Research Article Noncontact Detection and Analysis...

Research ArticleNoncontact Detection and Analysis of Respiratory FunctionUsing Microwave Doppler Radar

Yee Siong Lee1 Pubudu N Pathirana1 Robin J Evans2 and Christopher L Steinfort3

1School of Engineering Faculty of Science Engineering and Built Environment Deakin University Geelong VIC 3216 Australia2Department of Electrical and Electronic Engineering Melbourne University Parkville VIC 3010 Australia3University Hospital Geelong Geelong VIC 3220 Australia

Correspondence should be addressed to Pubudu N Pathirana pubudupathiranadeakineduau

Received 28 October 2014 Accepted 9 January 2015

Academic Editor Yu Chen Tsai

Copyright copy 2015 Yee Siong Lee et al This is an open access article distributed under the Creative Commons Attribution Licensewhich permits unrestricted use distribution and reproduction in any medium provided the original work is properly cited

Real-time respiratory measurement with Doppler Radar has an important advantage in the monitoring of certain conditions suchas sleep apnoea sudden infant death syndrome (SIDS) and many other general clinical uses requiring fast nonwearable and non-contact measurement of the respiratory function In this paper we demonstrate the feasibility of using Doppler Radar inmeasuringthe basic respiratory frequencies (via fast Fourier transform) for four different types of breathing scenarios normal breathingrapid breathing slow inhalation-fast exhalation and fast inhalation-slow exhalation conducted in a laboratory environment Ahigh correlation factor was achieved between the Doppler Radar-basedmeasurements and the conventional measurement device arespiration strapWe also extended this work from basic signal acquisition to extracting detailed features of breathing function (I Eratio) This facilitated additional insights into breathing activity and is likely to trigger a number of new applications in respiratorymedicine

1 Introduction

Respiration monitoring is essential in the diagnosis andtreatment of conditions such as chronic obstructive pul-monary disease heart disease and a number of sleeprelated conditions [1] Furthermore dysfunctional respira-tory patterns such as rapid or shallow breathing [2] or highfrequency breathing rates have also been associated withcertain psychosomatic conditions [3] all of which at presentare typically measured via respiration rates alone Howevera more detailed analysis of breathing patterns [4ndash9] will pro-vide physicians with new insights into diagnostic medicineparticularly if this can be performed noninvasively Non-contact Doppler Radar has already been considered in avariety of patient monitoring and measurement scenarios inhealthcare including heartbeat and respiration monitoring inplace of conventional methods such as the chest strap photo-plethysmograph [10] and ECG [11] Research reported usingDoppler Radar in measuring human physiological activity[12ndash18] has predominantly demonstrated the feasibility ofDoppler Radar in obtaining breathing frequency or heart

rate using FFT wavelet analysis or time-frequency analysis[14 19 20]

A complete respiration cycle is typically defined byinhalation (inspiration) and exhalation (expiration) statesaccompanied by a pause as described in [21] Breathing ratesare predominantly calculated independent of the inhalationto exhalation ratio (I E) for each breathing cycle For normaland spontaneous breathing there is an abundance of time forthe exhalation process from the inspired tidal volume but incertain pathological states for instance asthma and COPD(chronic obstructive pulmonary disease) reduced expiratoryflow would need longer time to empty the inspired lungvolume [22] Typically for adults a normal I E ratio is in therange of 1 2 but this varies between individuals depending onthe health and the physiological state of the individual[23] Consequently more information of each component isextremely important as it can be useful in early detection ofseveral respiratory disorders

Another important parameter associated with breathingis respiratory tidal volume [24 25] which can also be derivedfrom microwave radar due to the relationship between

Hindawi Publishing CorporationJournal of SensorsVolume 2015 Article ID 548136 13 pageshttpdxdoiorg1011552015548136

