1932 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 69, NO.

3, MARCH 2021

Vital Sign Detection and Radar Self-Motion

Cancellation Through Clutter Identification
Emanuele Cardillo , Member, IEEE, Changzhi Li , Senior Member, IEEE,
and Alina Caddemi , Member IEEE

Abstract— This article presents a novel technique to remove of interest, thus degrading or making the vital sign detection
the radar self-motion effects (RSMs) for an accurate detection of unfeasible. Many valuable research articles were aimed to face
human vital signs. As opposed to the commonly used techniques, this hot topic, significantly increasing the progress in this area
the proposed approach does not require any additional sensor,
and instead, it extracts the RSM from the signals reflected and achieving a higher technological maturity [11]–[16].
by stationary clutters. Since the proposed technique requires On the other hand, the problems arising from the radar self-
to accurately identify the clutter range, two procedures for motion (RSM) are less addressed even though they affect the
its automatic identification are proposed, aimed to detect both vital sign detection to the same extent. This serious issue
small and large radar motions. Besides allowing precise and can occur in a wide range of applications in which the
reliable vital sign detection, it provides a compact, lightweight,
comfortable, and cost-effective solution since potentially intrusive radar platform cannot be maintained stationary. For example,
additional sensors are not required. Simulations have been vibrations and shakes due to potholes are serious issues in both
carried out for validating the proposed approach, with an insight automotive and industrial environments, e.g., when the radar
on the influence of different clutter radar cross sections on the points toward the outside of the vehicle [17]–[21]. Different
sensitivity. Moreover, the effectiveness of the RSM cancellation applications are also characterized by the operator handshakes
has been experimentally demonstrated, showing its suitability for
different applications, e.g., radar on moving platforms, vibrating typical of handheld applications. This additional motion may
tools, handheld devices, unmanned aerial vehicles, and cars. in turn introduce further phase modulation to the received
Index Terms— Contactless sensor, ego motion, frequency- signal, making vital sign detection unfeasible [22]–[25].
modulated continuous-wave (FMCW) radar, interferome- Through the literature, some articles have addressed the
try radar, radar self-motion (RSM) cancellation, range RSM cancellation to achieve a more accurate vital signs
micro-Doppler, vital sign detection. detection.
I. I NTRODUCTION In [26] and [27], a radar aboard an unmanned aerial
vehicle was employed for detecting people in search and
D URING the last 50 years, the detection of human vital
signs with radar systems gained interest in a wide range
of applications [1]–[4]. The main concept is the contactless
rescue operations. First, the RSM effect was reduced by using
multiple additional sensors, such as GPS, LIDAR, inertial,
and ultrasonic devices to maintain a steady UAV altitude
detection of the signals reflected by a human target, which
and Yaw. Second, the residual phase distortion, measured
are modulated by the tiny movements of the chest due to
in [26] by LIDAR and ultrasonic sensors and in [27] by a
physiological activities [1]. Several innovative system-level
secondary ground radar, was subtracted from the measured
works have been reported by the microwave community
echo to recover the original breathing waveform. In [28]
concerning portable radars and applications at the human-
and [29], an RF tag was used to create a fixed reference for
microwave frontier [5]–[10]. Although great efforts have been
compensating the radar motion with the help of an adaptive
spent on various technical challenges, there are still issues to
filtering technique. Besides the limitation of requiring an
be addressed.
additional RF tag, the system was only tested for periodic radar
One of the challenging problems concerns the effect of
movements, which is overly simplified for some applications.
random body motion (RBM). Since the vital sign detection
In [30], decoupling RSM for physiological activity detection
is based on the measurement of tiny physiological motions,
with a handheld unit was addressed by placing a motion sensor
movements of the observed target might conceal the signal
on the radar antenna to record its vibrations. Removal of
Manuscript received September 22, 2020; revised November 10, 2020 and fidgeting interference for Doppler vital signs detection radar
November 25, 2020; accepted December 1, 2020. Date of publication was demonstrated in [31] and [32], which exploited empiri-
January 20, 2021; date of current version March 4, 2021. (Corresponding
author: Emanuele Cardillo.) cal mode decomposition techniques. Although representing a
Emanuele Cardillo and Alina Caddemi are with the Department of Engineer- valuable solution, it requires a relatively high computational
ing, University of Messina, 98100 Messina, Italy (e-mail: [email protected]; effort; moreover, the manual selection of the intrinsic mode
[email protected]).
Changzhi Li is with the Department of Electrical and Computer functions is often not feasible and it works properly only
Engineering, Texas Tech University, Lubbock, TX 79409 USA (e-mail: when the motion patterns are well defined. In [33], filtering
[email protected]). and adaptive noise cancellation were proposed based on prior
Color versions of one or more figures in this article are available at
https://doi.org/10.1109/TMTT.2021.3049514. knowledge of the radar motion. Besides requiring preliminary
Digital Object Identifier 10.1109/TMTT.2021.3049514 information concerning the radar motion, additional sensors
0018-9480 © 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See https://www.ieee.org/publications/rights/index.html for more information.

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY ROURKELA. Downloaded on September 27,2021 at 09:19:09 UTC from IEEE Xplore. Restrictions apply.
 CARDILLO et al.: VITAL SIGN DETECTION AND RSM CANCELLATION THROUGH CLUTTER IDENTIFICATION 1933

