Radar-Based Non-Contact Continuous Identity Authentication - MDPI
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
remote sensing
Review
Radar-Based Non-Contact Continuous
Identity Authentication
Shekh Md Mahmudul Islam * , Olga Borić-Lubecke, Yao Zheng and Victor M. Lubecke
Department of Electrical Engineering, University of Hawai‘i at Mānoa, Honolulu, HI 96822, USA;
olgabl@hawaii.edu (O.B.-L.); yaozheng@hawaii.edu (Y.Z.); lubecke@hawaii.edu (V.M.L.)
* Correspondence: shekh@hawaii.edu
Received: 11 June 2020; Accepted: 11 July 2020; Published: 15 July 2020
Abstract: Non-contact vital signs monitoring using microwave Doppler radar has shown great
promise in healthcare applications. Recently, this unobtrusive form of physiological sensing has also
been gaining attention for its potential for continuous identity authentication, which can reduce the
vulnerability of traditional one-pass validation authentication systems. Physiological Doppler radar
is an attractive approach for continuous identity authentication as it requires neither contact nor
line-of-sight and does not give rise to privacy concerns associated with video imaging. This paper
presents a review of recent advances in radar-based identity authentication systems. It includes
an evaluation of the applicability of different research efforts in authentication using respiratory
patterns and heart-based dynamics. It also identifies aspects of future research required to address
remaining challenges in applying unobtrusive respiration-based or heart-based identity authentication
to practical systems. With the advancement of machine learning and artificial intelligence, radar-based
continuous authentication can grow to serve a wide range of valuable functions in society.
Keywords: Doppler radar; non-contact measurement; respiration; heartbeat; sensor; authentication
1. Introduction
Doppler radar has been used in widespread applications, including weather forecasting, vehicle
speed measurement, structural health monitoring, and the monitoring of air and sea traffic [1].
This technology has most recently been recognized for promise in healthcare applications though long
term unobtrusive physiological monitoring [2–6]. The fundamental Doppler principle is illustrated in
Figure 1a, where a reflected signal undergoes a phase shift due to the subtle movement of the chest
surface caused by heartbeat and respiration [4–8]. Doppler radar remote life sensing of humans has been
widely reported, with proof of concepts demonstrated for various applications [7,8]. This non-contact
and non-invasive form of measurement has several potential advantages in medicine, especially
for the monitoring of neonates or infants at risk of sudden infant death syndrome [9], adults with
sleep disorders [10], and burn victims [11,12]. In addition, separation of respiratory signatures in
a multi-subject environment has also been investigated [13–16]. Moreover, this form of respiration
monitoring reduces patient discomfort and distress as electrodes need not be attached to the body.
The inherent advantage of this unobtrusive non-contact measurement technique broadens potential
applications beyond healthcare to include occupancy sensing [17] and related energy management in
smart homes [18,19] and baby monitoring [20].
Remote Sens. 2020, 12, 2279; doi:10.3390/rs12142279 www.mdpi.com/journal/remotesensing
Remote Sens. 2020, 12, 2279 2 of 22
Remote Sens. 2020, 12, x FOR PEER REVIEW 2 of 21
(a) (b)
Figure
Figure 1.
1. Basic
Basic principle
principle of
of Doppler
Doppler radar
radar physiological
physiological sensing
sensing (a),
(a), and
and non-contact
non-contact continuous
continuous
identity
identity authentication concept (b). A radar system typically consists of a transmitterand
authentication concept (b). A radar system typically consists of a transmitter andaareceiver.
receiver.
When
When aa transmitted
transmitted signal
signal of frequency ω
of frequency ω is
is reflected
reflected its
its phase changes, φφ(t),
phase changes, (t),in
indirect
direct proportion
proportion to
to
the
the subtle
subtle motion.
motion.
Growing interest
Growing interestininphysiological
physiologicalmotion motionsensing
sensingthrough
through radar
radar has has
ledled to to
thethe development
development of
of new front-end architectures [20,21], baseband signal processing
new front-end architectures [20,21], baseband signal processing methods [22], and system-level methods [22], and system-level
integration [21,23]
integration [21,23] to to improve
improve detection
detection accuracy
accuracy and and robustness
robustness [24].[24]. A A review
review of of recent
recent advances
advances
in Doppler
in Doppler radarradar sensors
sensors has has been
been reported
reported by by Li
Li et
et al.
al. [25].
[25]. One
One example
example is is the
the application
application of of this
this
non-invasive technology to monitor infants
non-invasive infants for forsudden
suddeninfant
infantdeath
deathsyndrome
syndrome(SIDS) (SIDS)[26,27]
[26,27]which
which is
one
is oneofofthetheleading
leadingcauses
causesofofinfant
infantmortality.
mortality.Moreover,
Moreover, Doppler
Doppler radar
radar has also been implementedimplemented
to monitor
to monitor the the health
health and and behavior
behavior of of terrestrial
terrestrial and aquatic animals [28–30]. Sleep Sleep monitoring
monitoring is is
another
anotheremerging
emergingapplication
applicationwhere where radar
radar alleviates
alleviatesthethe
measurement
measurement interference
interference introduced
introduced by theby
conventional
the conventional use ofuse obtrusive
of obtrusivedevices suchsuch
devices torsotorso
straps and or
straps andspirometers
or spirometers [31]. [31].
A clinical studystudy
A clinical was
performed
was performed to comparatively
to comparatively monitor patients
monitor suffering
patients sufferingfromfrom
sleepsleep
apnea apneausing a radar
using a radarsensor
sensorin
conjunction
in conjunction with traditional
with traditional intrusive
intrusive sleep
sleepmonitoring
monitoring equipment,
equipment, wherewhere the theradar
radarwaswas foundfoundto
provide
to provide stand-alone
stand-alone detection
detection of ofmost
most apnea
apnea events,
events,asaswellwellasascomplementary
complementarydetail detailto to facilitate
facilitate
conventional
conventional diagnostics
diagnostics [31]. [31]. Furthermore,
Furthermore,Food Foodand andDrug
Drug Administration
Administration(FDA) (FDA) approval
approval for for the
the
first
first commercial
commercial use use ofof wireless,
wireless, non-contact
non-contact respiratory
respiratory devices
devices inin the
the Unites
Unites States
States was
was granted
granted in in
2009
2009 [32].
[32].