2 Journal of Sensors

the chest wall displacement and the tidal volume Differenttypes of breathing can potentially be deduced from such chestwall or abdomendisplacement information during inhalationand exhalation This information can be used to identifydifferent types of breathing signatures such as shallow breath-ing deep breathing slow breathing fast breathing and othertypes of breathing patterns Indeed the displacement of thechest wall or abdomen in shallow breathing is expected tobe small and the complete breathing cycle would occur in ashorter time period compared to normal breathing

Doppler Radar operates by transmitting a radio wavesignal and receiving the modulated version of the signaldue to the motion triggered by the target [24 26] Thereflected wave is in the modulated form where it undergoes afrequency shift proportional to the radial velocity that can bedescribed using theDoppler effectWhen a target has a quasi-periodic motion the time varying position of the target canbe represented as a phase modulated signal and the phaseshift is directly proportional to the objectrsquos movement Thusthe movement of the chest wallabdomen for respiration dueto the inhalation exhalation and the pause states can bedetected and modelled using the reflected Doppler shiftedsignal the main focus of this paper We provide a com-prehensive description of the noncontact respiratory mea-surement via Doppler Radar which was then validated withindependent measurements using a respiration belt andbreathing cycle counts We also demonstrate the differenttypes of inhaling and exhaling states from data collectedusing our Doppler Radar systemThe purpose of this paper issummarized as follows

(i) investigation of Doppler Radarrsquos feasibility in captur-ing different types of breathing patterns under variousbreathing scenarios

(ii) correction of 119868119876 signal imbalance and cross-validation of Doppler breathing signal with stan-dard respiration measurement the respiration belt(MLT1132 iezo-respiratory belt transducer)

(iii) decomposition of the breathing signal (from DopplerRadar) into its respective inhalation and exhalationcomponents representing each component modelusing 4th polynomial fitting (see Table 2(a)) andclassifying decomposed breathing components intoits respective breathing scenarios

2 Methods

21 Respiration Monitoring via Microwave Doppler RadarThe Doppler effect occurs when there is a shift in thefrequency of the wave either reflected or radiated received byan object in motion [27] Consider a transmitted sine wavesignal with an angular frequency 120596

0

119879119909= sin (120596

0119905 + 1206010) (1)

where 119879119909is the transmitted signal 119905 is the time and 120601

0is the

arbitrary phase shift Assume that the target is stationary at adistance of 119903

0from the radar and the transmission time from

radar to target is 1199030119888where 119888 is thewave propagation velocity

The target range at time 119905 is given by equation below 119903(119905) =

1199030+ 119903(119905 minus 119905

0) where 119903 is the range of the target from the radar

and 119903 (velocity) is the rate of change of 119903 and 1199050is the time at

119903 = 1199030 The received signal at the stationary target is the same

as the transmitted signal at the time 1199030119888 which can be given

as

119877target = sin(1205960119905 minus

12059601199030

119888+ 1206010) (2)

The received signal from the target at time 119905 would havebeen sent Δ119905 seconds prior to time 119905 This can be representedas Δ119905 = 2119903

0119888 Referring to (1) signal can be depicted in the

same formulation given as

119877119909= sin (120596

0(119905 minus Δ119905) + 120601

0) (3)

Substituting Δ119905 = 21199030119888 into (3) the received signal is further

represented as

119877119909= sin(120596

0119905 minus

212059601199030

119888+ 1206010) (4)

For a target moving (radially) with respect to the radar thedistance will vary and by using 119903(119905) = 119903

0+ 119903(119905 minus 119905

0) and 120596

119889=

21205960119903119888 the received signal can be further derived as

119877119909= sin(120596

0(119905 minus

2119903 (119905)

119888) + 1206010)

= sin(1199080(119905 minus

21199030

119888minus2 119903 (119905 minus 119905

0)

119888) + 120601

0)

= sin(1205960(1 minus

2 119903

119888) 119905 minus

21205960

119888(1199030minus 1199031199050) + 1206010)

