Download Multi-frequency sensor for remote measurement of breath and

Adv. Radio Sci., 4, 79–83, 2006
www.adv-radio-sci.net/4/79/2006/
© Author(s) 2006. This work is licensed
under a Creative Commons License.
Advances in
Radio Science
Multi-frequency sensor for remote measurement of breath
and heartbeat
M. Jelen and E. M. Biebl
Fachgebiet Höchstfrequenztechnik, Technische Universität München, Germany
Abstract. Remote measurement of breath and heartbeat is
desirable in many situations. It avoids the discomfort resulting from electrodes applied on the skin for long-term patients
or during sports acvtivities. Also, surveillance of high security areas or finding survivors of disasters are interesting applications. Common methods identify the movement of heart
and thorax by using the range resolution provided by UWB
pulse radar systems. In this paper a low-cost approach is presented, that is based on detection of movement by means of
Doppler radar sensors. Combining three sensors working in
the ISM bands at 433 MHz, 2.4 GHz and 24 GHz, the presence of persons was reliably detected and the frequency of
breath and heartbeat was measured.
1 Introduction
Heartbeat and respiration are a common attribute of all humans. They have a characteristic form and frequency which,
in case of the heartbeat, can’t be affected voluntarily.
In technical systems (e.g. in cars) there is often the need to
detect the presence of a human. This task can be fullfilled by
a wide variety of sensors, such as mechanical switches (e.g.
in the seat) or ultrasonic sensors.
Also the frequencys of pulse and respiration are often measured. One obvious field of application is the medical sector, but numerous other applications (e.g. sport/fitness) can
be thougt of. Currently the measurements are performed by
sensors which are in contact with the person beeing observed.
This may be a serious disadvantage in situations were measurements must be taken over a long period of time (e.g. long
term patients in hospital) or were physical contact is not possible.
An approach of detecting humans by means of radar,
mainly for security applications, is described in Greneker
Correspondence to: E. M. Biebl ([email protected])
(1997). Bimpas et al. (2003) covers the use of a radar system
for the detection of persons trapped in building ruins after an
earthquake.
This article describes a multi-frequency-sensor approach
for both the reliable detection of persons and the measurement of heartbeat and respiration rate.
2 Physiological and technical basics
The human heart is approximately the size of a fist. It is located off-center in the thorax, two-third in the left part, one
third in the right. In resting state it beats abt. 70 times per
minute. With physical effort the activity can increase up to
200 bpm. The heart is covered by several layers, the most important are skin, fat and bone. The thickness of these layers
varies from person to person.
A movement detection of the heart by means of radar implies the transmission of electromagnetic waves into the human body as well as reflections at a border. To estimate the
amount of reflected energy it is necessary to know the dielectric properties of the different body tissues. A rather complete list can be found in Http://niremf.ifac.cnr.it/tissprop/.
For the calculation the well-known equations
!
ZF 2 − ZF 1 2
R[dB] = 10 log (1)
ZF 1 + ZF 2 and
!2
2 Z F 2 T [dB] = 10 log ZF 1 + ZF 2 ZF 2 ZF 1 (2)
can be used. Z F can be calculated from µr and r using the
following equation:
s
s
s
µ0 µr
µr
µr (1 − j tan δµ )
ZF =
= Z0
= Z0
(3)
0 r
r
r (1 − j tan δ )
Published by Copernicus GmbH on behalf of the URSI Landesausschuss in der Bundesrepublik Deutschland e.V.
Daempfung [dB]
Because the surface of the skin is also much larger than the
output of the mixer. If the object which reflects the wave
Because the surface of the skin is also much larger than the
moves around a fixed position (as it is the case with the beatsurface of the heart, the difference between the two reflected
moves around a fixed position (as it is the case with the beating heart) the amplitude of this voltage changes. But the
surface
of the heart, the difference between the two reflected
signals will be at last 30 dB. Those values are very similiar
ing heart) depends
the amplitude
this
voltage
changes.
2
Jelen
andposition
E.
M.ofBiebl:
Multi-fre
amplitude
also M.
onofthe
absolut
theBut
tar-the
signals
be at last 30 dB. Those values are very similiar
for 2.4will
GHz.
amplitude
depends
also
on
the
absolut
position
of
the
80
M.
Jelen
and
E.
M.
Biebl:
Multi-frequency
sensor
for
remote
measurement
get. This is illustrated in figure 3: A phase variation du to atarfor 2.4
GHz.
The reflection coefficent at the border air/skin at 24 GHz
get. This is illustrated in figure 3: A phase variation du to a
reflection
at the border air/skin
at which
24 GHz
isThe
about
-3.2 dB.coefficent
In good approximation
all power
is
0
second
partb feeds the LO-port of the receivin
Welleall
im Koerper,2,4
a
is transmitted
about -3.2 into
dB. the
In good
which
is
body approximation
can be considered
aspower
lost,GHzso
that no
a
b
transmitted
into
the
body
can
be
considered
as
lost,
so
that
no
ceived signal
feeds the IF-port of the mixer.
reflections from the heart can be measured at this frequency.
-5
reflections
fromprinciples
the heart can
bebe
measured
thistask:
frequency.
Two radar
could
used foratthis
ultraof the transmitted signal is ωT X , and the f
a’
Two radar principles could
be used for this task: ultrawide-band(UWB)-radar
and cw-doppler-radar.
a’
received signal is ωRX , and ϕTxXx and ϕR
wide-band(UWB)-radar
andpulses
cw-doppler-radar.
The
first one uses short
which are reflected at a tar-10
b’
sponding phases, the following signal can b
The The
first distance
one usesresolution
short pulses
reflectedis atinverse
a target.
thatwhich
can beareachieved
b’
get.
The distance
resolution
be achieved
is inverse
proportional
to the
length ofthat
thecan
pulses.