Beyond
Beyond healthhealth sensing,
sensing, Doppler
Dopplerradar radaralsoalsoholds
holdsgreat
great promise
promise to to enhance
enhance systemsystem security
security and and
privacy,
privacy, particularly
particularly in in the
the area of user authentication as illustrated in Figure 1b. Existing Existing system
system
authentication
authentication methodsmethods predominately
predominately employ employaaone-off,
one-off,interruptive
interruptiveapproach,
approach,which whichauthenticates
authenticates
only
only atatthe
theinitial
initiallog-in
log-inof ofaasession
session[33–36],
[33–36],withwithusers
usersactively
activelyengaging
engagingan aninput
inputdevice
deviceor orbiometric
biometric
reader.
reader. Such
Suchdesigns
designsare arevulnerable
vulnerableto toopen
opensession
sessionexploitations
exploitationsand andmay mayinterfere
interferewith withuseruseractivity.
activity.
There
There have
have beenbeen manymany studies
studies focused
focused on on continuity
continuity in user authentication.
authentication. Several Several havehave explored
explored
sensing
sensing technologies
technologies to toacquire
acquirecommoncommonphysiological
physiological traits,
traits, including
including fingerprint
fingerprint [37],[37],
palm palm
printprint
[38],
[38],
and and iris [39],
iris [39], used used to monitor
to monitor and and authenticate
authenticate usersthroughout
users throughoutaasession.session. Recent advancements
advancements
in
in wearable
wearable sensors
sensors and and pattern
pattern recognition
recognition further
further enable
enable system
system architects
architects to to collect
collect more
more subtle
subtle
physiological
physiological patterns,
patterns,such suchas asthose
thoseassociated
associatedwith withelectroencephalogram
electroencephalogram(EEG) (EEG)[40], [40],finger-vein
finger-vein[41],[41],
and
and gait
gait[42],
[42],to toverify
verifyusers
usersimplicitly
implicitlyand andcontinuously.
continuously.ComparedComparedto tothese
thesecontact-based
contact-basedsolutions,
solutions,
continuous authenticationusing
continuous authentication usingnon-contact,
non-contact, unobtrusive
unobtrusive techniques,
techniques, suchsuch as Doppler
as Doppler radar, radar, can
can further
further
improve improve
system system
usability usability
and expand and expand
the range theofrange of applications
applications into domainsinto domains
with known with privacy
known
privacy
concernsconcerns
[43,44]. [43,44]. For example,
For example, variousvarious
visible andvisible and thermal-based
thermal-based camerascamerasare employedare employed
to acquire to
acquire
face andface gaitand gait features
features for userfor user verification
verification [45–47]. [45–47].
However, However, image-based
image-based approaches
approaches suffersuffer
from
from
severalseveral irreconcilable
irreconcilable dilemmas,
dilemmas, including
including a lackaof lack of privacy
privacy and degraded
and degraded performance
performance underundera low
alight
lowambient
light ambient
conditions conditions [48,49]. Alternatively,
[48,49]. Alternatively, a solution unobtrusive
a solution leveraging leveraging radar unobtrusive
measurementradar
measurement
of cardiopulmonary of cardiopulmonary
motion can bemotion immune cantobesuch
immune to such and
deficiencies deficiencies
achieve and achieveand
consistent consistent
reliable
and reliable under
recognition recognition under privacy-sensitive
privacy-sensitive conditions [49–54].
conditions [49–54].
This paper reviews recently reported research efforts in identity authentication based on the use
of dedicated microwave Doppler radar or WiFi devices in a bi-static radar configuration for
recognition of cardiopulmonary signatures, heart activity patterns, and respiratory features. This
paper also critically evaluates the practical adoption of the proposed solutions and discusses future
requirements for solving some of the remaining challenges needed for practical application.
Remote Sens. 2020, 12, 2279 3 of 22
This paper reviews recently reported research efforts in identity authentication based on the use of
dedicated microwave Doppler radar or WiFi devices in a bi-static radar configuration for recognition
of cardiopulmonary signatures, heart activity patterns, and respiratory features. This paper also
critically evaluates the practical adoption of the proposed solutions and discusses future requirements
for solving some of the remaining challenges needed for practical application.
The remainder of this paper is organized as follows: Section 2 describes cardiopulmonary diversity
and physiological motion measurement. Section 3 describes research publications on the identification
of people from dedicated radar and WiFi-based bi-static radar measured cardiopulmonary patterns
and evaluates the applicability and limitations. Section 4 provides concluding remarks.
2. Cardiopulmonary Diversity and Physiological Motion Measurement
Doppler radar detects motion that occurs due to physiological events, including heartbeat, arterial
pulsation, and breathing [55]. This physiological motion is concentrated in the thorax, where the heart
and lung lie, but also includes the abdomen, which moves with respiration, and superficial pulses,
which are present at any points in the body [55–58]. Torso deformation, due to respiratory effort,
is a complex, three-dimensional pattern and it varies greatly with subject parameters and activity
context [59,60]. Respiratory effort motion can come from one of the two primary regions: chest or
abdomen, known as thoracic and diaphragmatic breathing, respectively [61]. When the heart contracts
to generate the pressure that drives blood flow, it moves within the chest cavity, hitting the chest
wall, and creating a measurable displacement at the skin surface that can be detected with Doppler
radar [54–57]. Non-contact Doppler radar can operate at frequencies where primarily skin-surface
motion is detected [54]. Changes in the volume and shape of the heart during systole move the ribs
and soft tissue near the heart, causing the chest to pulse with each heartbeat [54–57]. Thus, Doppler
radar systems can sense respiratory-related information as well as cardiac patterns.
It has been demonstrated in various clinical investigations that sedentary adult human subjects
exhibit a diversity in respiratory pattern while awake, not only in terms of tidal volume and inspiratory
and expiratory duration, but also in terms of air flow profile [52,55]. Every individual selects to exhibit
one pattern among the infinite number of possible ventilatory variables and air flow profiles [55–57].
These variabilities are non-random and may be explained by either central neural mechanisms or
chemical feedback loops [54,55]. In addition, each individual has a different physical size and shape of
lungs, as well as different rib cage and abdominal muscle strength that contributes to the variations in
breathing patterns [56].
In addition to variations in respiratory features, there is also significant variation in heart-based
geometry [54–57]. In general, the human heart contains two upper cavities (atria) and two bottom
chambers (ventricles) [62,63]. The successive contraction (systole) and relaxation (diastole) of both
atria and ventricles circulate the oxygen-rich blood throughout the whole human body [62,63].
