the ability to detect heartbeat by this particular UWB radar
system. These two features in general coincide with the original
model discussed in Section III, with the additional information
on the travel time location of the cardio-respiratory energy.
As to the results of real human subjects shown in Fig. 11, the
features of respiration seems to be complex, with a concentrated
distribution in both frequency and travel time directions. Since
both subjects LL and LZ stand at the same location (1 m from
the wall) during the experiment, strong reÀected energy appears
between 9–11 ns in all conditions of the two subjects. It also
clearly indicates that the frequency for normal breathing of LL
andLZisaround0.35Hz.
For the case of subjects holding breath, clearly, the energy
level is much lower for both subjects but still some energy ex-
ists, this may caused by human subjects’ whole body movement
or involuntary motions, rather than respiration. Furthermore, the
respiratory energy reduction was expressed in quite different
LIU et al.: THROUGH-WALL BIO-RADIOLOCATION WITH UWB IMPULSE RADAR: OBSERVATION, SIMULATION AND SIGNAL EXTRACTION 797
Fig. 11. The HHT time-frequency analysis results in terms of power (ampli-
tude squared) of UWB data collected upon Subjects LL and LZ for the three
conditions: normal breathing, breath holding, and speaking.
feature between these two subjects. For Subject LL, the energy
reduction is more likely expressed as lowering the energy level;
while for Subject LL, it is also expressed as the shrunk duration
of energy peak in travel time direction.
For the case of subjects speaking, comparing with the normal
breath case the adjustment of the respiratory energy is also sub-
stantially different between these two subjects. For subject LL,
the time-frequency distribution of the energy level of his respi-
ration has seldom increased, which implies that he spoke in a
relatively calm manner during the experiment. In contrast, Sub-
ject LZ has a much more powerful speaking level in lower fre-
quency band ( 0.2 Hz) in later travel time than that of LL, im-
plies that his speaking energy came from a relatively deeper part
of the chest than subject LL.
V. C ONCLUSION
In this paper, through-wall bio-radiolocation with UWB im-
pulse radar from both real experiment and numerical simula-
tion was discussed. These data pro¿les were presented in time
domain ¿rst to discuss the most obvious chest motion features;
then processed by more sophisticated methods using both the
FFT and the HHT methods. The main conclusions from this re-
search are as follows.
1) As the ¿rst order problem, the presence of human subject
can be clearly detected by UWB impulse radar system,
using both the time domain and the frequency domain
methods.
2) Using the advanced signal analysis tools (FFT and HHT)
distinct time-frequency distribution characteristics of the
respiratory energy can be well distinguished among the
three breathing conditions of each subject. Fore subject LL
the motion energy for breath-holding status has limited re-
duction than the normal breathing status, implies that either
LL did not hold the breath ¿rmly during taking the radar
data, or the whole body motion was occurred in the same
frequency band.
3) Using the advanced signal analysis tools (FFT and HHT)
distinct time-frequency distribution characteristics of the
respiratory energy can also be well distinguished between
the two subjects. For example, Subject LZ has a much more
powerful speaking level in lower frequency band ( 0.2
Hz) and later travel time than that of LL.
4) Due to its small amplitude in chest movement relative to
the resolution of the UWB radar used in this laboratory
experiment, the heartbeat signal is hard to be recognized
when compared with the respiration signal. A UWB radar
with higher central frequency may lend us more possibility
to have the cardiologic signal identi¿ed. This will be the
task for future study along the research route spearheaded
by this paper.
ACKNOWLEDGMENT
The authors are grateful to the anonymous reviewers and the
Associate Editor for their constructive comments.
REFERENCES
[1] A.S.Bugaev,I.A.Vasilyev,S.I.Ivashov,V.B.Parashin,I.K.Sergeev,
A. P. Sheyko, and S. I. Schukin, “Remote control of heart and respi-
ratory human system by radar,” Biomedical Technologies and Radio-
electronics, no. 10, pp. 24–31, 2004.
[2] S. I. Ivashov, V. V. Razevig, A. P. Sheyko, and I. A. Vasilyev, “Detec-
tion of human breathing and heartbeat by remote radar,” in Progress in
Electromagnetic Research Symp., Pisa, Italy, 2004, pp. 663–666.
[3] A.G.Yarovoy,L.P.Ligthart,J.Matuzas,andB.Levitas,“UWBradar
for human being detection,” IEEE Aero sp. Ele ctron. S yst. Ma g., vol.
21, pp. 10–13, Mar. 2006.
[4] A. M. Attiya, A. Bayram, A. Safaai-Jazi, and S. M. Riad, “UWB ap-
plications for through-wall detection,” in IEEE, Antennas Propag. Soc.
Int. Symp., Jun. 2004, vol. 3, pp. 3079–3082.
[5] I. J. Immoreev and S. V. Samkov, “Ultra-wideband radar for remote
detection and measurement of parameters of the moving objects on
small range,” in Ultra Wideband and Ultra Short Impulse Signals,Sev-
astopol, Ukraine, 2003, pp. 214–216.