As the movement
IF-port of the mixer:
-15
of the hearttois the
in the
sub-cm-region,
veryAs
short
(and
proportional
length
of the pulses.
thepulses
movement
highsub-cm-region,
bandwith) would
beshort
needed.
Despite
oftherefore
the heartaisvery
in the
very
pulses
(and
Fig. 3.
3. Influence
Influence
phase
IFofof phase
= Aon1 cos[(ω
− ωRX )t − (ϕT X − ϕR
Fig.
mixer output
voltage.
T Xvoltage
-20 frequency
from
regulation
issuies
this be
would
lead to
a sotherefore
a very high
bandwith)
would
needed.
Despite
Fig.
3.
Influence
of
phase
on
mixer
output
voltage
phisticated
broad-band
design,
not this
suitable
for alead
low-cost
from
frequency
regulation
issuies
would
to a apsoAcauses
ωRX )t
− (ϕT X − ϕR
2 cos[(ω
X +change
movement within ”‘a”’
the T
voltage
of ”‘a‘”’.
proach.
Staderini
covers
the
use
of
such
a
system
in
medical
phisticated
broad-band
design,
not
suitable
for
a
low-cost
ap-25
A
phase
variation
about
(the
same
absolut)
amount
”‘b”’
movement
within
thetovoltage
”‘a‘”’.
-5
0
5
10
15
20
25
30
35
applications
in detail.
The second
one is”‘a”’
much
simpler
design. change
Figure 2ofshows
. . .causes
proach.
Staderini
covers the
use
of such
a system in medical
Strecke
im Koerper
[mm]
on
another
absolut
position
causes
a
much
smaller
change
A
phase
variation
about
(the
same
absolut)
amount
”‘b”’
a simple block diagram of a doppler-radar circuit. Contrary
The second
one is much simpler to design. Figure 2 shows
applications
in detail.
of
”‘b‘”’.
Because
the
amplitude
of the
heartbeat
is much
on
another
absolut
position
causes
a
much
smaller
change
to the pulseradar,
the signal
is sent continously depending
and unmod- on the m
aThe
simple
block
a doppler-radar
second
onediagram
is muchofsimpler
to design.circuit.
FigureContrary
2 shows
A
und
Aamplitude
smaller
thanBecause
the1 used
wavelengths,
this
effect
is significant
2 are constants
of ”‘b‘”’.
the
of the
heartbeat
is The
much
ulated.
Therefore
only
a
small
bandwith
is
occupied.
Fig.
1.
Propagation
of
a
wave
at
440
MHz
through
the
thorax
and
a simple block diagram of a doppler-radar circuit. Contrary
for
the detection
ofused
heartbeats.
quency
of
the
sent
and
received
signals are
smaller
than
the
wavelengths,
this
effect
is
significant
Fig. 1. Propagation
of
a
wave
at
440
MHz
through
the
thorax
and
back.
information is contained in amplitude, frequency and phase
Local Oscillator
Powersplitter
Circulator
Tothe
overcome
this
problem
of ”‘blind spots”’, the approach
for
detection
of
heartbeats.
of the received
signal.
The signal of
thefollowing:
local oscillator is
tion
simplifies
the
back
shown here uses
three
sensors at to
three
different frequencies.
Local Oscillator
Powersplitter
Circulator
To overcome
this
problem
ofradiated
”‘blind
spots”’,
the approach
divided
in
two
parts:
One
part
is
via
the antenna,
the
In almost all cases at least one sensor will be able to achieve
shown
here
uses
three
sensors
at
three
different
frequencies.
second
part
feeds
the
LO-port
of
the
receiving
mixer.
The
=least
A1one
cos(∆ϕ)will be able to achieve
a meaningfullVmeasurement
result.
Out at
In almostsignal
all cases
received
feeds the
IF-portsensor
of the mixer. If the freLowpass
L.O.
reflected wave is attenuated
by approximately
22 dB. The
a remeaningfull
measurement
result.
quency
of the transmitted
signal
is ωT X , and the frequency
Asignal
dcsetup
voltage
depending
onarethe
can be
R.F.
of
received
is ωRX , and
ϕT X and ϕRX
the phase
cor3 the
Measurement
flection coefficent
air-skin I.F.is about
2.38 dB at this frequency.
Output
Lowpass
L.O.
responding phases,
following
signal can
observed
at which r
outputtheof
the mixer.
Ifbethe
object
Because Output
the surface of theI.F. skin isR.F.also much larger thanthe
3the
Measurement
setup
Measurements
were
performed at three different frequency
IF-port
ofmoves
the
mixer:
around a fixed position (as it is the ca
ranges. To make commercial applications possible, all chosurface of the heart, the difference between the two reflected
Measurements
were
performed
atthe
three
ingare
heart)
the
of frequency
thisSci-voltage ch
Fig. 2. Simple block diagram of a cw-doppler-radar
sen
frequencies
located
withinamplitude
ISMdifferent
(Industrial,
signals will
be at last 30 dB. Those values are very similiar
Iranges.
F = A1To
cos[(ω
− ωRX )t − applications
(ϕT X − ϕRX)possible,
]+
T Xcommercial
make
choentific, Medicine)
bands.
Table
1
gives
an
overview
over
amplitude
depends
also](Industrial,
on theallthe
absolut
pos
block
diagram
of is
a cw-doppler-radar.
cw-doppler-radar
Fig.
2. Simple
Simple
block the
diagram
sen frequencies
are
within
SciA2 cos[(ωTand
ωRX )tpowers:
−
(ϕT Xthe
− ϕISM
for 2.4 Fig.
GHz.
X +located
RX) +
used
frequencies
output
to 2.the
pulsradar,
signal
sent continously and unmodget. This
isTable
illustrated
figure
3:
A phase
entific,
1of
gives
an in
overview
over
Figure
block diagram
the system
used for
the
ulated. Therefore
only a small
occupied. The
. . Medicine)
. 4 shows a bands.