A diagrammatic section of the heart is shown in Figure 2a. The heart drives blood through the lungs
and to tissues throughout the body [54]. Cardiac motion consists of contraction and relaxation of both
atria and ventricles [58]. In one cardiac cycle, ventricles relax and passively fill with blood to 70% of
the total volume from atria through the open mitral valve [58]. At the same time, the atria contract
with heart muscles and pump blood. Figure 2b shows the motion of the heart through the phases of
the cardiac cycle [57]. The cardiac motion cycle consists of five distinct stages including: (1) ventricular
filling (VF), (2) atrial systole (AS), (3) isovolumetric ventricular contraction (IC), (4) ventricular ejection
(VE), and (5) isovolumetric ventricular relaxation (IR) [62,63]. These cycles are significantly unique
because of the different volumes, surface shape, and moving dynamics (speed, acceleration, etc.),
and also deformation of the heart [58,62,63]. These stages or cycles are different for each person
because of variations in size, position, anatomy of the heart and chest configuration, and various other
factors [63,64]. It has also been demonstrated in various clinical investigations that no two persons
have the same cardiac blood circulation [62,63].
Remote Sens. 2020, 12, 2279 4 of 22
Remote Sens. 2020, 12, x FOR PEER REVIEW 4 of 21
Figure
Figure 2. Diagrammaticsection
2. Diagrammatic sectionof
of the
the heart,
heart,(a),
(a),[55]
[55]and
andmotion
motionof the heart
of the through
heart the phases
through of the of
the phases
cardiac cycle, (b) [56]. The arrows indicate direction of blood flow.
the cardiac cycle, (b) [56]. The arrows indicate direction of blood flow.
Thus, cardiopulmonary motion is a unique identity marker for everyone, due to differences in
Thus, cardiopulmonary
physical motion
organ size and shape, andismuscle
a unique identity
strength [62].marker for everyone,
In addition, due to differences
cardiopulmonary motion is in
physical organ
controlled bysize andneural
central shape, and muscle
mechanisms strength feedback,
or chemical [62]. In addition, cardiopulmonary
which makes motion
it extremely difficult to is
controlled by central
counterfeit or hide neural
a livingmechanisms or chemical
individual’s motion feedback,
signature [52]. which makes it extremely difficult to
counterfeit or hide a living individual’s motion signature [52].
3. Radar-based Continuous Identity Authentication Research
3. Radar-based
Identity Continuous Identity
authentication Authentication
using microwave Research
Doppler radar is gaining attention as it can add an
extra layerauthentication
Identity of security to the vulnerable
using microwave traditional
Doppler one-pass
radarvalidation
is gainingapproach
attention (e.g., fingerprint,
as it can add an
password, and facial/iris) [58]. Pattern recognition and unique identification
extra layer of security to the vulnerable traditional one-pass validation approach (e.g., fingerprint, are always challenging
for this non-contact technology because of variations in human breathing patterns due to physical
password, and facial/iris) [58]. Pattern recognition and unique identification are always challenging
activity and emotional stress [55]. As future big data analyses emerge and machine learning algorithms
for this non-contact technology because of variations in human breathing patterns due to physical
improve, Doppler radar-measured physiological signals can be turned into increasingly useful data
activity and emotional stress [55]. As future big data analyses emerge and machine learning
and knowledge [58]. In particular, diverse respiratory motion patterns have good potential to be used
algorithms improve, Doppler radar-measured physiological signals can be turned into increasingly
as biometric identifiers [58,64–72].
useful data
Identityandtheftknowledge
continues[58]. In everyday
to pose particular,challenges
diverse for respiratory
consumers motion
and thepatterns
associated havethreatgood
potential to be used as biometric identifiers [58,64–72].
is increased as traditional identity authentication systems are targeted [58,69]. Traditional identity
Identity theftmethods,
authentication continuessuch to pose everydaypassword,
as fingerprint, challengesand forfacial
consumers and the
recognition, all associated threat is
require an initial
increased
spot checkas traditional
at the start identity authentication
of user session, systems are
which potentially targeted
conveys [58,69].
personal Traditional
information like identity
bank
authentication
account, social methods,
securitysuchnumber,as fingerprint,
and credit cardpassword,
and socialand facial recognition,
networking all require
account details [72]. Inan initial
2018,
spotover
check at the start
14 million people ofwere
uservictims
session,ofwhich
identitypotentially
fraud in the conveys
United personal
States [69].information
Identity fraud likecan bank
be significantly
account, reduced
social security by implementing
number, and credit multi-factor
card and socialauthentication
networking systems,
account which can [72].
details be further
In 2018,
overenhanced
14 million through
people integration
were victims of unobtrusive
of identitycontinuous
fraud in the radar-based identity
United States authentication
[69]. Identity fraud [58].can be
significantly reduced by implementing multi-factor authentication systems, which can bebased
In this section, radar-based sensing authentication is categorized in two different ways, further
either on
enhanced breathing-related
through integrationfeatures, or heart-based
of unobtrusive features.
continuous Breathing identity
radar-based motion isauthentication
generally periodic, [58].
and respiratory-related features can be extracted from the time domain signature
In this section, radar-based sensing authentication is categorized in two different ways, based of the reflected phase
signal and rate information extracted by performing an FFT [4–6]. Heartbeat motion is modulated on top
either on breathing-related features, or heart-based features. Breathing motion is generally periodic,
of respiratory motion and the larger resulting breathing signal is dominant over the heartbeat signal [4].
and respiratory-related features can be extracted from the time domain signature of the reflected
This leads to a classical problem in FFT, where the stronger signal at given frequency leaks into other
phase signal and rate information extracted by performing an FFT [4–6]. Heartbeat motion is
frequencies and can mask a weaker signal at nearby frequencies [4,5]. Generally, the radar captured
modulated on top of respiratory motion and the larger resulting breathing signal is dominant over
signal is filtered outside the 0.005–0.5-Hz frequency band for extracting respiration information and
the 0.8–2
heartbeat
Hz forsignal [4]. This
extracting leads to a classical
heartbeat-related problem
information. All theinAll
FFT,
the where the strongerresearch
radar authentication signal at given
cited
frequency leaksisinto
in this paper other
focused on frequencies
extracting two and can mask
separate a weakerfeatures,
distinguishable signal at nearby
based frequencies
on either respiration [4,5].
Generally, the radar captured signal is filtered outside the 0.005–0.5-Hz frequency
or heartbeat. Extracting both simultaneously has the potential for stronger authentication; however, band for extracting
respiration information
such a process and 0.8–2
may increase Hz for extracting
computational heartbeat-related
complexity. information.