were again required. In [34], a bistatic radar with the receiver

placed near the human subject was employed to detect both
the signal transmitted by the radar and the echo from the
human. If the distance from the receiver to the human is
shorter than that from the transmitter to the human, an RSM
reduction might be achieved. A fundamental-and-harmonic
dual-frequency Doppler radar system was described in [35]
for canceling the RSM noise. Since the signal reflected from
Fig. 1. Transmitted signal and key parameters.
a stationary reflector includes the RSM signature, its contribu-
tion can be removed. However, this method requires accurate II. O PERATING P RINCIPLE OF THE RSM C ANCELLATION
and specific placement of both the target and the stationary
The task of decoupling the radar motion from the
reflector. Finally, in [36] and [37], the effect of the RSM
breath of a human target is accomplished by employing a
noise was removed with the help of an ultrasonic sensor.
frequency-modulated continuous-wave (FMCW) radar, whose
Various ego-motion estimation solutions were reported in the
operations are hereafter described. A voltage-controlled oscil-
automotive field. Even though the main concept is similar to
lator (VCO) driven by a saw-tooth waveform generates the
this work, both the magnitude of the considered motions and
signal that is amplified and then transmitted by means of the
the final purpose are different. Relevant examples are reported
antenna. A small portion of the transmitted signal is routed
in [38]–[41], where the main goal was locating vehicles during
to the receiving stage for signal demodulation. Characteris-
the covered route. This is often a matter of detecting ranges
tic parameters of the transmitted signal are the modulation
and large velocities instead of accurately measuring millime-
bandwidth, the duration time τchirp , and the repetition time
ter shifts, which can be considered large when compared
τrep of the transmitted signal. The sequence of modulated
with the tiny physiological displacements. Some technical
time-limited signals, i.e., chirps, requires an appropriate mod-
concerns still exist; while some articles involve relatively
ulating bandwidth to achieve a suitable spatial resolution. The
high computational costs and user intervention [31], [32],
relationship between resolution δ R and bandwidth B is
the most common limitation is the requirement of additional
sensors [28]–[30], [33]–[37]. c
δR = (1)
This work aims to develop a solution that allows efficient 2B
suppression of the RSM for measuring the breathing rate with- where c is the speed of light. As a consequence, if the
out additional sensors. The main idea consists of inferring the spatial resolution is not enough for resolving two different
RSM from echo signatures of stationary objects/clutters. While objects, the bandwidth should be increased at the expense of
RSM and physiological signals are mixed in human-associated higher computational cost, which is also limited by spectrum
echoes, they are rather isolated in the echoes from stationary allocation regulations. The duration time represents the period
objects around the human subject. Analyzing phase histories of the chirp, whereas the repetition time is the time between
provides precious information concerning the radar motion the beginning of a chirp and the next one. The former affects
that can be in turn isolated from the human echo. It is worth the beat frequency f b , i.e., the frequency of the signal at the
noting that the term “clutter” will be used as a synonym of output of the quadrature mixer, which is related to the range
stationary objects around the primary target of interest, i.e., the R by means of [44]
human. cτchirp f b
To automatically identify the human target and clutters, two R= . (2)
procedures are proposed: the former based on the autocorrela- B 2
tion applied to the received phase histories and the latter based The latter concerns the so-called Doppler dilemma: the
on the cross correlation applied to the measured range-Doppler longer the repetition time, the greater the unambiguous range
profiles. The proposed approaches will first be validated by but the smaller the detectable speed and vice versa.
simulating small and large RSM and the effect of different Moreover, longer repetition times allow obtaining shorter
clutter radar cross sections (RCSs). Then, the simulated cases duty cycles, thus maintaining the transmitter switched OFF
will be experimentally reproduced for testing and validation. longer and thus reducing the energy consumption.
To the best of the authors’ knowledge, this is the first work A graphical representation of the transmitted signal and
where the radar motion is canceled relying only on the the main characteristic parameters are reported in Fig. 1. The
measurement of a single radar. The absence of additional transmitted signal stx (t) can be expressed as [45]
 
sensors is an important feature for obtaining a compact and j 2π f c t+π τ B
t 2 +ρ0
lightweight system, particularly for future applications where stx (t) = σT e chirp (3)
vital sign detectors would be integrated into mobile devices where f c is the central operating frequency, σT is the amplitude
such as smartphones [42], [43]. of the transmitted signal, and ρ0 is the initial phase. If an
This article is organized as follows. The operating principle object reflects the transmitted signal, it will be received by
and the simulations have been reported in Sections II and III, the antenna and amplified. The received signal srx (t) can be
respectively. Section IV shows the experimental setup and expressed as
analyzes the results. Finally, the conclusions are drawn in
srx (t) = σ R e j (2π fc (t−t)+π τ (t−t) +ρ 0 )
B 2
Section V. (4)

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY ROURKELA. Downloaded on September 27,2021 at 09:19:09 UTC from IEEE Xplore. Restrictions apply.
 1934 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 69, NO. 3, MARCH 2021

where σ R is the amplitude of the amplified signal and t is

the round-trip time proportional to the object range
2(R + x(t))
t = (5)
c
where x(t) is the range shift due to the relative radial motion
between the radar and the object. If the target is a human and
the radar is not still, x(t) will include the superposition of two
terms: the former accounting for the radar motion and the latter
accounting for the displacement of the human chest. The signal
demodulation is achieved by means of a homodyne receiver
with a quadrature architecture that is beneficial for avoiding
the onset of null point issues in continuous-wave (CW) radars.
Indeed, by exploiting two 90◦ out-of-phase channels, there is
always one channel that is not at the null point [46], making
this architecture very attractive for the task of vital signs
measurement [47]. Although this might not be beneficial for
null point issues in FMCW radar, the radar board employed
Fig. 2. System block diagram.
in Section IV presents the same architecture, and thus, it has
been maintained for this work. The beat signal sb (t) at the
be a challenging task due to the signal degradation as an
output of the in-phase and quadrature (I /Q) demodulator is
effect of the RSM. Two different methods are identified to
reported in the following equation:
  distinguish between stationary and human targets based on the
j 2π τ B
tt+2π f c t+θ1 autocorrelation applied to the received phase histories and on
sb (t) = σ e chirp (6)
the cross correlation between the measured range-Doppler pro-
where files, respectively. The former exploits the inherent periodicity
B of the breathing signal with the aim to highlight the breathing
θ1 = π t 2 (7)
τchirp rate ω by calculating the phase histories autocorrelation for
each measured range bin.
is a residual phase contribution that usually can be neglected in
Let n be a single range bin and N be the maximum number
short-range applications, i.e., when the square of the round-trip
of available range bins; RϕAn ,ϕn represents the autocorrelation
time t 2  tt [48], [49].
of the phase history at the range bin n, and then, Rϕn ,ϕn can be
The first term in the argument of the exponential is usually
calculated for all the N range bins as a function of the delay τ
exploited for extracting the target range. Indeed, since both
the radar motion and the breathing displacement are negligible  
RϕAn ,ϕn = E ϕn (t)ϕn∗ (t − τ ) (9)
compared to the target range, t is only attributed to R. The
second term represents the phase ϕ(t) of the received signal. where −τmax < τ < τmax , with τmax the measuring time and
The measurement of the phase history through multiple chirps E{} the expected value operator.
is very advantageous for achieving high-precision motion When the target is a breathing person, RϕAn ,ϕn is a periodic
measurements with the only drawback of possible phase function whose period is 2π/ω. On the other hand, when the
ambiguities if the phase step between two consecutive chirps target is a stationary clutter whose phase history only accounts
exceeds 2π. The phase history has been extracted by applying for the radar motion, RϕAn ,ϕn is a function with a maximum in
chirp-by-chirp the arctangent demodulation on the complex τ = 0, thus clearly distinguishable from the autocorrelation
range profile. The ambiguity due to the phase discontinuity, of the person. In the present case, the phase measured at the
i.e., the phase is limited within the interval (−π, π), whereas person range is not a pure periodic signal since it is affected
the chest movement is not, is solved by means of a phase by the radar motion.
unwrapping process. The relationship between x(t) and ϕ(t) is Therefore, the breathing signal can be effectively recognized
ϕ(t) when the radar motion is not so large to hide the breath-
x(t) = c. (8) ing. The second method allows to automatically identify the
4π f c
suitable clutters positions even for large radar movements,
For the case of a moving radar platform, the phase of the by applying the cross correlation to the range bins of the
human echo ϕ(t) is composed of the mixed contribution of the range-Doppler profile. In detail, let ψn and ψm be the Doppler
radar phase shift ϕ R (t) and the human breathing displacement components at the range bins n and m, respectively ; RψXn ,ψm
ϕ B (t). On the other hand, the phase of the clutter echo represents the cross correlation between ψn and ψm , and then,
includes only the term ϕ R (t). By subtracting ϕ R (t) from RψXn ,ψm can be calculated for all the N range bins as a function
ϕ(t), it will be possible to extract ϕ B (t). A block diagram of the Doppler shift γ
of the proposed radar system is shown in Fig. 2. Measuring
 