(4)the
The reflection
coefficent
at thebandwith
borderis air/skin
at 24 GHz
used frequencies and output powers:
to information
the pulsradar,
the signalinisamplitude,
sent continously
and
unmodmeasurements.
is contained
frequency
and
phase
is aboutulated.
-3.2
dB.
Inansignal.
good
approximation
allMHz
power
is 440 4MHz,
Figure
shows
estimation
ofbandwith
a wave
430
being
Figure
shows
a block diagram
of the system
used for
only
aThe
small
occupied.
The
of
the 1Therefore
received
signal
of theatislocal
oscillator
iswhichFor
a circulator
was available,
thus a system
ac-the
A1 und A2 are constants dependinga on the mixer. bIf the frepropagated
through
the
body
towards
the
heart
and
being
remeasurements.
information
is contained
inpart
amplitude,
frequency
and
phase
divided
two
parts:
radiated
via the as
antenna,
the
cording
to figure 2 could be set up. The RF was supplied by
transmitted
intoin the
bodyOne
can
beis considered
lost,
so that
no
quency of the sent and received signals are equal, the equaflected
back on signal.
the border
fat/heart. of
The
that ais
of
thefrom
received
The
thegraph
localshows
For 440 MHz, a circulator was available, thus a system acreflections
theis heart
canbysignal
be
measured
atoscillator
this
frequency.
tion simplifies to the following:
reflected
wave
attenuated
approximately
22
dB.
The
redivided in two parts: One part is radiated via the antenna, the
cording to figure 2 could be set up. The RF was supplied by
air-skin
is about
dB atfor
this this
frequency.
Two flection
radar coefficent
principles
could
be2.38
used
task: ultraa’
Because the surface of the skin is also much larger than the
VOut = A1 cos(1ϕ)
(5)
wide-band(UWB)-radar
and cw-doppler-radar.
surface of the heart, the difference between the two reflected
The first
one
which
reflected
signals
willuses
be at short
least 30pulses
dB. Those
values are
are very
similiar at a tarb’ on the phase can be observed at the
A dc voltage depending
for
2.4
GHz.
get. The distance resolution that can be achieved is inverse
output of the mixer. If the object which reflects the wave
The reflection coefficent at the border air/skin at 24 GHz
proportional
the
of the pulses.
As which
the movement
moves around a fixed position (as it is the case with the beatis aboutto
-3.2
dB.length
In good approximation
all power
is
ing
heart) the amplitude of this voltage changes. But the
of the heart
is ininto
the
very
short
pulses
(and
transmitted
thesub-cm-region,
body can be considered
as lost,
so that
no
amplitude depends also on the absolute position of the tarreflections
from
the
heart
can
be
measured
at
this
frequency.
therefore a very high bandwith) would be needed. Despite
get. This is Fig.
illustrated
in Fig. 3. of
A phase
phase variation
dueoutput
to
3. Influence
on mixer
voltag
Two radar principles could be used for this task: ultrafrom frequency
regulationandissuies
this would lead to aa movement
sowithin ‘a’ causes the voltage change of ‘a’. A
wide-band(UWB)-radar
cw-doppler-radar.
phase
phisticated
broad-band
design,
suitable
for aat low-cost
ap- variation about (the same absolut) amount ‘b’ on anThe
first one uses short
pulsesnot
which
are reflected
a tarother absolutmovement
position causeswithin
a much ”‘a”’
smaller causes
change ofthe
‘b’. voltage c
get.
The
distance
resolution
that
can
be
achieved
is
inverse
proach. Staderini covers the use of such a system in medical
Because the amplitude
heartbeat is
much smaller
than absolu
proportional to the length of the pulses. As the movement
A phaseof the
variation
about
(the same
applications
detail.
the used wavelengths, this effect is significant for the detecof the in
heart
is in the sub-cm-region, very short pulses (and
on another absolut position causes a much
tion of heartbeats.
therefore one
a very
bandwith)
would
needed.Figure
Despite 2 shows
The second
ishigh
much
simpler
to be
design.
of this
”‘b‘”’.
the amplitude
To overcome
problemBecause
of ‘blind spots’,
the approach of the he
frequency regulation issues this would lead to a soa simplefrom
block
diagram of a doppler-radar circuit. Contrary
phisticated broad-band design, not suitable for a low-cost apshown here uses
three
sensors
at
three
different
frequencies. this eff
smaller than the used wavelengths,
proach. Staderini covers the use of such a system in medical
In almost all cases at least one sensor will be able to achieve
for the detection of heartbeats.
a meaningfull measurement result.
applications
in detail.
Local Oscillator
Powersplitter
Circulator
Adv. Radio Sci., 4, 79–83, 2006
To overcome this problem of ”‘blind spots
shown here
uses three sensors at three diffe
www.adv-radio-sci.net/4/79/2006/
In almost all cases at least one sensor will b
a meaningfull measurement result.
1,25 cm
24 - 24,25 GHz
0 dBm
Table 1. Used frequencys in the ISM range
M.
M.Jelen
Jelenand
and E.
E. M.
M. Biebl:
Biebl: Multi-frequency
Multi-frequency sensor...
sensor for remote measurement
Wavelength
70 cm
440 MHz
12,5
cm
+ 17 dBm
1,25 cm
Frequency range
433,05 - 434,79 MHz
Splitter/
2,4 - 2,5 GHz
Mixer
24 - 24,25 GHz
813
Output Power
0 dBm
0 dBm
0 dBm
Table 1. Used frequencys in the ISM range
2,4 GHz
+ 14 dBm
Splitter/
Mixer
440 MHz
+ 17 dBm
Splitter/
24 Mixer
GHz −
Radar module
AF −
2,4Amplifier
GHz
+ 14 dBm
Splitter/
A/D
−
Mixer
Converter
Fig. 5. The sensor used for 2.4 GHz
PC
AF −
Amplifier
the time domain representation the filtered signal was transformed back into the time domain.