Table 1 provides a summaryAll the
of All the radar
published
authentication researchnon-contact
work on radar-based cited in this paper isidentity
continuous focused on extracting
authentication two separate
considered in this distinguishable
paper. In the
features, based
next two on eitherdetails
subsections, respiration or heartbeat.
are provided on theseExtracting
two differentbothunique
simultaneously has the potential
features (breathing and heart,for
stronger authentication; however, such a process may increase computational complexity. Table 1
provides a summary of published work on radar-based non-contact continuous identity authentication
considered in this paper. In the next two subsections, details are provided on these two different unique
features (breathing and heart, respectively), including related identification demonstrations along with
associated challenges for further development. There have also been attempts to use Doppler
Remote Sens. 2020, 12, 2279 5 of 22
respectively), including related identification demonstrations along with associated challenges for
further development. There have also been attempts to use Doppler modulation of WiFi signals to
authenticate people and this research is described in the third subsection.
Table 1. Systematic review on radar-based non-contact continuous identity authentication.
Reference Year Hardware
Identification Features Outcome
of Publication (RF Frequency)
• Respiration-based
• Accuracy: 90%
[52] A. Rahman 2.4 GHz # Power
• Neural network classifier
et al., 2016 Doppler Radar # spectral density
• Participants: 3
# Packing density
# Inspiratory time
• Heart-based dynamics • Accuracy: 98.61%
[58] F. Lin et al., 2.4 GHz • Support Vector Machine
2017 Doppler Radar # Cardiac-motion cycle
• Participants: 78
# Five points
• Respiration-based • Accuracy: 95%
[68] A. Rahman 2.4 GHz • K-nearest neighbor
et al., 2018 Doppler Radar # Inhale-exhale area ratio
• Participants: 6
# Minor component
• Accuracy: 100%
[70] S. M. M. • Support Vector Machine
2.4 GHz • Respiration-based
Islam et al., • Participants: 10
Doppler Radar # FFT-based feature
2019 • Only sedentary breathing
• Accuracy: 98.8% (normal)
• Respiration-based • Accuracy: 92% (combined)
[71] S. M. M. • Support vector machine
2.4 GHz
Islam et al., # Exhale Area: Air flow • Mixture of sedentary and after
Doppler Radar
2020 short exertion breathing
# Breathing depth
• Participants: 10
• Respiration-based OSA patient • Accuracy: 93%
[72] S. M. M. 2.4 GHz and • OSA patient recognition
Islam et al., 24-GHz # Peak power spectral density
• Support Vector Machine
2020 Doppler Radar # Linear envelop error
• Participants: 6
# Inspiratory duration
• Heart-based dynamics • Accuracy: 82%
[73] D.
2.4 GHz # Cardiac motion • K-nearest neighbor
Rissacher et al.,
Doppler Radar # Wavelet based time and • Participants: 20
2015
frequency feature
• Accuracy: 94.6%
[74] K. Shi et al., 24 GHz • Heart based dynamic • Support Vector Machine
2018 Doppler Radar # Heartbeat signal complexity • Participants: 4
Remote Sens. 2020, 12, 2279 6 of 22
Table 1. Cont.
Reference Year Hardware
Identification Features Outcome
of Publication (RF Frequency)
• Accuracy: 92.8%
[75] T. Okano 24 GHz • Heart based dynamics • Autoregressive analysis
et al., 2017 Doppler Radar # Power spectral density • Participants: 11
• Accuracy: 98.5%
• Heart based dynamics • Convolutional
• Short-time Fourier Transform Neural Network
[76] P. Cao et al., 24 GHz
2020 Doppler Radar # Heartbeat • Participants: 10
# Energy • Mixture of normal and
# Bandwidth abnormal breathing
• Accuracy: 93%
[77] J. Zhang WiFi router & • Channel state Information
• K-nearest neighbor
et al., 2016 Laptop # Gait Pattern • Participants: 16
• Respiration-based • Accuracy: 95%
[78] J. Liu et al., WiFi router & Morphological pattern • Deep learning
2020 Laptop • Fuzzy Wavelet based • Participants: 20
3.1. Radar-Based Identity Authentication through Respiration Related Features
One of the first attempts at radar-based identity authentication using breathing pattern was
reported by a research group at the University of Hawaii, where they integrated a neural network
classifier to recognize individual human beings [52]. In this work they extracted three different
respiratory features (peak power spectral density, linear envelop error, and packing density) from
respiratory motion measurements, as illustrated in Figure 3. A 2.4 GHz quadrature direct conversion.
Doppler radar system was used for this experiment, assembled from coaxial components. A data
acquisition system recorded the data and post processing was performed in MATLAB (MathWorks,
Natick, MA, USA). Three subjects with similar breathing rates were selected and various features
were investigated, such as power spectral density, linear envelop error, and packing density, which
convey the breathing energy and air flow profile related phenomena. The research concentrated on
using the Levenberg–Marquadrt back propagation algorithm to perform classification [52]. Figure 4
illustrates the experimental setup and reported results for training and applying a neural network
classifier to recognition of the Doppler radar physiological measurements [52]. The overall classification
accuracy was above 90%, which clearly illustrates that the proposed technique can be effective for this
application. However, this work was limited to identifying only three participants. Another issue is
that the experiment was entirely focused on measuring a single subject at a time.
illustrates the experimental setupback
using the Levenberg–Marquadrt and propagation
reported results for training
algorithm and applying
to perform a neural
classification [52].network
Figure 4
classifier
illustratestotherecognition
experimental of setup
the Doppler radarresults
and reported physiological
for trainingmeasurements
and applying[52]. The network
a neural overall
classification
classifier to accuracy
recognition wasof
above
the 90%, whichradar
Doppler clearly illustrates that
physiological the proposed [52].
measurements technique
The can be
overall
effective for this application. However, this work was limited to identifying only three
classification accuracy was above 90%, which clearly illustrates that the proposed technique can be participants.
Another
effectiveissue is that
for this the experiment
application. wasthis
However, entirely
workfocused on measuring
was limited a single
to identifying onlysubject
three at a time.
participants.
Remote Sens. 2020, 12, 2279 7 of 22
Another issue is that the experiment was entirely focused on measuring a single subject at a time.