[6] E. Zaikov, J. Sachs, M. Aftanas, and J. Rovnakova, “Detection of
trapped people by UWB radar,” in German Microwave Conf.,Ham-
burg, Germany, 2008.
[7] O. Sisma, A. Gaugue, C. Liebe, and J.-M. Ogier, “UWB radar: Vision
through a wall,” Telecommun. Syst., vol. 38, no. 1–2, pp. 53–59, 2008.
[8] A.G.Yarovoy,X.Zhuge,T.G.Savelyev,andL.P.Ligthart,“Com-
parison of UWB technologies for human being detection with radar,”
in Proc. 4th European Radar Conf., Munich, Germany, 2007, pp.
295–298.
[9] N. E. Huang and Z. Wu, “A review on Hilbert-Huang transform:
Method and its application to geophysics studies,” Rev. Geophys., vol.
46, no. 2, Jun. 2008.
[10] R. M. Narayanan, “Through-wall radar imaging using UWB noise
waveforms,” J. Franklin Inst., vol. 345, no. 6, pp. 659–678, Sep. 2008.
[11] K. Yee, “Numerical solution of initial boundary value problems in-
volving Maxwell’s equations in isotropic media,” IEEE Trans. An-
tennas Propag., vol. 14, pp. 302–307, May 1966.
[12] J. P. Berenger, “A perfectly matched layer for the absorption of electro-
magnetic waves,” J. Comput. Phys., vol. 114, pp. 185–200, Oct. 1994.
[13] A. TaÀove and S. C. Hagness, Computational Electrodynamics: The
Finite-Difference Time-Domain Method, 2nd ed. Norwood, MA:
Artech House, 2000.
[14] L. Liu and S. A. Arcone, “Numerical simulation of the wave-guide ef-
fect of the near-surface thin layer on radar wave propagation,” J. Env-
iron. Eng. Geophys., vol. 8, no. 2, pp. 133–141, 2003.
[15] J. D. Bronzino, The Biomedical Engineering Handbook, 2nd ed.
Boca Raton, FL: CRC Press, 1999.
[16] S. Egot-Lemaire, J. Pija nka, J.S ul’e-Suso, and S. Semenov, “Dielectric
spectroscopy of normal and malignant human lung cells at ultra-high
frequencies,” Phys. Med. Biol., vol. 54, no. 8, pp. 2341–2357, 2009.
[17] S. Saha and P. A. Williams, “Electric and dielectric properties of wet
human cortical bone as a function of frequency,” IEEE Trans. Biomed.
Eng., vol. 39, no. 12, pp. 1298–1304, Dec. 1992.
798 IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING, VOL. 4, NO. 4, DECEMBER 2011
Lanbo Liu received the B.S. and M.S. degrees in
geophysics from Peking University, Beijing, China,
and the M.S. degree in civil and environmental en-
gineering and the Ph.D. in geophysics from Stanford
University, Stanford, CA.
He was the Carnegie Fellow at Carnegie Institute
of Washington before joining the faculty of Geology
and Geophysics of the University of Connecticut. He
was the Summer Faculty Fellow at Schlumberger-
Doll Research, and at NASA’s Goddard Space Flight
Center. He is also serving as an Expert for the US
Army Corps of Engineers. He received the US Army R&D Achievement Award
for his work on radio wave propagation in complex terrains. He was a Fulbright
Scholar to Norway in 2009–2010.
He is currently an Associate Professor at the University of Connecticut. He
has more than 100 publications in peer-refereed journals, conference proceed-
ings, and technical reports. He served for Geophysics and is serving for the
Journal of Environmental and Engineering Geophysics as an Associate Editor.
His current research concentrates on numerical modeling and imaging with elec-
tromagnetic, acoustic, and seismic waves for exp loration, military, geotechnical,
environmental, and biomedical engineering applications.
Zijian Liu (S’10) received the B.S. degree in
biomedical engineering from Huazhong University
of Science and Technology, Wuhan, China, in
2008. He is currently pursuing the Ph.D. degree in
biomedical engineering at the University of Con-
necticut. His research interests include numerical
simulation and imaging in biomedical engineering
using electromagnetic and ultrasonic waves.
Benjamin Barrowes (M’99) received the B.S. and
M.S. degrees in electrical engineering from Brigham
Young University, Provo, UT, in 1999, and the
Ph.D. degree from the Massachusetts Institute of
Technology, Cambridge, MA, in 2004.
He was named top high school math student in the
state of Utah (1991), received two Rocky Mountain
Space Grant Consortium (RMSGC) grants, and
was awarded an NSF graduate fellowship. During
2004–2005, he was a Director’s funded Postdoc
at Los Alamos National Laboratory in the Physics
Division. Currently, he is a physicist with the ERDC Cold Regions Research
and Engineering Laboratory.
Heistheauthororcoauthorofover50scienti¿c publications. His research
interests center on electromagnetic wave theory and modeling with applications
including wind-wave interaction, electromagnetic scattering from the sea sur-
face as well as from random media, nano-scale energy generation techniques,
computer interface methodologies, electromagnetic induction models for non-
spherical geometries, and biological electromagnetic phenomena. Other inter-
ests include automatic code conversion/translation and arbitrary precision com-
puting.