The achievable frequency resolution ∆f is direct proportional to the length of the captured sequence N :
Fig. 5.
5. The
The sensor
sensor used
Fig.
used for
for 2.4
2.4 GHz.
GHz
24 GHz −
Radar module
Fig.
diagram
of theofsystem
used forused
the presented
measureFig.4.4.block
Block
diagram
the system
for the presented
AF −
Amplifier
ments
measurements.
A/D −
Converter
fs
(6)
N
The
GHz measurements
performed
using
com-the
the
timea24domain
representation
the filtered
wasa into
transWith
constant
sample ratewere
this
can be signal
converted
mercially
available
doppler
radar module with a horn antenna
formed
back
into
the
time
domain.
recording time ts :
with
20 dBi frequency
antenna gain.
The output
this module
Theabout
achievable
resolution
∆f isofdirect
proporwas ac-coupled,
response
to
very slow
movements
1the lengthsoofthe
tional
to
the
captured
sequence
N
:
∆f =
(7)
(e.g. breath
ts movement) was strongly attenuated or couldn’t
fs at all. This can be seen in the measurement rebe detected
(6)be
∆f
=resolution
If
a
of 0.1 Hz is needed, the capture time must
sults. N
atAll
least
10 seconds.
were
fed into
professional
16 bit A/D
With
aoutputs
constant
sample
rate athis
can be converted
intoconthe
verter
card
connected
to
a
notebook.
The
captured
data
recording time ts :
was processed using MATLAB. First, the dc-offset was re4 Measurement results
moved 1from the signal by calculating the mean value and
∆f =
(7)
subtracting
it from all samples. Then the captured time dot
For thes actual measurements a sample rate fs = 8 kHz was
main data was windowed by a hanning-window to reduce the
For theoffirst
measurement
a person
Ifused.
a resolution
0.1 Hz
is needed, the
capturewas
timeseated
must on
be a
leakage-effect and converted to the frequency domain using
asked to sit quiet without moving and hold the
atchair
least and
10 seconds.
the fast-fourier-transformation (FFT). There the signal was
breath for 30 seconds. The data from all three sensors were
band-pass-filtered with a passband from 0.2 Hz to 5 Hz. For
taken parallel. Figure 6 shows the raw time- and frequency
time domain representation
the filtered signal was trans4the
Measurement
results
domain
datainto
as captured
the sensors, figure 7 shows the
formed
back
the time from
domain.
same data
after digital
filtering.
frequency
resolution
proporForThe
theachievable
actual measurements
a sample 1f
rate is
f direct
= 8 kHz
was
As
one
can
see
in
figure
6 the
spectralNscomponents
tional to
thethe
length
the captured
: seated on aare
used.
For
first of
measurement
asequence
person was
limited to very low frequencies. Therefore the maximum
chair and
to sit quiet without moving and hold the
fs asked
frequency
range in figure 7 was reduced to 5 Hz. The fre1f = for 30 seconds. The data from all three sensors were
(6)
breath
quencyNresolution with a measurement time of 30 seconds is
taken
parallel.
Figure
6
shows
the
raw
timeand
frequency
1
A peak sample
is visible
at this
1.2 Hz.
This
equals ainto
hearttherate
30 Hz.
With
a constant
rate
can be
converted
domain
data as captured
from
the sensors,
figure 7 shows
the
of
72
bpm
which
was
confirmed
by
a
parallel
”‘classical”’
recording
same
data time
after tdigital
filtering.
s:
measurement.
As one can see in figure 6 the spectral components are
For 1the second measurement the person was again seated
limited
1f = to very low frequencies. Therefore the maximum
(7)
in
the tchair
about
m from
the sensors
sit frequiet,
s range
frequency
in2figure
7 was
reducedand
to 5asked
Hz. to
The
this time
breathing.
Figure 8 shows the
The breath
quency
resolution
with
timeresults.
of 30
seconds
is
If a resolution
of 0.1
Hzaismeasurement
needed, the capture
time
must be
1movement is clearly visible in the graphs resulting from the
Hz.
A
peak
is
visible
at
1.2
Hz.
This
equals
a
heart
rate
at
30 least 10 s.
at was
440 confirmed
MHz and 2.4
As mentioned
ofmeasurements
72 bpm which
by aGHz.
parallel
”‘classical”’before,
these
low
frequencies
were
cut
by
the
hardware
at 24
measurement.
GHz.
The
spectrum
has
a
peak
at
about
0.26
Hz,
which
4 For
Measurement
results
the second measurement
the person was again seated
in the chair about 2 m from the sensors and asked to sit quiet,
For time
the actual
measurements
a sample
fs = The
8 kHzbreath
was
this
breathing.
Figure 8 shows
therate
results.
used.
For
the
first
measurement
a
person
was
seated
on
a
movement is clearly visible in the graphs resulting from the
chair
and
asked
to
sit
quiet
without
moving
and
hold
the
measurements at 440 MHz and 2.4 GHz. As mentioned before, these low frequencies were cut by the hardware at 24
GHz. The spectrum has a Adv.
peakRadio
at about
Hz, which
Sci.,0.26
4, 79–83,
2006
∆f =
PC
Table 1. Used frequencies in the ISM range.
AF − signal generator. The maximum available outa commercial
Amplifier
put power was +17 dBm. Because a resistive power splitter
Wavelength
range
Output Power
was used
to split the Frequency
transmit and
reference
signal and due
70
cmdiagram
433,05
- 434,79
MHz
0 dBm
Fig.the
4. block
the system
used
for the presented
to
losses
in the of
circulator,
approximately
+10 measuredBm (10
ments were available at the antenna.
mW)
12,5 cm
2,4 - 2,5 GHz
0 dBm
For 2.4 GHz no circulator was available, so two separate
1,25
cm used for24
- 24,25
GHz
0 dBm
antennas
were
both
the transmit
and the
receive path.
a
commercial
signal
generator.