Figure 3. Radar-measured respiratory features from 30 second epochs. Pioneering efforts at
recognizing
Figure subject identity
Figure3.3.Radar-measured
Radar-measured from radar
respiratory
respiratory measured
features fromrespiratory
features 30 second
from signalsPioneering
30 epochs.
second (a) involved
epochs. extraction
efforts of three
at recognizing
Pioneering efforts at
different
subject features:
identity breathing
from radar rate (b),
measured linear envelop
respiratory error
signals(c),
(a)and packing
involved density
extraction (d).
recognizing subject identity from radar measured respiratory signals (a) involved extraction of threeof Linear
three envelop
different
error shows
features:
different the peak
breathing
features: ratedistribution
(b), linear
breathing differences
envelop
rate (b), and
error
linear envelop (c),packing
and (c),
error density
packing illustrates
density
and packing the(d).
differences
(d). Linear
density envelop
Linear in air
error
envelop
flow
errorprofile
shows shows with
the peak the the
peak inhale
distribution and exhale
differences
distribution areapacking
and episodes.
differences Taken
anddensity
packing from [52].
illustrates
density the differences
illustrates the in air flow profile
differences in air
with
flowthe inhale
profile withandthe
exhale
inhalearea
andepisodes.
exhale areaTaken from [52].
episodes. Taken from [52].
(a)
(a)
(b)
(b)
Figure 4. Human identification experiment using a radar system. The test set-up is shown, (a), along
with the neural network recognition results, (b). From [52].
Subsequent research by the same group reported on continuous authentication based on dynamic
segmentation where they used inhale and exhale area ratios of the captured respiratory pattern as
unique features for six different participants [68]. Dynamic segmentation evaluates the displacement
and identifying points in the range of 30–70% of both exhale and inhale episodes, which defines four
boundary points of a trapezium [65]. The ratio of these two areas provides a useful feature which
indicates how quickly the next cycle of inhalation begins [68]. Figure 5 illustrates the inhale/exhale area
ratio features for two different subjects, which differ significantly. Based on extracted unique features,
pattern as unique features for six different participants [68]. Dynamic segmentation evaluates the
displacement and identifying points in the range of 30%–70% of both exhale and inhale episodes,
which defines four boundary points of a trapezium [65]. The ratio of these two areas provides a useful
feature which indicates how quickly the next cycle of inhalation begins [68]. Figure 5 illustrates the
Remote Sens. 2020, 12, 2279 8 of 22
inhale/exhale area ratio features for two different subjects, which differ significantly. Based on
extracted unique features, a K-nearest algorithm was integrated to identify each person, which
a K-nearest
showed algorithm
a classification was integrated
accuracy to identify
of almost eachInperson,
90% [68]. which
order to showed
increase theaaccuracy
classification accuracy
of the proposed
of almost 90% [68]. In order to increase the accuracy of the proposed method, minor component
method, minor component analysis was performed on subject data sets which showed overlapping
analysis was performed on subject data sets which showed overlapping inhale/exhale area ratios.
inhale/exhale area ratios. For minor-component analysis, a linear demodulation technique was
For minor-component analysis, a linear demodulation technique was employed [68]. The variation in
employed [68]. The variation in minor component shows the radar cross section and high frequency
minor component shows the radar cross section and high frequency component of respiration and
component of respiration
heart signal modulation. and heart
Higuchi signal
fractal modulation.
dimension analysisHiguchi fractalondimension
was performed analysisof was
minor components
performed
the radaroncaptured
minor components of the
signals to identify radar captured
overlapped signals
inhale/exhale toratios
area identify overlapped
of subjects, whichinhale/exhale
increased
area the
ratios of subjects, which increased the classification accuracy to 95% [68]. The
classification accuracy to 95% [68]. The proposed method clearly shows efficacy. However, proposed method
clearly
theshows
number efficacy.
of subjectsHowever,
of tested the
wasnumber
small andoffurther
subjects of tested was
investigation small and
is required furtherlarger
to establish investigation
data
set functionality.
is required to establish larger data set functionality.
Figure
Figure 5. Respiratory
5. Respiratory pattern
pattern classifiers used
classifiers used for
for subject
subjectrecognition.
recognition.Dynamically
Dynamically segmented
segmented
inhale/exhale area ratios of two subjects significantly differs, (a), as do signal patterns relating to the
inhale/exhale area ratios of two subjects significantly differs, (a), as do signal patterns relating to the
dynamics of breathing near the points where the inhale and exhale transition occurs, (b). From [68].
dynamics of breathing near the points where the inhale and exhale transition occurs, (b). From [68].
Another limitation of this approach is the reliance on two different parameters (inhale/exhale area
Another
ratio limitation
and minor of thisanalysis).
component approach is theinvestigations
Further reliance on alsotwodemonstrated
different parameters (inhale/exhale
that, as inhale/exhale
area ratio
ratiobecome
and minor component
more similar, analysis).may
false classification Further investigations
occur [70]. also demonstrated
To increase performance further, an that,
FFT as
inhale/exhale ratioextraction
based feature become more approachsimilar, false classification
was applied may occur
with an integrated [70]. To increase
support-vector machines performance
(SVM)
classifier
further, an FFTusing a radial
based basis extraction
feature function [70]. The performance
approach was appliedof the with
proposed system increased
an integrated as the
support-vector
machines (SVM) classifier using a radial basis function [70]. The performance of the proposedrate,
FFT based feature extraction approach contains all breathing dynamics related features (breathing system
breathing
increased depth,
as the FFT inhale
basedrate,feature
exhale rate and airflow
extraction profile). contains
approach Figure 6 illustrates the FFTdynamics
all breathing based feature
related
extraction approach used for six different participants. As the data set and number of participants
features (breathing rate, breathing depth, inhale rate, exhale rate and airflow profile). Figure 6
was small, continued experimentation remains needed to verify the efficacy of the FFT-based feature
illustrates the FFT based feature extraction approach used for six different participants. As the data
extraction approach. For further investigation, the feasibility of the FFT-based approach for extracting
set and number of participants was small, continued experimentation remains needed to verify the
identifying features from radar respiratory traces for sedentary subjects was tested, along with
efficacy of the FFT-based
measurements featurejust
of the subjects extraction approach.
after performing For further activities
physiological investigation,
(walkingtheupstairs)
feasibility of the
[69].