= sin((1205960minus 120596119889) 119905 minus

212059601199030

119888+ 1205961198891199050+ 1206010)

(5)

where the frequency of the reflected signal is shifted by 120596119889

and the phase angle by 1205961198891199050 Therefore the Doppler shift 120596

119889

can also be denoted by 120596119889= 2120587119891

119889 where 119891

119889= 2 119903119891

0119888 is the

Doppler shift in Hertz and 1198910is the transmitted frequency

Using 120582 = 1198881198910 119891119889can be written as 119891

119889= minus2 119903120582 where

the negative sign accounts for the fact that if 119903 is negative(when the target is approaching) the Doppler frequency willbe positive or vice versa [27] From (5) the phase angle Φ ofthe received signal is given as120596

1198891199050Therefore the transmitted

wave from the radar to the target will be reflected to thereceiver with some phase shifting and can be represented asphase modulation given as

Φ =21205960119903

1198881199050=4120587 (119903)

120582 (6)

The measurement model for human respiration usingDoppler Radar can be derived as follows Generally theDoppler shift in frequency is given by

119891119889 (119905) =

2119891V (119905)119888

=2V (119905)120582

(7)

where V(119905) is the velocity of the target 120582 is the wavelengthof the transmitted signal and 119888 is velocity of the propagating

Journal of Sensors 3

wave Assuming the target to be stationary or undergoing aperiodic movement of 119909(119905) with no net velocity the Dopplerfrequency shift can be represented in the form of nonlinearphase modulation as the phase signal Φ

119903(119905) given by Φ

119903(119905) =

4120587119909(119905)120582 where 119909(119905) is the displacement of the chest wallor abdomen Using a continuous wave (CW) radar thetransmitted signal is represented by

119879 (119905) = cos (1205960119905 + 1206010(119905)) (8)

where 119879(119905) is the transmitted signal and 1206010is the arbitrary

phase shift or the phase noise of the signal source if thetransmitted wave 119879(119905) is reflected by the targetsubject at anominal distance 119889

0with a time varying displacement of 119909(119905)

which is caused by the movement of the torso (abdomen)Thus the distance [28] between the transmitter and the targetis given as 119889(119905) = 119889

0+ 119909(119905) The measurement of the time

delay between the transmitter and the target is denoted asthe distance travelled over the signalrsquos propagation velocitygiven as 119889(119905)119888 Thus due to the movement of the abdomenduring the process of respiration the distance between theantenna and the abdomen at the time of reflection is denotedby 119889(119905minus119889(119905)119888) and the round trip time can be further derivedas 119905119889= 2(1198890+ 119909(119905 minus 119889(119905)119888))119888

Using the similar formulation shown in (3) along with1205960= 2120587119891 and 119888 = 119891120582 the received signal 119877(119905) can be

represented as

119877 (119905) = cos [1205960(119905 minus 119905119889) + 120601 (119905 minus 119905

119889)]

= cos [1205960(119905 minus

21198890+ 2119909 (119905 minus 119889 (119905) 119888)

119888)

+ 120601(119905 minus21198890+ 2119909 (119905 minus 119889 (119905) 119888)

119888)]

(9)

and further approximated as

119877 (119905) asymp cos(2120587119891119905 minus41205871198890

120582minus4120587119909 (119905)

120582+ 120601(119905 minus

21198890

119888)) (10)

Demodulation of the phase is used to determine the motionsignature which can be detected at the receiver In the directconversion system the received signal will be mixed withlocal oscillator to obtain the baseband output given as

119861 (119905) = cos(120579 + 4120587119909 (119905)

120582+ Δ120601 (119905)) (11)

In a quadrature receiver system the received signal will besplit into two forms which are an in-phase (119868

119861(119905)) and a

quadrature phase (119876119861(119905)) signal where the phase difference

will be 1205872 Therefore general two orthogonal basebandoutputs of the quadrature receiver system can be denoted by

119868119861(119905) = cos(120579 + 4120587119909 (119905)