ϕ B (t) requires the proper and automatic identification of the RψXn ,ψm = E ψn (v)ψm∗ (v − γ ) (10)
human and stationary objects position that in real scenarios
that are unknown. As introduced in Section I, this might where v is the Doppler speed.

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY ROURKELA. Downloaded on September 27,2021 at 09:19:09 UTC from IEEE Xplore. Restrictions apply.
 CARDILLO et al.: VITAL SIGN DETECTION AND RSM CANCELLATION THROUGH CLUTTER IDENTIFICATION 1935

It is interesting to study two cases where the cross correla-

tion is applied between the Doppler components of: 1) station-
ary object and human and 2) two different stationary objects.
In case 1), the Doppler components of the stationary object
are due to the radar motion, whereas those of the human are
due to the combination of both the breathing displacement and
the radar motion. Therefore, the cross correlation will result in
a poor similarity grade. In case 2), the Doppler components
of two different stationary objects are both attributed to the
radar movements, thus resulting in high degree of similarity. Fig. 3. Simulated range profile for small radar movements.
The method based on the Doppler cross correlation is a
reliable solution to identify the clutter position required for
the next step of the breathing recovery. However, it requires
to compute the Doppler map and the cross correlation between
all the available range bins. When a low computational cost
is required, e.g., in real-time applications, the method based
on the phase autocorrelation might be preferred for moderate
radar movements. Another advantage of the phase autocor-
relation method is the capability to measure the period of
the movements. As an example, if the radar undergoes the
vibrations of an air conditioning vent, the vent vibration
Fig. 4. Demodulated phase histories for small radar movements of: person
period is usually known. Therefore, it might be exploited to (blue curve) and stationary objects at 5 m (black curve) and 7 m (yellow
selectively find and remove the corresponding range bins. The curve).
steps of the proposed method are as follows.
1) Select all possible targets from the range profile.
2) Implementing one of the proposed methods to distin-
guish person from clutters.
3) Removing the clutter demodulated phase from the
human one for extracting the vital sign information.
It is worth noting that the first step of targets selection is
implemented based on manual inspection of the range-profile
graph. To allow automatic target selection, a software-based
algorithm can be implemented. Since this contribution is
focused on the task of the RSM cancellation, this topic
is not further investigated. Nevertheless, different solutions
have been proposed in the literature for various applications,
including the examples reported in [50]–[52]. In Section III,
the methods’ effectiveness is simulated in different scenarios.
III. RSM S IMULATION
The RSM is tested by simulating a radar system with a
transmitted bandwidth sweeping from 24 to 24.2 GHz, i.e.,
200-MHz bandwidth, providing the range resolution of 75 cm.
The duration time and the repetition time are τchirp = 2 ms and
Fig. 5. Auto-correlation of (a) all the phase histories and (b) person (blue
τrep = 9 ms, respectively. The number of employed chirps Nc curve) and stationary object at 7 m (red curve).
depends on the measurement time τmeas , e.g., if τmeas = 30 s,
Nc can be calculated by τmeas /τrep = 3333 chirps. In Fig. 5(a), the autocorrelation highlights the breathing peri-
odicity at the person range, whereas this is not the case
A. Small Radar Movements for the stationary objects. In Fig. 5(b), the autocorrelation is
In the first case-of-study, the radar moves by following reported only at 2 and 7 m, allowing the stationary object
a random pattern whose maximum speed is 2.2 mm/s. The detection. Since RϕAn ,ϕn is a function of the time delay τ , all
presence of a person at the distance of 2 m and two stationary Fig. 5 concerning the autocorrelation method have the lag
targets at 5 and 7 m, respectively, is simulated. Thereafter, unit on the x-axis expressed in seconds (s). On the other
the phase history is extracted for each simulated range bin. hand, in Fig. 5 reporting the cross correlation RψXn ,ψm , the lag
Figs. 3 and 4 show the range profile and the phase histories unit depends on the Doppler shift γ and it is expressed in
at 2, 5, and 7 m, respectively. millimeter-per-seconds (mm/s). The time intervals considered
Since the nature of the targets is unknown, the phase for calculating the autocorrelations are the measured periods
autocorrelation method described in Section II is performed. for the demodulated phase history plots. As an example,

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY ROURKELA. Downloaded on September 27,2021 at 09:19:09 UTC from IEEE Xplore. Restrictions apply.
 1936 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 69, NO. 3, MARCH 2021

Fig. 6. Simulated range-Doppler map.

the time interval to compute the autocorrelation of Fig. 5 is the

period considered in Fig. 4. The second method exploits the
cross correlation between the Doppler components reported
in Fig. 6. The cross correlation between the person and the
other range bins is shown in Fig. 7(a), demonstrating a lack
of similarity. On the other hand, the cross correlation between
the stationary object at 7 m and the other range bins of
Fig. 7(b) shows a peak at 5 m-0 lag, i.e., high similarity.
Indeed, it is characterized by the same Doppler components
only accounting for the radar motion. This different behavior Fig. 7. Cross correlation between (a) person and other range bins and
(b) object at 7 m and other range bins.
allows identifying both person and clutter position.
For the case of small radar movements, both the proposed
methods allow recovering the breathing signal. However,
cross-correlating the Doppler components requires a higher
computational cost compared to the phase autocorrelation.
The fluctuations noticed around the main target range are
due to spectrum leakage and the limited range resolution. A
Blackman windowing is implemented on the received I /Q
data to limit this effect. When properly restricted within the
adjacent range bins, this effect can be considered beneficial
because it would allow the proper algorithm operation without
considering the exact target location. This outcome is expected Fig. 8. Simulated range profile for large radar movements.
because the phase history of the adjacent range bins is almost
the same.