FFT-based approach
It was found that for extracting
subject identifying
recognition features
still worked from
but was notradar respiratory
as effective traces for sedentary
after performing short
subjects was tested,
exertions as it wasalong with measurements
for sedentary of the subjects
subjects [71]. Experimental justdemonstrated
results after performing physiological
that, after short
activities (walking
exertion, upstairs)segmented
the dynamically [69]. It was found
exhale areathat subject recognition
and breathing stillby
depth increased worked
more thanbut1.4
was
timesnot as
for all
effective participants,
after performing which made
short evident the
exertions as uniqueness of the residual
it was for sedentary heart [71].
subjects volume after expiration
Experimental results
for recognizing each subject, even after short exertions [71]. They also integrated a machine learning
classifier SVM, with a radial basis function kernel which resulted in an accuracy of 98.55% for subjects
in a sedentary condition and almost 92% for a combined mixture of conditions (sedentary and after
short exertions) [71]. Furthermore, they also investigated identity authentication of patients with
obstructive sleep apnea (OSA) symptoms based on extracting respiratory features (peak power spectral
demonstrated that, after short exertion, the dynamically segmented exhale area and breathing depth
increased by more than 1.4 times for all participants, which made evident the uniqueness of the
residual heart volume after expiration for recognizing each subject, even after short exertions [71].
They also integrated a machine learning classifier SVM, with a radial basis function kernel which
resulted in an accuracy of 98.55% for subjects in a sedentary condition and almost 92% for a combined
Remote Sens. 2020, 12, 2279 9 of 22
mixture of conditions (sedentary and after short exertions) [71]. Furthermore, they also investigated
identity authentication of patients with obstructive sleep apnea (OSA) symptoms based on extracting
respiratory features
density, packing (peakand
density, power
linearspectral
envelopdensity,
error) forpacking density, paradoxical
radar captured and linear envelop
breathingerror) for
patterns,
radar captured paradoxical
in a small-scale breathing
clinical sleep patterns, in three
study integrating a small-scale
differentclinical
machinesleep study integrating
learning three
classifiers (SVM,
different
k-nearestmachine
neighborlearning
(KNN), classifiers
and random (SVM, k-nearest
forest). neighbor (KNN),
Their proposed OSA-based andauthentication
random forest). Their
method
proposed
was testedOSA-based
and validated authentication
for five OSA method was
patients withtested andaccuracy,
93.75% validatedusing
for five
a KNNOSAclassifier
patientswhich
with
93.75% accuracy,
outperformed otherusing a KNN[72].
classifiers classifier which
This study wasoutperformed other
limited to only classifiers
six supine [72]. in
subjects This
thestudy was
controlled
limited to onlyofsix
environment supine
a sleep subjects in the controlled environment of a sleep center.
center.
Figure6.
Figure FFT/SVMbased
6.FFT/SVM basedsubject
subjectrecognition
recognitionstudy.
study. A
A2.4-GHz
2.4-GHzradar
radarsystem
system(a)
(a)was
wasused
usedtotomeasure
measure
subjects in seated position (b) and FFT-based extracted features up to 10 Hz were used to identify
subjects in seated position (b) and FFT-based extracted features up to 10 Hz were used to identify six six
different participants (c). From [70].
different participants (c). From [70].
3.2. Radar-Based Identity Authentication through Heart-Based Features
3.2. Radar-Based Identity Authentication through Heart-based Features:
One of the first attempts at recognizing people from their heart-based features (cardiac cycle)
One of the first attempts at recognizing people from their heart-based features (cardiac cycle)
from Doppler radar was reported on by a research group at Clarkson University [73]. They used a
from Doppler radar was reported on by a research group at Clarkson University [73]. They used a
2.4-GHz heterodyne radar system from which cardiac data was extracted, and an ensemble average
2.4-GHz heterodyne radar system from which cardiac data was extracted, and an ensemble average
was computed using ECG as time reference [73]. A continuous wavelet transform was integrated to
was computed using ECG as time reference [73]. A continuous wavelet transform was integrated to
provide time-frequency analysis of the average radar-measured cardiac cycle and a k-nearest neighbor
provide time-frequency analysis of the average radar-measured cardiac cycle and a k-nearest
algorithm was used to recognize people with an accuracy of 82%. This was the first reported attempt
neighbor algorithm was used to recognize people with an accuracy of 82%. This was the first reported
for applying cardiac-based features using a cardiac-radar system as biometric identification tool [73].
attempt for applying cardiac-based features using a cardiac-radar system as biometric identification
The low classification accuracy occurred as there was overlap in the ensemble average of the cardiac
tool [73]. The low classification accuracy occurred as there was overlap in the ensemble average of
cycle; therefore, further investigation and experimentation is required to demonstrate efficacy for more
the cardiac cycle; therefore, further investigation and experimentation is required to demonstrate
reliable recognition of subjects from radar captured signals.
efficacy for more reliable recognition of subjects from radar captured signals.
A study conducted by a research group at the University of Buffalo [58] proposed a continuous
A study conducted by a research group at the University of Buffalo [58] proposed a continuous
identity authentication system named “Cardiac Scan”. This system used a 2.4-GHz Doppler radar
identity authentication system named “Cardiac Scan”. This system used a 2.4-GHz Doppler radar
transceiver with two antennas (one for transmit and another for receive functions) each having a
transceiver with two antennas (one for transmit and another for receive functions) each having a
beam width of 45◦ . The radar power consumption was 650 mW with a 5V-volt source and 130 mA of
beam width of 45°. The radar power consumption was 650 mW with a 5V-volt source and 130 mA of
current [58]. The customized Doppler radar was placed in front of the subject at 1 meter [58]. Figure 7
current [58]. The customized Doppler radar was placed in front of the subject at 1 meter [58]. Figure
shows the experimental set up of the proposed cardiac scan system, from which five different points
7 shows the experimental set up of the proposed cardiac scan system, from which five different points
were extracted from the radar captured respiration patterns which were hypothesized to fully represent
were extracted from the radar captured respiration patterns which were hypothesized to fully
cardiac motion. Based on the hypothesis the experiment illustrated that these heart-based geometry
represent cardiac motion. Based on the hypothesis the experiment illustrated that these heart-based
measures differ from person to person due to difference in size, position and anatomy of the heart,
chest configuration, and various other factors [58]. From their experimental results, it was also clear
that no two subjects had the same heart, tissue, and blood circulation system, as there were significant
differences in their cardiac cycle points measured in the radar data set [58]. Figure 8 illustrates the
cardiac motion marker for one segment captured from the radar respiration measurement. In this
work, the user’s cardiac-motion related features were stored in the system. A SVM with a radial basis
clear that no two subjects had the same heart, tissue, and blood circulation system, as there were
heart, chest configuration, and various other factors [58]. From their experimental results, it was also
significant differences in their cardiac cycle points measured in the radar data set [58]. Figure 8
clear that no two subjects had the same heart, tissue, and blood circulation system, as there were
illustrates the cardiac motion marker for one segment captured from the radar respiration
significant differences in their cardiac cycle points measured in the radar data set [58]. Figure 8
measurement. In this work, the user’s cardiac-motion related features were stored in the system. A
illustrates the cardiac motion marker for one segment captured from the radar respiration
SVM with a radial basis function (RBF) kernel classifier was employed to uniquely identify different
Remote Sens. 2020, 12,
measurement. In 2279
this work, the user’s cardiac-motion related features were stored in the system. 10 of 22
A
participants. A study of a 78 subject data set was reported, and the proposed system achieved an
SVM with a radial basis function (RBF) kernel classifier was employed to uniquely identify different
accuracy of 98.61% and a 4.42% equal error rate [58]. One of the limitations of the above proposed
participants.