The
maximum
available
Both antennas were realised as patch antennas etched outon a
put power was +17 dBm. Because a resistive power splitter
low-loss substrate. To keep losses low, the power splitter and
was used to split the transmit and reference signal and due
mixer were also included on the same substrate. Figure 5
to
losses in the setup
circulator, approximately +10 dBm (10
3 the
Measurement
shows
the system. The power was also generated by a labomW) were available at the antenna.
ratory generator, and also about +10 dBm were fed into the
Measurements
performed at three different frequency
For 2.4 GHz were
transmit
antenna.no circulator was available, so two separate
ranges.
To
make
commercial
allpath.
choantennas were used for both theapplications
transmit andpossible,
the receive
The
24 GHz measurements
werethe
performed
using a comsen
frequencies
are
located
within
ISM
(Industrial,
SciBoth antennas were realised as patch antennas etched on
a
mercial
available doppler
radar 1module
with
a hornover
antenna
entific,
Medicine)
bands.
Table
gives
an
overview
the
low-loss substrate. To keep losses low, the power splitter and
with
20 dBi
antenna
gain. The output of this modused about
frequencies
and
output
mixer
were also included
onpowers.
the same substrate. Figure 5
ule
was
ac-coupled,
so
the
response
very
slowused
movements
Figure
shows aThe
block
diagram
oftothe
system
the
shows
the 4system.
power
was also
generated
by afor
labo(e.g.
breath
movement)
was
strong
attenuated
or
couldnt
be
measurements.
For
440
MHz,
a
circulator
was
available,
thus
ratory generator, and also about +10 dBm were fed into the
detected
ataccording
all. This can
be seen
in the
results.
a systemantenna.
to Fig.
2 could
be measurement
set up. The RF
was
transmit
All
outputs
were
fed
into
a
professional
16
bit
A/D
consupplied
by
a
commercial
signal
generator.
The
maximum
The 24 GHz measurements were performed using a comverter
card
connected
to
a
notebook.
The
captured
data
available
output
power
was
+17
dBm.
Because
a
resistive
mercial available doppler radar module with a horn antenna
was
processed
using
MATLAB.
First,
the
dc-offset
was
repower
splitter
was
used
to
split
the
transmit
and
reference
with about 20 dBi antenna gain. The output of this modmoved
from
the
signal
by
calculating
the
mean
value
and
signal
and
due
to
the
losses
in
the
circulator,
approximately
ule was ac-coupled, so the response to very slow movements
subtracting
itmovement)
from were
all samples.
Then
captured
time do+10 dBm
(10
mW)
available
atattenuated
thethe
antenna.
(e.g.
breath
was
strong
or couldnt
be
main
data
was
windowed
by
a
hanning-window
to
reduce
the
For
2.4
GHz
no
circulator
was
available,
so
two
separate
detected at all. This can be seen in the measurement results.
leakage-effect
and
converted
to
the
frequency
domain
using
antennas
were
used
for
both
the
transmit
and
the
receive
path.
All outputs were fed into a professional 16 bit A/D conthe
fast-fourier-transformation
There
signalon
was
Both
antennas
were realised
as(FFT).
patch antennas
etched
a
verter
card connected
to a notebook.
The the
captured
data
low-loss
substrate.
To
keep
losses
low,
the
power
splitter
and
band-pass-filtered
with
a
passband
from
0.2
Hz
to
5
Hz.
For
was processed using MATLAB. First, the dc-offset was remixer were
included
the same the
substrate.
Figure
moved
from also
the signal
by on
calculating
mean value
and5
shows
the
system.
The
power
was
also
generated
by
a
labosubtracting it from all samples. Then the captured time doratorydata
generator,
and alsobyabout
+10 dBm weretofed
into the
the
main
was windowed
a hanning-window
reduce
transmit
antenna.
leakage-effect and converted to the frequency domain using
the fast-fourier-transformation (FFT). There the signal was
band-pass-filtered
with a passband from 0.2 Hz to 5 Hz. For
www.adv-radio-sci.net/4/79/2006/
44 482
M.M.
Jelen
Jelen
and
Biebl:Multi-frequency
sensor...
and
E.E.
M.M.
Biebl:
sensor...