B. Large Radar Movements

In the second case-of-study, the radar maximum speed is
set to 22 mm/s, whereas both human and clutters are the
same as in Section III-A. The corresponding range profile is
reported in Fig. 8. The key difference with the case-of-study
of Section III-A is that the spatial amplitudes of radar motion
are 20 times larger than the breathing displacement, as shown
in Fig. 9. In this case, the phase autocorrelation shows its Fig. 9. Demodulated phase-histories for large radar movements of: person
limitation since the radar motion hides the breathing activity, (blue curve) and stationary objects at 5 m (black curve) and 7 m (yellow
curve).
making it very arduous to recognize the breathing, as shown
in Fig. 10. On the other hand, the Doppler cross correlation 1) observing the energetic peaks on range profile, which
shows reliability. correspond to either person or clutter;
In Fig. 11, the different Doppler signatures between person 2) implementing the autocorrelation/cross-correlation
and clutters are already visible, whereas in Fig. 12, the cross methods to distinguish person from clutters;
correlations between person and object at 7 m and objects at 3) exploiting the information of point b to calculate the
7 and 5 m highlight a high similarity among clutters, allowing difference in the demodulated phase between person and
their easy differentiation. clutter bins for extracting the vital sign information.
From Figs. 5 and 7, it is straightforward to figure out that
C. Detected Vital Sign Information the first energetic strip in the range profile of Fig. 3 is the
In Sections III-A and III-B, the first two steps of the person echo, whereas the two energetic strips at 5 and 7 m
algorithm procedure have been fulfilled according to the are clutters. For the case of large radar movement, the same
following schedule: conclusion can be inferred from the cross-correlation results of

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY ROURKELA. Downloaded on September 27,2021 at 09:19:09 UTC from IEEE Xplore. Restrictions apply.
 CARDILLO et al.: VITAL SIGN DETECTION AND RSM CANCELLATION THROUGH CLUTTER IDENTIFICATION 1937

Fig. 10. Autocorrelation of person (blue curve) and stationary object at 7 m

(red curve). The two curves overlap.

Fig. 13. Demodulated raw phase-histories of person (blue curve) and

difference between person and stationary object at 7 m (red curve) for the
Fig. 11. Simulated range-Doppler map. cases of (a) small and (b) large radar movement.

Fig. 12. Cross correlation between person and object at 7 m (blue curve) Fig. 14. (a) Range spectrum for σ R equal to 2 (blue inverted triangles), 1
and objects at 7 m and 5 m (red curve). (red dashed line, 0.5 (cyan triangles), 0.25 (green squares), 0.05 (black solid
line), and 0.005 (yellow circles).
Fig. 12. The last step of the procedure has been implemented
for both the cases of Sections III-A and III-B. The results D. Influence of RCS
reported in Fig. 13(a) and (b) for the case of small and large In this section, the effect of different clutter RCSs on the
radar movements, respectively, showing the effectiveness of performance of the proposed solution is analyzed by arranging
the algorithm as, in both cases, the vital sign detection has the three targets two meters apart, with a clutter of variable
been improved to a very large extent. In both cases, the peak RCS placed in the middle, in detail:
detection to distinguish between stationary and breathing tar- 1) person at 2 m;
gets was based on manual inspection. Like the case of the 2) clutter with variable RCS at 4 m;
target selection, the readers can refer to [50]–[52] for further 3) clutter at 6 m.
details. The above analysis assumes that all the surrounding The impact of different RCSs is studied by assigning differ-
targets are stationary. In real applications, this might represent ent weights to the amplitude of the signal after the low-noise
a limitation, thus deserving special consideration. Two cases amplification stage σ R [reported in (4)]. It is worth noting that
may occur: 1) periodic and 2) nonperiodic motion of the the range windowing has not been applied to the range Fast
surrounding objects. The first case is challenging because the Fourier transform (FFT).
periodical motion may be confused with the vital signs and Otherwise, the spectrum sidelobes might be suppressed by
thus requires complex additional processing. Indeed, echoes the window itself, in turn masking the effect of different
from stationary targets may appear to the radar due to the RCS on the other frequency components. Fig. 14 shows the
physiological activity. The second case has the unique effect range spectrums for different RCSs. Thereafter, the normalized
of reducing the number of stationary targets/clutters available prominence factor P is introduced
for the RSM cancellation method, which is based on their  
similarity. A moving target will show a Doppler behavior
X
D Rhmn,obj1 ,R obj1,obj2
X

different from that of clutters and will be discarded in favor P=   (11)

1 ,R obj1,obj2
X X
of suitable stationary targets. M Rhmn,obj

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY ROURKELA. Downloaded on September 27,2021 at 09:19:09 UTC from IEEE Xplore. Restrictions apply.
 1938 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 69, NO. 3, MARCH 2021

TABLE I
S IMULATED N ORMALIZED RCS VALUES

   
X
where D Rhmn,obj1 , R obj1,obj2
X X
and M Rhmn,obj1 , R obj1,obj2
X
are
functions that return the distance and the middle point between
the peak value of human/object cross correlation at 4 m and the Fig. 15. Picture of the outdoor experimental setup.
cross correlation between objects at 6 and 4 m, respectively. P
quantifies the capability of the method in separating humans
from stationary targets. In Table I, the values of σ R are
normalized using the value of the human subject and reported
with the corresponding prominence values. Increasing the
RCS is beneficial since the cross correlation between the
two stationary objects increases, which improves the detection
capability. This result is not trivial, as it might be mistakenly
expected that a larger RCS would increase the spectrum
leakage due to the higher sidelobe levels, enhancing the
similarity between human and nonhuman responses and in
turn challenging the cross-correlation-based approach. In such Fig. 16. Measured range profile for small radar movements.
a case, the prevalent effect is the increased absolute value
of the clutter main peak that in turn increases the value of
the cross correlation between clutters at 0 lag. By decreasing
the RCS, although the amplitudes decrease, the clutters cross
correlation is always characterized by the main peak at 0
lag, thus allowing to discern between human and nonhuman
targets.
It is worth noting that by varying the RCS from 0.05 to
0.005, the cross correlation does not change, as can be
inferred from the almost constant value of the prominence
values shown in Table I. This might be due to the level of the Fig. 17. Measured phase-histories for small radar movements of: person
(blue curve) and stationary objects at 5.5 m (black curve) and 13 m (yellow
clutter spectrum at 4 m that is lower than the clutter sidelobes curve).
at 6 m.
By considering the order of magnitude of the ranges mea-
IV. E XPERIMENTAL VALIDATION
sured in this article, this effect is negligible for both range and
A 24-GHz radar board was employed to test the effec- Doppler measurements. Two different cases of studies have
tiveness of the proposed approach. It exploits an Infineon been reproduced for testing small and large radar motions.
BGT24MTR11 SiGe transceiver chipset [53] with both trans-
mitting and receiving antennas aboard. The row I /Q data are A. Small Radar Movements
sent to a computer through a UART interface for the signal The scenario is composed of a human target at 2.5 m,
processing steps. The main radar parameters were set to be a stationary target at 5.5 m, and a weak clutter at 13 m.
the same as the simulations, i.e., bandwidth sweeping from The corresponding range profile is shown in Fig. 16. Once
24 to 24.2 GHz, duration time, and repetition time equal to 2 the ranges of person and clutter are detected, it is possible
and 9 ms. The block diagram of the real system is the same to retrieve the real breathing activity by subtracting the con-
as in Fig. 2, whereas a picture of the experimental setup is tribution of the radar motion from the person phase history,
reported in Fig. 15. It is worth noting that the radar azimuth as shown in Fig. 17. As described in Section II, the breathing
beamwidth is 20◦ . As an example, by considering the radar measurement in case of RSM requires measuring the phase
moving on a line directed toward the main lobe, the worst case history at the range of the human and clutter. Since their ranges
occurs for a target lying on the main lobe line and a target are unknown, the phase autocorrelation method is applied for
at the edge of the radar field of view, i.e., ±10◦ . Therefore, their automatic distinction and detection. In Fig. 18, the object
the range difference can be computed by dividing the range at 13 m does not show the periodicity typical of the breathing
by cos 10◦ = 0.985. activity, thus allowing a proper target distinction. In this