function (RBF) Akernel
study classifier
of a 78 subject data settowas reported, and the proposed system achieved an
technique is that the complete was studyemployed uniquely
was performed withidentify
healthydifferent participants.
sedentary persons andA study
only offora
accuracy of 98.61% and a 4.42% equal error rate [58]. One of the limitations of the above proposed
single-subject
78 subject datameasurements.
set was reported, If subjects have cardiovascular
and the proposed diseases,
system achieved unique identification
an accuracy of 98.61% and may be
a 4.42%
technique is that the complete study was performed with healthy sedentary persons and only for
problematic
equal error rate as the
[58].cardiac
One ofcycle would be of
the limitations affected.
the aboveFurther studytechnique
proposed is also required to complete
is that the verify that the
study
single-subject measurements. If subjects have cardiovascular diseases, unique identification may be
proposed heart-based
was performed cardiacsedentary
with healthy cycle points remain
persons andconsistent after subjects measurements.
only for single-subject perform varying If degrees
subjects
problematic as the cardiac cycle would be affected. Further study is also required to verify that the
of physiological
have cardiovascular activity.
diseases, unique identification may be problematic as the cardiac cycle would be
proposed heart-based cardiac cycle points remain consistent after subjects perform varying degrees
affected. Further study is also required to verify that the proposed heart-based cardiac cycle points
of physiological activity.
remain consistent after subjects perform varying degrees of physiological activity.
Figure 7. Experimental setup for cardiac scan continuous authentication system using microwave
Doppler
Figure 7.radar. A data acquisition
Experimental device and
setup for cardiac scanLABVIEW interfaces
continuous were used
authentication to capture
system signals. A
using microwave
Figuresensor
pulse 7. Experimental
and Achest setup
werefor cardiac scan continuous authentication system using microwave
Doppler radar. databelt used
acquisition for reference
device measurements.
and LABVIEW From
interfaces [58].
were used to capture signals.
Doppler radar. A data acquisition device and LABVIEW interfaces were used to capture signals. A
A pulse sensor and chest belt were used for reference measurements. From [58].
pulse sensor and chest belt were used for reference measurements. From [58].
Figure
Figure 8.
8. Cardiac
Cardiacmotion
motionmarker.
marker. The
Thecardiac
cardiacmotion
motioncycle
cycledefined
definedby
byfive
fivedifferent
different points
points(red
(reddots)
dots)
within five
fivedifferent
differentpoints of displacement
points and timing
of displacement was calculated
and timing as a unique
was calculated as feature for recognizing
a unique feature for
Figure 8.From
people.
recognizingCardiac
[58].motion
people. From marker.
[58]. The cardiac motion cycle defined by five different points (red dots)
within five different points of displacement and timing was calculated as a unique feature for
In another
another reported
recognizing
In people.
reported study,
From cardiac
[58].
study, measurement
cardiac measurement of different persons
of different was used
persons wastoused
uniquely identify
to uniquely
each using
identify a 24-GHz
each using continuous wave radar system
a 24-GHz continuous employing
wave radar system six-port measurement
employing six-porttechnology
measurement [71].
FigureIn 9another reported
represents the study, cardiac
hardware setup measurement
used for this of differentApersons
experiment. six-port was used to uniquely
measurement system
technology [71]. Figure 9 represents the hardware setup used for this experiment. A six-port
identify of
consists each using
input aports
twosystem 24-GHz continuous
andoffour wave The
radar system signals
employingsuperimposed
six-port measurement
measurement consists twooutput
input ports.
ports and two
fourinput
output ports.areThe two input signalsto extract
are
technology
phase [71]. Figure
shift information due9 to
represents the hardware
chest displacement. setup used
This particular forfocused
work this experiment.
on extractingAheartbeat
six-port
measurement
signal system
information forconsists of two input
each participant, ports
as the andposition
exact four output
and ports.
angle ofThethetwo input
heart signals
in the are
thorax,
as well as the anatomy of the thorax itself, is a little different for every person due to varying tissue and
muscle/fat components [74]. Due to these differences, the radar-captured heartbeat signal involved
different propagation and attenuation characteristics. As each person has a different heart position and
dimensions, dominant features exist in the heartbeat signal which form a complex and unique pattern.
Figure 10 illustrates the heartbeat signal variation for each participant. Initially a 5-second heartbeat
signal was used for identifying unique features for each participant. Integrated machine learningangle of the heart in the thorax, as well as the anatomy of the thorax itself, is a little different for every
participant. Integrated machine learning classifiers were also used to recognize people. A quadratic
person due to varying tissue and muscle/fat components [74]. Due to these differences, the radar-
SVM outperformed
captured heartbeat other classifiers,
signal involved with an accuracy
different propagationofand 74.2%. To increase
attenuation the accuracy,
characteristics. As eacha 7-second
heartbeat segment
person was used,
has a different heartwhich increased
position the classification
and dimensions, accuracy
dominant features tothe
exist in 94.6%. The signal
heartbeat study provides
a clearwhich
indication
Remote form
Sens. that
12, 2279heart-based
a complex
2020, geometry
and unique pattern. Figurecan be usedthe
10 illustrates asheartbeat
a unique feature
signal to for
variation identify
11each
of 22 people.
participant. Initially a 5-second heartbeat signal was used for identifying
However, validation for this study only included four different participants. Thus, further unique features for each
participant. Integrated machine learning classifiers were also used to recognize people. A quadratic
investigation iswere
classifiers required
also usedfor larger data setsAhaving varying conditions, especially those an involving
SVM outperformed othertoclassifiers,
recognize people.
with quadratic
an accuracy SVM outperformed
of 74.2%. other
To increase the classifiers,
accuracy, with
a 7-second
measurements
accuracy of
heartbeat made
74.2%.after
segment was physical
To increase
used, which activities.
the accuracy,
increaseda the
7-second heartbeat
classification segment
accuracy to was used,
94.6%. Thewhich
studyincreased
provides
athe classification
clear indicationaccuracy to 94.6%. geometry
that heart-based The study can
provides a clear
be used as aindication that heart-based
unique feature to identifygeometry
people.
can be used as a unique feature to identify people. However, validation for
However, validation for this study only included four different participants. Thus, this study only included
further
four different is
investigation participants.
required for Thus, further
larger data investigation is required
sets having varying for larger
conditions, data setsthose
especially having varying
involving
conditions, especially
measurements thosephysical
made after involving measurements made after physical activities.
activities.