M. Jelen and E. M. Biebl:
Multi-frequency
sensor
forMulti-frequency
remote measurement
−3
0.01
0.01
0.01 0.01
66
6
6
0
0
44
4
4
−0.01
−0.01
−0.01−0.01
22
2
2
00
−0.02
−0.02
−0.02−0.020
0
00
10
20
10
10 10
20 20
t[s]20
t[s]
t[s] t[s] 2,4
Zeitbereich
2,4 GHz
Zeitbereich
Zeitbereich
2,4
Zeitbereich
2,4GHz
GHz GHz
0.1
0.1
0.1 0.1
30
30
30 30
0.03
0.03
0.03 0.03
0
0
0.02
0.02
0.02 0.02
−0.05
−0.05
−0.05−0.05
0.01
0.01
0.01 0.01
−0.1
−0.1
−0.1 −0.10
0
00
0.04
0.04
0.04
0.04
10
20
t[s]20 20
10
10 10t[s]
t[s] t[s]20
Zeitbereich
24 GHz
GHz
Zeitbereich
24
Zeitbereich
Zeitbereich24
24GHz
GHz
30
30
30 30
0.02
0.02
0.02
0.02
00
−0.04
−0.04
−0.04
0
−0.04
0
00
10
20
10
20
t[s]20
10
t[s]
10
20
t[s]
t[s]
30
30
30
30
00
−4
10−4
−4
xx 10
−4
44
−5
−5
00
30
30
30
30
30
30
30
30
30
30
30
30
2
22 2
−5
−5
−5 −5
1
11 1
−10
−10
−10 −1000
0
0
10
20
t[s] 20 20
10
10 10 t[s]
t[s] t[s] 20
Zeitbereich
24 GHz
GHz
Zeitbereich
24
Zeitbereich
24 GHz
GHz
Zeitbereich 24
0.01
0.01
0.01
0.01
30
30 30
30
0
0
0
1
2 f[Hz] 3
3
00
0
1
2
f[Hz]
−4
00
11−4
22 f[Hz]
44
10
f[Hz] 33 24
Spektrum
24 GHz
GHz
xx 10
−4
Spektrum
4−4
10
4
xx 10
Spektrum 24
24 GHz
GHz
Spektrum
44
4
4
55
55
4
4
55
55
2
2
11
1
1
0
00 000
3
4
11
22 f[Hz]
f[Hz]
00
11
22 f[Hz]
33 3
44 4
−3
f[Hz]
x−310−3
Spektrum
24 GHz
GHz
Spektrum
24
x 10
10x−310
x1.5
Spektrum
24 GHz
GHz
Spektrum
24
1.5
1.5
1.5
5
55 5
5
55 5
11
11
00
0.5
0.5
0.5
−0.01
−0.01
−0.01
0
0 0
0
0.5
0.5
0
0
0
0
0
00 000
3
4
11
22 f[Hz]
f[Hz]
00
11
22 f[Hz]
33 3
44 4
−3
f[Hz]
−310−3
x−3
Spektrum
2,4
GHz
x
10
Spektrum
GHz
Spektrum
2,4 2,4
GHz
x 10
3 x 10
Spektrum
2,4
GHz
3
3
3
00
2
1
1
00
00
30
30 30
30
10
20
t[s] 20 20
10 10 t[s] t[s]
30
30
30
00
0
00
0
−3
Spektrum 440
440 MHz
MHz
Spektrum
Spektrum
Spektrum 440
440 MHz
MHz
0
0
0
1
2 f[Hz] 3
3
00
0
1
2
f[Hz]
−3
00
11−3
22 f[Hz]
44
10
f[Hz] 33 2,4
Spektrum
2,4 GHz
GHz
xx 10
−3
−3
Spektrum
1.5
xx1.5
10
10
Spektrum
Spektrum 2,4
2,4 GHz
GHz
1.5
1.5
22
10
20
10
20
t[s]
t[s]
20
20
t[s]
t[s]
−5
−5
−5 −500
10
20
0
10
20 20
0
10 10
t[s] 20
−3
t[s]
−3
x−310−3
t[s] t[s] 2,4
Zeitbereich
2,4 GHz
GHz
xx 10
Zeitbereich
2,4 GHz
GHz
10
5 x 10 Zeitbereich
Zeitbereich
2,4
5
5
5
11
1
22 f[Hz]
33
f[Hz]
22 f[Hz]
33
f[Hz]
44
44
55
55
Fig.
Breath
and
heartbeat
captured
from
a distance
Fig.
Breath
and
heartbeat
captured
from
aadistance
ofof2of
m.2m
Fig.
8.8.8.
Breath
and
heartbeat
captured
from
distance
2m
2
2
11
10
10
2
22 2
−1
0
0
−5
−5
0
0
4
xx 10
10
4
22
0.5
0.5
00
00
1
0
0
−5
−5
0
10
20
−5
−5
0
10
20
t[s]
t[s]
−3 10
20
00
20
10−3 10
t[s]
t[s]
Zeitbereich
24 GHz
GHz
xx 10
Zeitbereich
24
5−3−3
10
5
xx10
Zeitbereich24
24GHz
GHz
Zeitbereich
55
0
0
−0.005
−0.005
−0.005
0
00 00
0
20
40 f[Hz] 60
60
80
100
20
40
80
100
00
20
40
60
80
100
f[Hz]
20
40 f[Hz]
80
100
f[Hz] 60
−3
4
44 4
0
Fig.
6. Raw
Raw
timefrequency
domain
data.
Fig.
6.
timefrequency
domain
data
Fig.
6.
timeand
frequency
domain
data
Fig.