120582+ Δ120601 (119905))

119876119861 (119905) = sin(120579 + 4120587119909 (119905)

120582+ Δ120601 (119905))

(12)

Here 120579 = 41205871198890120582 is the constant phase shift dependent on

the nominal distance to the target and Δ120601(119905) is the residual

phase noise The benefit of using a quadrature receiver is toovercome the null problem [11] where at least one output(either 119868119876) is not null when the other is null

22 Signal Processing Decomposition and Identification Acomplete breathing cycle is comprised of inhalation (119868)exhalation (119864) and pause components where the ratio of I Ecan certainly be asymmetric [23] Therefore computation ofbreathing rates purely based on simple single frequencysignatures computed via fast Fourier transforms (FFT) is notsufficient to provide detailed breathing pattern features par-ticularly for the identification and analysis of respiratory con-ditions Firstly the basic received signal is sent to the 119868119876 (in-phase and quadrature phase) demodulator for direct conver-sion into its baseband differential 119868119876 signal and then sam-pled at 1000Hz using NI-DAQ (National Instrument DataAcquisition System) The differential signals were then con-verted to a single ended baseband signal removing any DCcomponents of the raw signals and then processed in twodifferent approaches In the first approach the preprocessedraw data was modelled using a piecewise linear least squaresapproach [29] In the second approach the raw data was pro-cessed using a SG (Savitzky-Golay polynomial least square)[30] smoothing filter and further analysed using Fourier fil-tering [31]The first approach offers a simplemethod applica-ble for real-time processing while the second approach offersmore accurate identification of the respiration cycle compo-nents and their properties the main focus in this paper

23 Correction of 119868119876 Amplitude and Phase Imbalance Twoorthogonal outputs (119868 and119876) are obtained from a quadraturereceiver system but in practice (due to the imperfection ofcomponents in the hardware design) it suffers from ampli-tude and phase imbalance which affects the accuracy of therecovered data at the output [32] Consequently phase andamplitude corrections are necessary to increase accuracyThere are a number of approaches to correct the amplitudeand phase imbalance [33 34] In [34] a final form of twoorthonormal vectors using a method similar to the GramSchmidt orthogonalization (GSO) [32] has been proposed asshown in (17) The derivation of this is as follows The ideallyreceived signal 119877

119909(119905) is defined by

119877119909 (119905) = 119883119868 cos (1199080119905) + 119883119876 sin (1199080119905) (13)

where 119883119868and 119883

119876are the in-phase and quadrature phase of

the information signal respectively In our approach withthe presence of amplitude imbalance and phase offset thereceived signal at the mixer can be represented as

1198771015840

119909(119905) = 119877

119909(119905) lowast cos (119908

0119905) + 119877

119909(119905) lowast 119860

119890lowast sin (119908

0119905 + 120601)

(14)

where 119860119890and 120601 are the amplitude and phase imbalance

Demodulation of received signal is as follows

1198681015840= 119877119909(119905) lowast cos (119908

0119905)

1198761015840= 119877119909(119905) lowast 119860

119890lowast sin (119908

0119905 + 120601)

(15)


Expanding the derivation

1198681015840= 119883119868cos (119908

0119905) cos (119908

0119905) + 119883

119876sin (119908

0119905) cos (119908

0119905)

1198761015840= 119883119868cos (119908

0119905)

lowast 119860119890(sin (119908

0119905) cos (120601) + cos (119908

0119905) sin (120601))

+ 119883119876sin (119908

0119905)

lowast 119860119890(sin (119908

0119905) cos (120601) + cos (119908

0119905) sin (120601))

(16)

After the low pass filtering and ignoring the term 12representation of orthogonal119883

119868and119883

119876in matrix form

[

119883119868

119883119876

] =[[

[

1 0

minus tan (120601) 1

119860119890cos (120601)

]]

]

[

1198681015840

1198761015840] (17)