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY ROURKELA. Downloaded on September 27,2021 at 09:19:09 UTC from IEEE Xplore. Restrictions apply.
 CARDILLO et al.: VITAL SIGN DETECTION AND RSM CANCELLATION THROUGH CLUTTER IDENTIFICATION 1939

Fig. 18. Autocorrelation of person (blue curve) and stationary object at 13 m Fig. 21. Cross correlation between person and object at 8.5 m (blue curve)
(red curve). and objects at 5 and 8.5 m (red curve).

Fig. 19. Measured phase-histories for large radar movements of: person (blue
curve) and stationary objects at 5 m (black curve) and 8.5 m (yellow curve).

Fig. 22. Measured phase histories of person (blue curve) and difference
between person and stationary object (red curve) for the cases of (a) small
and (b) large radar movement.

Fig. 20. Measured range-Doppler map. represents a valuable solution when a low computational cost
is required.
case, also, the cross-correlation method can be applied to the
range-Doppler map to identify the clutters’ ranges. C. Detected Vital Signs
As demonstrated in Section III-C, where the simulated
B. Large Radar Movements vital signs were properly extracted, the information provided
For the case of large radar motion, the environment includes by the autocorrelation/cross-correlation methods allows dis-
a person and two stationary objects at the distance of 2.5, tinguishing the person from clutters in a real environment.
5, and 8.5 m, respectively. The related phase histories have For the case of small radar motion, target identification is
been reported in Fig. 19, where the larger radar motion accomplished either by observing the periodicity typical of
completely hides the breathing rate. Therefore, the clutter the breathing physiological motion from the autocorrelation
detection through the phase autocorrelation is not feasible. reported in Fig. 18 or implementing the cross-correlation
On the other hand, the Doppler cross correlation allows to method. For the case of large radar motion, the peak of the
effectively recognize different Doppler signatures of the person cross correlation reported in Fig. 21 allows to separate the
and the clutters. The range-Doppler map and the outcome of clutters at 5 and 8.5 m from the person at 2.5 m.
the Doppler cross correlation are reported in Figs. 20 and 21, The difference signal is calculated for both the cases of
respectively. The experimental activity confirmed the prelimi- small and large radar movements, thus allowing to recognize
nary results obtained by means of the simulations. For small the human vital sign from the phase-unwrapped information
radar movements, both the solutions allow the effective iden- of the person echo and clutter echo, as reported in Fig. 22.
tification of suitable clutters for the next step of the breathing Moreover, by comparing the spectrum of the noisy raw signal
signal recovery. While the Doppler cross correlation can be belonging to the person echo with the extracted breathing
effectively applied to different scenarios at the expense of a signal in the frequency domain shown in Fig. 23, the advan-
higher computational cost, the phase autocorrelation method tageous features of the proposed solution are demonstrated.

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY ROURKELA. Downloaded on September 27,2021 at 09:19:09 UTC from IEEE Xplore. Restrictions apply.
 1940 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 69, NO. 3, MARCH 2021

Fig. 25. Measurements in an indoor scenario. (a) Range profile,

Fig. 23. FFT of the extracted vital sign (blue curve) and raw person phase (b), phase-histories of: person (blue curve), stationary objects at 4 m (black
history (red curve) for the cases of (a) small and (b) large radar movement. curve) and 6 m (yellow curve), and extracted vital sign (red curve) multiplied
by ten.

for better visualization. The main limitations of the proposed

solution might be summarized as follows.
1) It requires stationary targets/clutters for the RSM
cancellation.
2) Periodical movements of either the human body or the
radar may be confused with the vital sign.
References [26]–[30] and [33]–[37] are not affected by the
first limitation.
However, they require additional sensors, which may limit
realistic applications. In [31] and [32], the RSM cancellation
can be fulfilled if the motion patterns are well defined.
However, they involve a high computational cost and require
Fig. 24. Picture of the indoor experimental setup. manual selection of the intrinsic mode functions. In very
cluttered scenarios without periodic random motion of either
During the measurement, the breathings-per-minutes (bpm) moving objects or the radar platform, the proposed solution
of the subject under test have been monitored by manually may be preferred.
counting the chest displacements per unit time, obtaining the V. C ONCLUSION
values of 14 and 10 bpm for the case of small and large radar
In this article, a novel approach to remove the RSM
motion, respectively. These values, corresponding to 0.23 and
effects from the human vital sign is proposed. It employs
0.17 Hz, respectively, are in good agreement with the signals
the radar signal for both the breathing detection and the
reported in Fig. 23. Finally, a measurement was performed
RSM effect removal, without the need for any additional
for testing the method in a complex indoor environment
sensor. Moreover, two new techniques to distinguish clutters
and for reproducing higher RSM velocities. A sofa, chairs,
from humans are developed. The choice between the two
walls, and different additional objects were present within the
solutions depends on the tradeoff between the extent of the
scenario to reproduce a real house case-of-study, as shown
radar motion and the available computational resources. Both
in Fig. 24. Contrary to what could be thought, it turned
simulations and experimental tests have been performed to
out that more cluttered scenarios might be beneficial for
validate the approaches. This work might represent a new
the proposed RSM cancellation. Indeed, a cluttered scenario
solution for vital sign detection on a nonstationary platform,
may offer a larger number of suitable stationary targets,
in a compact, lightweight, comfortable, and cost-effective
which are necessary for the proposed method. In Fig. 25(a),
way.
the corresponding range profile is reported and, due to the
large extent of the RSM, the Doppler cross-correlation method R EFERENCES
is proposed. In Fig. 25(b), the demodulated phase histories [1] J. C. Lin, “Noninvasive microwave measurement of respiration,” Proc.
are reported, and the extracted vital sign is multiplied by 10 IEEE, vol. 63, no. 10, p. 1530, Oct. 1975.