6. Raw
Raw
timeandand
frequency
domain
10−3
−3
Zeitbereich 440
440 MHz
MHz
xx 10
−3
Zeitbereich
1
xx10
10
Zeitbereich
1
Zeitbereich440
440MHz
MHz
11
0.5
0.5
0.5
0.5
0
0
00
−0.5
−0.5
−0.5
−0.5
−1
−1
0
10
20
−1
−1
0
10
20
t[s]20
−3 10
00
20
t[s]
10−3 10
Zeitbereich
2,4 GHz
GHz
xx 10
t[s]
t[s]
Zeitbereich
2,4
5−3−3
xx10
10
5
Zeitbereich
Zeitbereich2,4
2,4GHz
GHz
55
55
0.005
0.005
0.005
0.005
1
1
0.5
0.5
0.5
0.5
−0.02
−0.02
−0.02
−0.02
−3
−310−3
x−3
Spektrum
440
MHz
10
Spektrum
MHz
Spektrum
440440
MHz
xx 10
6 x 10
Spektrum
440
MHz
66 6
5
5
0
0
0
00 0 0
20
40 f[Hz] 60
60
80
100
40
00 0 20
60
80
20−3 20 40
40 f[Hz]
60
80 80 100
100 100
f[Hz]f[Hz]
−3
10−3
−3
Spektrum
24 GHz
GHz
xx 10
Spektrum
24
xx1.5
10
10
Spektrum
Spektrum 24
24 GHz
GHz
1.5
1.5
1.5
11
0
0
−3
−3
x−310−3
Zeitbereich
440
MHz
xx10
10
Zeitbereich
440 440
MHzMHz
10x 10 Zeitbereich
Zeitbereich
440
MHz
10
10 10
Spektrum
440 MHz
Spektrum
440
Spektrum
440
Spektrum
440 MHz
MHzMHz
0
00 0
0 20 20
20 40 40
40 f[Hz]
60
80
100
00 0
60
20
40 f[Hz]
60 60 80
80 80 100
100 100
f[Hz]f[Hz]
Spektrum
2,4 GHz
GHz
Spektrum
2,4
Spektrum
2,4
GHz
0.04
Spektrum
2,4
GHz
0.04
0.04
0.04
0.05
0.05
0.05 0.05
00
−3
−3
x 10
−3
x 10
8 x 10
8
88 x 10
Zeitbereich
440 MHz
MHz
Zeitbereich
440
Zeitbereich
440
Zeitbereich
440MHz
MHz
0.02
0.02
0.02 0.02
2 f[Hz] 3
3
2
f[Hz]
22 f[Hz]
f[Hz] 33
44
4
4
55
55
−4
x 10
10−3
Zeitbereich 440
440 MHz
MHz
x−3
Zeitbereich
x 10
Zeitbereich 440 MHz
22
6
11
4
00
−1
−1
−2
−2
10
20
−2
00
10
20
t[s] 20
0
10
t[s]
Zeitbereich
2,4 GHz
GHz
t[s]
Zeitbereich
2,4
0.01
0.01
Zeitbereich 2,4 GHz
0.01
0.005
0.005
0.005
00
0
−0.005
−0.005
−0.005
−0.01
−0.01
10
20
−0.01
00
10
20
t[s]
t[s]
−3
20
0
x 10
10−3 10 Zeitbereich
t[s]
24 GHz
GHz
x−3
Zeitbereich
24
44
x 10
Zeitbereich 24 GHz
4
22
2
00
0
−2
−2
−2
−4
−4
10
20
00
10
20
−4
t[s]
t[s]
20
0
10
t[s]
2
30
30
30
2
30
30
30
30
22
55
55
44
22
00
0
00
11
22 f[Hz]
44
f[Hz] 33
−4
0 x 101
22 f[Hz]
44
−4
f[Hz] 3324
Spektrum
24 GHz
GHz
x−4
10
−4
Spektrum
44
x 10
Spektrum
Spektrum 24
24 GHz
GHz
4
2
30
44
00
0
00
11
22 f[Hz]
44
f[Hz] 33
−4
0 x 101
22 f[Hz]
−4
f[Hz] 332,4 GHz44
Spektrum
x−4
10
−4
Spektrum
2,4 GHz
x 10
66
Spektrum 2,4
2,4 GHz
GHz
Spektrum
6
4
30
10−4
Spektrum 440
440 MHz
MHz
−4
xx−4
10
Spektrum
x 10
Spektrum 440
440 MHz
MHz
66
Spektrum
55
55
22
00
00
0
0
1
11
22 f[Hz] 33
22 f[Hz]f[Hz]
3
f[Hz] 3
44
44
55
55
Fig.7.
7. Filtered
Filtered timetime- and
and frequency
domain
Fig.
frequency
domaindata.
data
Fig.
7. Filtered
Filtered
time- and
and frequency
frequency
domain
Fig.
7.
timedomain
data
Fig. 9. Heartbeat captured from a distance of 2m
Fig.9.9.Heartbeat
Heartbeat captured
captured from
m.
from aa distance
distanceof
of22m
Fig.
breath aforbreath
30 s. frequency
The data from
all three
were taken
equals
about
15.6sensors
per minute.
This
equals
breath
frequency
of of
equals
aa breath
frequency
of
about
15.6
per
minute. domain
This
parallel.
Figure
6
shows
the
raw
timeand
frequency
was
in
fact
the
breath
frequeny
of
the
test
person.
wasdata
in fact
fact
the breath
breath
frequeny
of theFig.
test7 person.
was
in
the
frequeny
as captured
from
the sensors,
shows
the same
After
the
measurement
the
person
was asked
to data
stop
After
the measurement
measurement
the person was
After
the
the
asked to stop
after
digital
filtering.
breathing
so
that
only
the
heartbeat
could
be
detected.
Figure
breathing
so that
that
only
the
heartbeat
breathing
so
only
heartbeat
could be
detected. Figure
As one
can
see inthe
Fig.
the spectral
components
are lim-at
9 shows
the
result.
Here
a6 peak
in the spectrum
is visible
shows
the
result.
Here
a peak
peak
99 shows
the
result.
Here
a
in
the
spectrum
is
visible
at
ited
to
very
low
frequencies.
Therefore
the
maximum
freall three measurement frequencies at 1.3 Hz. As a reference,
all three
three
measurement
frequencies
all
measurement
frequencies
at
1.3
Hz.
As
a
reference,
quency
range
in
Fig.
7
was
reduced
to
5
Hz.
The
frequency
the heart rate of the test person was measured with a pulse
the resolution
heart rate
rate with
of the
the
test person
person time
the
heart
of
test
measured
a Hz.
pulse
a measurement
of 30 This
s iswith
1/30
A
watch
which
showed
79 beats was
per minute.
equals
1.317
watch
which
showed
79Hz.
beats
watch
which
showed
79
beats
perequals
minute.
This rate
equals
1.317
peak
is visible
at 1.2
This
a heart
of 72
bpm
Hz.
Hz.which was confirmed by a parallel ‘classical’ measurement.
Hz.
Comparing the peak powers of breath and heartbeat with
Comparing
the peak
peak
powers of the
breath
andwas
heartbeat
with
For the second
measurement
person
again seated
Comparing
the
powers
the formula
the
formula
in
the
chair
about
2
m
from
the
sensors
and
asked
to
sit
quiet,
the formula
peakrespiration
this timepeak
breathing.