Using (17) correction on amplitude and phase imbalance canbe performed Simulation results of using this approach willbe discussed in Section 3

24 The Piecewise Linear Fitting Method This method fitsnonlinear typically noisy waveforms by choosing an optimalsegmentation of the waveform and then fitting each segmentwith a linear function [29] Here the segmentation process iscritical and in this case appropriate lengths of nonoverlap-ping segments were used Also we used fixed nonoverlappingsegments of 200ms to accommodate the Doppler Radarsignal

25 The Savitzky-Golay Method and Fourier Filtering TheSavitzky-Golay filter is a least square polynomial filter [30]By applying the filter to the noisy data obtained from thechemical spectrum analysers Savitzky and Golay demon-strated how it reduces noise while preserving the shape andheight of waveform peaks Here the SG filter was used tosmooth the input raw data after the DC components wereremoved The output from the SG filter improved the shapeof the signal significantly where noise and redundancy werefiltered extensively as shown in Figure 3 (data set 1) ((a) and(c))

The signals were smoothed by SG filter and then recon-structed using Fourier filtering This was to extract absolutemaxima and minima points of the breathing curve thatdenotes each of the inhalation and exhalation componentsFourier filtering from [31] has already been used as oneof the processing algorithms to further eliminate noise andto reconstruct the signals It is a filtering function thatmanipulates specific frequency components of a signal bytaking the Fourier transform of the corresponding signalswhich later either attenuate or amplify frequencies of interestIn this paper the Fourier filter was used to eliminate noiseemploying a band pass filter depending on the desiredbreathing frequency range while not distorting the signalsignificantly The shape of the Fourier filtered signal wasquite similar to the resulting signal from piecewise linearfitting but was smoother and local minima and maxima wereprominent

26 Breathing Signal Decomposition For the breathing cyclesobtained fromDoppler Radar we assumed that the transitionfrom local minima to local maxima on the curve representsthe inhalation component and vice versa for exhalation com-ponent respectively A peak detection algorithm was thenused to determine the maximum and minimum points ofeach transition defining the inhalation and exhalation com-ponents respectively These components were extracted sep-arately and represented by a fourth-order polynomial Wethen computed the average representation for normal andfast breathing components (inhalation and exhalation) to beused as a model for component identification as discussed inSection 532

27 Identification-Dynamic Time Warping Dynamic timewarping (DTW) is used to optimally align two time serieswhere one time series is transformed to best fit the other[35] This technique has been extensively used in speechrecognition to identify the similarity of spoken phases fromtwowaveforms as the duration of each spoken sound can varywith similar overall waveform shapes DTW has also beenused in other areas such as data mining and gait recognition[36] Typically similarity between two time series for thepurpose of classification often requires distancemeasurementbetween the twoComputation of Euclidean distance betweenthe two time series may not yield accurate results if oneof the two time series is slightly shifted along the timeaxis To overcome this limitation DTW was introduced asdescribed in [35] Here we use DTW for registering andcomparing breathing components to determine temporalfeatures (extracted breathing component model)

3 Experiment Mechanism

Measurement of humans respiration was approved by theFaculty of Science and Technology Ethics SubcommitteeHEAG (Faculty Human Ethics Advisory Groups) DeakinUniversity and all participants provided their writteninformed consent to participate in this study

A Doppler Radar system (Figure 1(a)) has a continuouswave (CW) that operates at 27 GHz with 214 dBm twopanel antennae where one is (Tx) and the other (Rx) 119868119876

demodulator (Analog Device AD8347) and a data acquisi-tion module (NI-DAQ) were used The received signals weredirectly converted into 119868119876 decomposition using AD8347where the demodulated signal was then sent to a DAQ forfurther processing using MATLAB