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY ROURKELA. Downloaded on September 27,2021 at 09:19:09 UTC from IEEE Xplore. Restrictions apply.
 CARDILLO et al.: VITAL SIGN DETECTION AND RSM CANCELLATION THROUGH CLUTTER IDENTIFICATION 1941

[2] J. C. Lin, J. Kiernicki, M. Kiernicki, and P. B. Wollschlaeger, [23] J. Zhang, H. Lan, M. Liu, and Y. Yang, “A handheld nano through-wall
“Microwave apexcardiography,” IEEE Trans. Microw. Theory Techn., radar locating with the gain-enhanced vivaldi antenna,” IEEE Sensors
vol. 27, no. 6, pp. 618–620, Jun. 1979. J., vol. 20, no. 8, pp. 4420–4429, Apr. 2020.
[3] E. F. Greneker, “Radar sensing of heartbeat and respiration at a distance [24] E. Cardillo and A. Caddemi, “Insight on electronic travel aids for
with applications of the technology,” in Proc. Radar Syst. (RADAR), visually impaired people: A review on the electromagnetic technology,”
1997, pp. 150–154. Electron., vol. 8, pp. 1–12, Nov. 2019.
[4] E. Cardillo and A. Caddemi, “A review on biomedical MIMO radars for [25] E. Cardillo, C. Li, and A. Caddemi, “Empowering blind people mobility:
vital sign detection and human localization,” Electronics, vol. 9, no. 9, A millimeter-wave radar cane,” in Proc. IEEE Int. Workshop Metrology
p. 1497, Sep. 2020. Ind. 4.0 IoT, Rome, Italy, Jun. 2020, pp. 213–217.
[5] M. Li and J. Lin, “Wavelet-transform-based data-length-variation tech- [26] R. Nakata, S. Clemens, A. Lee, and V. Lubecke, “RF techniques for
nique for fast heart rate detection using 5.8-GHz CW Doppler radar,” motion compensation of an unmanned aerial vehicle for remote radar
IEEE Trans. Microw. Theory Techn., vol. 66, no. 1, pp. 568–576, life sensing,” in IEEE MTT-S Int. Microw. Symp. Dig., San Francisco,
Jan. 2018. CA, USA, May 2016, pp. 1–4.
[6] E. Yavari and O. Boric-Lubecke, “Channel imbalance effects and com- [27] R. H. Nakata, B. Haruna, T. Yamaguchi, V. M. Lubecke, S. Takayama,
pensation for Doppler radar physiological measurements,” IEEE Trans. and K. Takaba, “Motion compensation for an unmanned aerial vehicle
Microw. Theory Techn., vol. 63, no. 11, pp. 3834–3842, Nov. 2015. remote radar life sensor,” IEEE J. Emerg. Sel. Topics Circuits Syst.,
[7] C. Gu, Z. Peng, and C. Li, “High-precision motion detection using low- vol. 8, no. 2, pp. 329–337, Jun. 2018.
complexity Doppler radar with digital post-distortion technique,” IEEE [28] A. Rahman, E. Yavari, A. Singh, V. M. Lubecke, and O.-B. Lubecke,
Trans. Microw. Theory Techn., vol. 64, no. 3, pp. 961–971, Mar. 2016. “A low-IF tag-based motion compensation technique for mobile Doppler
[8] C. Ding et al., “Inattentive driving behavior detection based on portable radar life signs monitoring,” IEEE Trans. Microw. Theory Techn., vol. 63,
FMCW radar,” IEEE Trans. Microw. Theory Techn., vol. 67, no. 10, no. 10, pp. 3034–3041, Oct. 2015.
pp. 4031–4041, Oct. 2019. [29] A. Singh and V. Lubecke, “Adaptive noise cancellation for two frequency
[9] T.-Y. Huang, L. F. Hayward, and J. Lin, “Noninvasive measurement and radars using frequency doubling passive RF tags,” in IEEE MTT-S Int.
analysis of laboratory Rat’s cardiorespiratory movement,” IEEE Trans. Microw. Symp. Dig., Montreal, QC, Canada, Jun. 2012, pp. 1–3.
Microw. Theory Techn., vol. 65, no. 2, pp. 574–581, Feb. 2017. [30] I. Mostafanezhad, O. Boric-Lubecke, V. Lubecke, and A. Host-Madsen,
[10] J. Tu, T. Hwang, and J. Lin, “Respiration rate measurement under 1- “Cancellation of unwanted motion in a handheld Doppler radar used for
D body motion using single continuous-wave Doppler radar vital sign non-contact life sign monitoring,” in IEEE MTT-S Int. Microw. Symp.
detection system,” IEEE Trans. Microw. Theory Techn., vol. 64, no. 6, Dig., Atlanta, GA, USA, Jun. 2008, pp. 1171–1174.
pp. 1937–1946, Jun. 2016. [31] I. Mostafanezhad, O. Boric-Lubecke, V. Lubecke, and D. P. Mandic,
[11] C. Li and J. Lin, “Random body movement cancellation in Doppler “Application of empirical mode decomposition in removing fidgeting
radar vital sign detection,” IEEE Trans. Microw. Theory Techn., vol. 56, interference in Doppler radar life signs monitoring devices,” in Proc.
no. 12, pp. 3143–3152, Dec. 2008. Annu. Int. Conf. IEEE Eng. Med. Biol. Soc., Minneapolis, MN, USA,
[12] F. K. Wang, T. S. Horng, K. C. Peng, J. K. Jau, J. Y. Li, and C. C. Chen, Sep. 2009, pp. 340–343.
“Single-antenna Doppler radars using self and mutual injection locking
[32] I. Mostafanezhad, E. Yavari, O. Boric-Lubecke, V. M. Lubecke, and
for vital sign detection with random body movement cancellation,” D. P. Mandic, “Cancellation of unwanted Doppler radar sensor motion
IEEE Trans. Microw. Theory Techn., vol. 59, no. 12, pp. 3577–3586, using empirical mode decomposition,” IEEE Sensors J., vol. 13, no. 5,