Figure 8 shows the results. The breath
(8)
V = 20log
peakrespiration
respiration
=
20log ispeak
movement
clearly
visible in the graphs resulting from(8)
the
VV =
20log
heartbeat
peakheartbeat
peak
heartbeat
a difference of about 16 dB for 2.4 GHz and 20 dB for 440
difference
of Sci.,
about
16
dB for
for
aa difference
of
about
dB
2.4
GHz andpower
20 dBisfor
440
Adv. is
Radio
4,16
79–83,
2006
MHz
found.
Assuming
that
all returning
reflected
MHz is
is found.
found. Assuming
Assuming that
that all
all returning power is reflected
MHz
at the border air/skin at the breath measurement, these values
at the
the border
border air/skin
air/skin at
at the
the breath
breath measurement, these values
at
are close to those predicted before. This may be a hint that
predicted
before.
This As
maymentioned
be a hintbethat
are
close to those
measurements
at 440
MHz and
2.4 GHz.
actually reflections
heart
were
measured.
at at
thethe
heart
were
fore, these
low frequencies
were
cut measured.
by the hardware in the
actually
reflections
If
only
information
needed
aabout
person
present
24IfGHz
sensor.
The spectrum
has wether
a wether
peak at
0.26
Hz,
only
thethe
information
is is
needed
a person
is is
present
within
the
observed
range,
a
threshold
in
the
power
spectrum
which the
equals
a breath
frequency
of about
per spectrum
minute.
within
observed
range,
a threshold
in the15.6
power
could
beindefined.
This
was
tested
different
constellations.
This
was
fact the
breath
frequeny
ofdifferent
the
test constellations.
person.
This
was
tested
forfor
could
be defined.
In all cases the personwere
weresuccesfully
succesfullyif ifthey
theywere
were inthe
the
After
thethe
measurement
the person was asked to in
stop
In all
cases
person
line of sight of the antennas. Even through obstacles, such as
antennas.
Even through
obstacles,
breathing
the heartbeat
could be
detected.such
Fig-as
line
of sightso
ofthat
theonly
a closed
wooden
door, reliable recognition of the presence of
9 shows
the door,
result.reliable
Here arecognition
peak in the of
spectrum
is vis-of
the presence
aure
closed
wooden
persons could be achieved. Table 2 gives an overview over
ible at all
three
measurement
frequencies
at 1.3
Hz. Asover
a
achieved. Table
2 gives an
overview
persons
could
be
the increase
of the low-frequency-spectrum compared to a
reference,
the
heart
rate of the test person was
measured
low-frequency-spectrum
compared
to a
the
increase
of
the
reference
measurement
in an empty
roomper minute. This
with
a pulse
watch which
79room
beats
in showed
an empty
reference
measurement
equals 1.317 Hz.
Comparing the peak powers of breath and heartbeat with
5 Conclusions
5 Conclusions
www.adv-radio-sci.net/4/79/2006/
CW-doppler-radar sensors
at three different frequencies were
sensors at three different frequencies were
CW-doppler-radar
sucessfully used to detect the presence of persons and to measucessfully used to detect the presence of persons and to mea-
M. Jelen and E. M. Biebl: Multi-frequency sensor for remote measurement
Table 2. Increase of maximum compared to reference measurement.
Increase
Measurement distance
440 MHz
2,4 GHz
24 GHz
Distance 1 m
> 35 dB
> 35 dB
> 35 dB
Distance 3 m
22,5 dB
23,5 dB
> 40 dB
Distance 3 m + door
16,5 dB
31,1 dB
–
Distance 8 m
21 dB
14 dB
30 dB
Distance 12 m
2,5 dB
15,5 dB
16,9 dB
the equation
V = 20log
peakrespiration
peakheartbeat
(8)
a difference of about 16 dB for 2.4 GHz and 20 dB for
440 MHz is found. Assuming that all returning power is reflected at the border air/skin at the breath measurement, these
values are close to those predicted before. This may be a hint
that actually reflections at the heart were measured.
If only the information is needed whether a person is
present within the observed range, a threshold in the power
spectrum could be defined. This was tested for different constellations. In all cases the persons were succesfully detected
if they were in the line of sight of the antennas. Even through
obstacles, such as a closed wooden door, reliable recognition
of the presence of persons could be achieved. Table 2 gives
an overview over the increase of the low-frequency-spectrum
compared to a reference measurement in an empty room.
www.adv-radio-sci.net/4/79/2006/
5
83
Conclusions
CW-doppler-radar sensors at three different frequencies were
sucessfully used to detect the presence of persons and to measure heartbeat and respiration rate over a distance of several
meters. This was achieved with a low-cost hardware setup
which could be easily implemented, for example, in the interior of a car. As expected, 440 MHz and 2.4 GHz show very
similiar properties for this application, whereas 24 GHz can
only detect body movements due to its very small penetration
depth.
The system still has some potential for improvements, for
example the sensitivity of the radar sensors could be increased and therefore the output power could be decreased
even further. Also only very basic DSP techniques were used
for this measurements. A combination of the output from different sensors in combination with more sophisticated DSP
algorithms could help to eliminate unwanted responses and
overcome the disadvantage of the dependence from the absolute position.
References
Http://niremf.ifac.cnr.it/tissprop/.
Bimpas, M., Nikellis, K., Paraskevopoulos, N., Economou, D., and
Uzunoglu, N.: Devlopement and testing of a detector system for
trapped humans in building ruins, in: 33rd European Microwave
Conf., 999–1002 , 2003.
Greneker, E.: Radar Sensing of Heartbeat and Respiration at a Distance with Security Applications, Proc. of SPIE, 22–27, 1997.
Staderini, E.: UWB Radars in Medicine, IEEE Aerospace and Electronic Systems Magazine, 13–18, 2002.
Adv. Radio Sci., 4, 79–83, 2006