Figure 9. Experimental setup for unique identification of a human from radar captured respiration
patternFigure
which Experimental
Experimental
9. includes setup
setup for
six-port for unique
unique identification
technology.identification
From [74]. of
of a human from radar captured respiration
pattern which includes six-port technology.
technology. From [74].
Figure 10. Heartbeat curves recorded by a 24-GHz radar for four subjects. Each signal is periodic but
for each subject the pattern is a bit different which serves as a unique feature for recognition of identity.
From [74].Figure 10. Heartbeat curves recorded by a 24-GHz radar for four subjects. Each signal is periodic but
for each subject the pattern is a bit different which serves as a unique feature for recognition of
identity. From [74].
Remote Sens. 2020, 12, 2279 12 of 22
Another study used a 24-GHz radar system to extract heartbeat related unique features to
recognizeAnother
eleven different
study used participants
a 24-GHz [75].
radarAn autoregressive
system (AR) model-based
to extract heartbeat frequency
related unique analysis
features to
was introduced,
recognize elevenwhich is superior
different to FFT,
participants having
[75]. a window length
An autoregressive of 100 milliseconds,
(AR) model-based frequency from which
analysis
powerwasspectral density
introduced, could
which be calculated
is superior to FFT, [75].
having Each peak inlength
a window this analysis represents the
of 100 milliseconds, fromcontraction
which
power spectral
and extraction density
of the could
heart. Thebe calculated
first peak was [75].used
Eachas
peak in this analysis
a reference and thenrepresents the contraction
a period of 0.2 s before
and extraction
and after, 0.4 s, wasof the heart.
used as aThe first peakTemplate
template. was used as a reference
matching andused
was then atoperiod
detectofall
0.2heartbeats.
s before and The
after,of
average 0.4the
s, was used asspectral
power a template. Template
density was matching
used was
as aused to detect
unique all heartbeats.number
identification The average
forofeach
the power spectral density was used as a unique identification number for each participant. Figure 11
participant. Figure 11 illustrates the power spectral density features extracted from the radar
illustrates the power spectral density features extracted from the radar respiration signal and PSD
respiration signal and PSD profile for eleven different participants. The success rate was 92.8%. The
profile for eleven different participants. The success rate was 92.8%. The proposed method clearly
proposed method clearly demonstrates heart-based PSD feature extraction efficacy for recognizing
demonstrates heart-based PSD feature extraction efficacy for recognizing people. However, if the
people. However,
position between if the position
the radar and between the radar
human subject and
changes or human
the heartsubject changesgreatly
rate fluctuates or thethen
heart
the rate
fluctuates greatly then the proposed method produces false classification. Motion
proposed method produces false classification. Motion artifacts and multi-subject scenarios were not artifacts and multi-
subject scenarios
considered were
and not aconsidered
remain and remain
significant challenge a significant
for this approach. challenge for this approach.
FigureFigure
11. 11. Measured
Measured 24-GHz
24-GHz heartbeatpatterns
heartbeat patterns for
for autoregressive
autoregressivePSD
PSDanalysis based
analysis subject
based subject
recognition. Measurement of heartbeats are shown for electrocardiogram (reference) (a), Doppler radar
recognition. Measurement of heartbeats are shown for electrocardiogram (reference) (a), Doppler radar
(b), PSD of Doppler sensor output (c), and (d) PSD for 15-s Doppler radar for eleven participants [75].
(b), PSD of Doppler sensor output (c), and (d) PSD for 15-s Doppler radar for eleven participants [75].
Recently, another study demonstrated the efficacy of radar-based identity authentication using a
Recently, anotherTransform
short-time Fourier study demonstrated theperson
(STFT) [76]. Each efficacy
satof radar-based
a 1.5 m distance andidentity authentication
physiological using
signatures
a short-time Fourier Transform (STFT) [76]. Each person sat a 1.5 m distance
were recorded for about 6 seconds of the breathing pattern, using a 24-GHz continuous wave radar. and physiological
signatures
An STFT were
was recorded for aboutthe
used to characterize 6 seconds of thesignature
micro-Doppler breathing pattern,
of ten differentusing a 24-GHz
participants, continuous
followed by
basic image transformation methods like translation, rotation, zoom, mirror,
wave radar. An STFT was used to characterize the micro-Doppler signature of ten different and cropping. The STFT
image was
participants, used to represent
followed by basicheart-based features for each
image transformation differentlike
methods subject. A deep convolutional
translation, rotation, zoom,
mirror, and cropping. The STFT image was used to represent heart-based features micro-Doppler
neural network (DCNN) was used to classify subjects based on their radar captured for each different
signatures. Figure 12 illustrates the STFT images for four participants which are significantly different
subject. A deep convolutional neural network (DCNN) was used to classify subjects based on their
for each subject and include unique features for identification. From the spectrogram, three different
radar captured micro-Doppler signatures. Figure 12 illustrates the STFT images for four participants
heart-based features were extracted, such as the period of the heartbeat, the energy of the heartbeat,
which are significantly different for each subject and include unique features for identification. From
and the bandwidth of the signal. A deep convolutional neural network was then trained, and the
the spectrogram, three different
resulting classification heart-based
accuracy was almost 98.5%.features were extracted, such as the period of the
heartbeat,Tothe energy
extract of the heartbeat,
heart-based andinformation,
or respiratory the bandwidth of the signal.generally
data segmentation A deep plays
convolutional neural
an important
network
role. was then trained,
Segments correspondand theFFT
to the resulting
window classification
size and shouldaccuracy
contain was almost
at least 98.5%.
one full respiration cycle
and multiple cardiac cycles [58,70]. The number of segments used for a data set also plays an important
role for authentication time and accuracy [58]. Increasing the FFT window size will bring a benefit in
resolution as a higher number of samples are included in the operation, but this will also increase the
time delay and complexity of authentication and is not generally justified for real-time operation.You can also read