Dec. 2011.
pp. 1897–1904, May 2013.
[13] C. Gu, G. Wang, Y. Li, T. Inoue, and C. Li, “A hybrid radar-camera
[33] A. Rahman, E. Yavari, X. Gao, V. Lubecke, and O. Boric-Lubecke,
sensing system with phase compensation for random body movement
“Signal processing techniques for vital sign monitoring using mobile
cancellation in Doppler vital sign detection,” IEEE Trans. Microw.
short range Doppler radar,” in Proc. IEEE Top. Conf. Biomed. Wire-
Theory Techn., vol. 61, no. 12, pp. 4678–4688, Dec. 2013.
less Technol., Netw., Sens. Syst. (BioWireleSS), San Diego, CA, USA,
[14] M.-C. Tang, C.-Y. Kuo, D.-C. Wun, F.-K. Wang, and T.-S. Horng, Jan. 2015, pp. 1–3.
“A self- and mutually injection-locked radar system for monitoring
vital signs in real time with random body movement cancellation,” [34] I. Mostafanezhad, B.-K. Park, O. Boric-Lubecke, V. Lubecke, and
IEEE Trans. Microw. Theory Techn., vol. 64, no. 12, pp. 4812–4822, A. Host-Madsen, “Sensor nodes for Doppler radar measurements of life
Dec. 2016. signs,” in IEEE MTT-S Int. Microw. Symp. Dig., Honolulu, HI, USA,
Jun. 2007, pp. 1241–1243.
[15] K. Han and S. Hong, “Differential phase Doppler radar with collocated
multiple receivers for noncontact vital signal detection,” IEEE Trans. [35] F. Zhu, K. Wang, and K. Wu, “A Fundamental-and-Harmonic dual-
Microw. Theory Techn., vol. 67, no. 3, pp. 1233–1243, Mar. 2019. frequency Doppler radar system for vital signs detection enabling
[16] T. Zhang, J. Sarrazin, G. Valerio, and D. Istrate, “Estimation of human radar movement self-cancellation,” IEEE Trans. Microw. Theory Techn.,
body vital signs based on 60 GHz Doppler radar using a bound- vol. 66, no. 11, pp. 5106–5118, Nov. 2018.
constrained optimization algorithm,” Sensors, vol. 18, no. 7, p. 2254, [36] A. Rahman, E. Yavari, X. Gao, J. Xu, V. Lubecke, and O.-B. Lubecke,
Jul. 2018. “See-through-wall life sensing using mobile Doppler radar,” in Proc.
[17] C. Zhang, M. Cao, Y. Gong, Y. Li, Y. Huang, and H. Wang, “Cali- 87th ARFTG Microw. Meas. Conf. (ARFTG), San Francisco, CA, USA,
bration of motional frequency spread for wide-band FMCW automo- May 2016, pp. 1–4.
tive millimeter-wave radar,” IEEE Access, vol. 8, pp. 14355–14366, [37] E. F. Greneker, III, and D. J. Zywicki, “Stabilizing motion in a radar
Jan. 2020. detection system using ultrasonic radar range information,” U.S. Patent
[18] S. Dias Da Cruz, H.-P. Beise, U. Schroder, and U. Karahasanovic, 6 909 397 B1, Jun. 21, 2005.
“A theoretical investigation of the detection of vital signs in presence of [38] J. Dickmann et al., “Automotive radar the key technology for
car vibrations and RADAR-based passenger classification,” IEEE Trans. autonomous driving: From detection and ranging to environmental
Veh. Technol., vol. 68, no. 4, pp. 3374–3385, Apr. 2019. understanding,” in Proc. IEEE Radar Conf. (RadarConf), Philadelphia,
[19] S. Agarwal and D. Singh, “An adaptive statistical approach for non- PA, USA, May 2016, pp. 1–6.
destructive underline crack detection of ceramic tiles using millimeter [39] D. Kellner, M. Barjenbruch, J. Klappstein, J. Dickmann, and
wave imaging radar for industrial application,” IEEE Sensors J., vol. 15, K. Dietmayer, “Instantaneous ego-motion estimation using multiple
no. 12, pp. 7036–7044, Dec. 2015. Doppler radars,” in Proc. IEEE Int. Conf. Robot. Autom. (ICRA), Hong
[20] E. Cardillo and A. Caddemi, “Feasibility study to preserve the health of Kong, May 2014, pp. 1592–1597.
an industry 4.0 worker: A radar system for monitoring the sitting-time,” [40] M. Barjenbruch, D. Kellner, J. Klappstein, J. Dickmann, and
in Proc. 2nd Workshop Metrol. Ind. 4.0 IoT (MetroInd4.0&IoT), Naples, K. Dietmayer, “Joint spatial- and Doppler-based ego-motion estimation
Italy, Jun. 2019, pp. 254–258. for automotive radars,” in Proc. IEEE Intell. Vehicles Symp. (IV), Seoul,
[21] S. Abdulatif, Q. Wei, F. Aziz, B. Kleiner, and U. Schneider, “Micro- South Korea, Jun. 2015, pp. 839–844.
Doppler based human-robot classification using ensemble and deep [41] S. H. Cen and P. Newman, “Precise ego-motion estimation with
learning approaches,” in Proc. IEEE Radar Conf. (RadarConf18), millimeter-wave radar under diverse and challenging conditions,” in
Oklahoma City, OK, USA, Apr. 2018, pp. 1043–1048. Proc. IEEE Int. Conf. Robot. Autom. (ICRA), Brisbane, QLD, Australia,
[22] M. Ulrich and B. Yang, “Short-duration Doppler spectrogram for person May 2018, pp. 6045–6052.
recognition with a handheld radar,” in Proc. 26th Eur. Signal Process. [42] J. Lien et al., “Soli: Ubiquitous gesture sensing with millimeter wave
Conf. (EUSIPCO), Rome, Italy, Sep. 2018, pp. 1227–1231. radar,” ACM Trans. Graph., vol. 35, no. 4, pp. 1–9, Jul. 2016.

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY ROURKELA. Downloaded on September 27,2021 at 09:19:09 UTC from IEEE Xplore. Restrictions apply.
 1942 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 69, NO. 3, MARCH 2021

[43] T. Fan et al., “Wireless hand gesture recognition based on continuous- Changzhi Li (Senior Member, IEEE) received
wave Doppler radar sensors,” IEEE Trans. Microw. Theory Techn., the B.S. degree in electrical engineering from
vol. 64, no. 11, pp. 4012–4020, Nov. 2016. Zhejiang University, Hangzhou, China, in 2004,
[44] E. Cardillo and A. Caddemi, “A novel approach for crosstalk mini- and the Ph.D. degree in electrical engineering from
mization in FMCW radars,” Electron. Lett., vol. 53, pp. 1379–1381, the University of Florida, Gainesville, FL, USA,
Sep. 2017. in 2009.
[45] Y. Xiong, Z. Peng, G. Xing, W. Zhang, and G. Meng, “Accurate and He is currently a Professor with Texas Tech
robust displacement measurement for FMCW radar vibration monitor- University, Lubbock, TX, USA. His research
ing,” IEEE Sensors J., vol. 18, no. 3, pp. 1131–1139, Feb. 2018. interests include biomedical applications of
[46] C. Li, V. M. Lubecke, O. Boric-Lubecke, and J. Lin, “A review on recent microwave technology, wireless sensors, and
advances in Doppler radar sensors for noncontact healthcare monitor- RF/analog circuits.
ing,” IEEE Trans. Microw. Theory Techn., vol. 61, no. 5, pp. 2046–2060, Dr. Li was a recipient of the IEEE Microwave Theory and Techniques
May 2013. Society (MTT-S) Outstanding Young Engineer Award, the IEEE Sensors
[47] G. Sacco, E. Piuzzi, E. Pittella, and S. Pisa, “An FMCW radar for Council Early Career Technical Achievement Award, the ASEE Frederick
localization and vital signs measurement for different chest orientations,” Emmons Terman Award, the IEEE-HKN Outstanding Young Professional
Sensors, vol. 20, no. 12, p. 14, Jun. 2020. Award, the NSF Faculty Early CAREER Award, and the IEEE MTT-S
[48] W. G. Carrara, R. S. Goodman, and R. M. Majewski, Spotlight Synthetic Graduate Fellowship Award. He is an Associate Editor of the IEEE
Aperture Radar: Signal Processing Algorithms. Boston, MA, USA: T RANSACTIONS ON M ICROWAVE T HEORY AND T ECHNIQUES , and the
Artech House, 1995. IEEE J OURNAL OF E LECTROMAGNETICS , RF AND M ICROWAVES IN
[49] E. Cardillo and A. Caddemi, “Radar range-breathing separation for the M EDICINE AND B IOLOGY. He served as an Associate Editor of the IEEE
automatic detection of humans in cluttered environments,” IEEE Sensors T RANSACTIONS ON C IRCUITS AND S YSTEMS I AND II from 2016 to
J., early access, Sep. 21, 2020, doi: 10.1109/JSEN.2020.3024961. 2019 and 2014 to 2015, respectively. He served as a TPC Co-Chair for the
[50] X. Wei, Y. Yang, J. Urena, J. Yan, and H. Wang, “An adaptive peak IEEE MTT-S International Microwave Biomedical Conference from 2018 to
detection method for inspection of breakages in long rails by using 2019, and the IEEE Wireless and Microwave Technology Conference from
barker coded UGW,” IEEE Access, vol. 8, pp. 48529–48542, Mar. 2020. 2012 to 2013.
[51] J.-Y. Kim, J.-H. Park, S.-Y. Jang, and J.-R. Yang, “Peak detection
algorithm for vital sign detection using Doppler radar sensors,” Sensors,
vol. 19, no. 7, p. 1575, Apr. 2019.
[52] D. Oh, J. S. Strattan, J. K. Hur, J. Bento, A. E. Urban, G. Song,
and J. M. Cherry, “CNN-Peaks: ChIP-seq peak detection pipeline using
Alina Caddemi (Member, IEEE) received the
convolutional neural networks that imitate human visual inspection,” Sci.
degree (Hons.) in electronic engineering and the
Rep., vol. 10, Dec. 2020, Art. no. 7933.
Ph.D. degree from the University of Palermo,
[53] Infineon BGT24MTR11 Datasheet, Rev. 3.1, 2014-03-25, Infineon
Palermo, Italy, in 1982 and 1987, respectively.
Technol. AG, Munich, Germany, 2014.
She joined the Electrical Engineering Department,
University of Utah, Salt Lake City, UT, USA,
in 1984, and the Electrical and Computer Engineer-
Emanuele Cardillo (Member, IEEE) received the ing Department, University of Colorado, Boulder,
M.Sc. degree in electronic engineering from the CO, USA, in 1985, as a Visiting Researcher in
University of Messina, Messina, Italy, in 2013 and the field of microwave bioelectromagnetics. From
the Ph.D. degree from the University Mediterranea 1990 to 1998, she was with the Department of
of Reggio Calabria, Reggio Calabria, Italy, in 2018. Electrical Engineering, University of Palermo, as an Assistant Professor.
He is currently a Research Fellow with the Uni- In 1998, she joined the University of Messina, Messina, Italy, where she is
versity of Messina. His current research interests currently a Full Professor of electronics and the Supervisor of the Microwave
are focused on the microwave electronics field, Electronics (ELEMIC) Lab. Her current research interests are in the fields
mainly on compact radar systems for short-range of temperature-dependent linear and noise characterization techniques for
applications, design of active and passive planar solid-state devices, cryogenic measurements and modeling of field-effect
hybrid microwave integrated circuits (HMIC), linear transistors, noise modeling of bipolar and field-effect transistors, neural
and noise modeling of microwave field-effect transistors, linear and noise network modeling of field-effect transistors, design and realization of hybrid
microwave measurements, and realization of HMIC circuits and systems. low-noise components and circuits, and compact radar systems for short-
Dr. Cardillo was served as a member for the organizing committee of the range applications. She has authored or coauthored more than 250 papers
International Workshop on Integrated Nonlinear Microwave and Millimeter- on international scientific journals and conference proceedings. She has been
wave circuits (INMMIC 2015). He serves as a member for the Scientific the partner team leader for several national and international projects.
Committee of the International Conference on Microelectronic, Devices and Dr. Caddemi has served as an Associate Editor for the International Journal
Technologies (MicDAT). He was the recipient of the “IEEE S ENSORS of Numerical Modeling: Electronic Networks, Devices and Fields and in
J OURNAL Best Student Paper Award” (2017/2018) and the “IEEE Microwave the Editorial Board of Microwave Review, a publication of the National
Theory and Techniques (MTT-S) award” (2018). He is the Chair of the Society for Microwave Technique, Technologies and Systems for Serbia
Technical Committee “Microwave and Millimeter-Wave Radar Sensors” of and Montenegro IEEE MTT-S Chapter. She is presently an Editorial Board
the IEEE Sensors Council Italy Chapter. Member of Electronics (MDPI).

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY ROURKELA. Downloaded on September 27,2021 at 09:19:09 UTC from IEEE Xplore. Restrictions apply.