For this experiment the subjectwas positioned 05mawayfrom the antenna (transmitter Tx and receiver Rx) Thepanel antennae were aligned to focus on the abdomen tocapture a better Doppler effect due to respirationThe subjectwith normal clothing (see Figure 1(a)) and was asked to standin front of the antenna and breathe in specific ways for adetermined period of time as follows ldquonormal breathing(maintaining consistency in inhalation and exhalation rate)rdquoldquofast breathing (fast inhalation and fast exhalation)rdquo ldquofastinhalation and slow exhalationrdquo and ldquoslow inhalation and fastexhalationrdquo


NI‐DAQ

AD8347

AD

MATLAB

I Q

Local oscillator

AD8347

(IQ demodulator)

MLT1132 piezo-respiratorybelt transducer

Matlab environment

Tx and Rx

Tx

Rx

(a) Doppler Radar system

Piecewise linear filter

Savitzky-Golay filter

Fourier filter

FFT (spectral analysis)

Extraction of atomic component of breathing

Polynomial modeling

Atomic component identification

Filtering stage

Approximation of breathing rate

Raw data (IQ)

Atomic component decomposition modelling and classification

Local maxima andminima detection

(b) Signal processing flow

Figure 1 Doppler Radar system and signal processing flow

For each breathing pattern the number of breathingcycles was manually counted and recorded independently tobe compared with those computed using the proposed signalprocessing techniques as shown in Figure 1(b)

For validation purposes a respiband (MLT1132 piezo-respiratory belt transducer) attached to PowerLab (ADIn-struments) was used as a reference signal to compare with theDoppler measurements Results in Figure 2(b) show the nor-malized raw respiration signal obtained from the respirationbelt and normalized filtered Doppler Radar signals

From (17) the imbalance factors of 119860119890and 120601 need to be

estimated for 119868119876 correctionThis procedure is similar to theGSO procedure as the quadrature phase signal is orthogonalto the in-phase signal The simulation was performed byassuming that the breathing frequency is in the vicinity of

02Hz in the 119868 and 119876 representation In the simulationresults shown in Figure 2(a)(C) the phase offset of 25∘ withamplitude imbalance in quadrature signal was simulated inthe noisy signal We have estimated the amplitude imbalanceratio and phase offset between 119868 and119876 signal is corrected thesignal using (17) as shown in Figure 2(a) Amplitude imbal-ance was obtained by taking the average ratio of119876119868while thephase offset was estimated by computing the phase differencebetween the 119868 and 119876 signals

Estimated parameters would be slightly different fromthe real value due to the noise in the signal but it will beadequate to correct the 119876 signal based on the 119868 signal Fromthe results shown in Figure 2(a)(E) the corrected 119876 signal issimilar to the simulated noiseless signal (Figure 2(a)(A)) ofthe amplitude and the phase offset The same approach was


20 30

(A)

0 10

05

10

Am

plitu

de

Simulated noiseless signal

minus5

t (s)

(C)

0 10 20 30

05

10

Am

plitu

de

Simulated noisy signal(amplitude and phase imbalance)

minus5

t (s)

(E)

I

Q

0 10 20 30

05

10

Am

plitu

de

Simulated corrected signal

minus5

t (s)

(B)

0 5 10

minus505

10 Complex noiseless signal

minus10minus10 minus5

I

Q

(D)

0 5 10

05

10Complex noisy signal

minus10

minus5

minus5minus10

I

Q

(F)

0 5 10

05

10 Corrected complex signal

minus10

minus5

minus10 minus5I

Q

(a) 119868119876 amplitude and phase imbalance correction simulation

Respiration signal3

2

1

0

0 5 10 15 20 25 30 35

minus1

minus2

Am

plitu

de

t (s)

Respiration beltQ signal from Doppler RadarCorrected Q signal from Doppler Radar

(b) Comparison of respiration belt signal versus Doppler Radar signal

Figure 2 119868119876 imbalance simulation and results evaluation

used with the real data and subsequently compared with therespiration belt signalThe corrected119876 signal is slightly betterthan the uncorrected119876 signal as the mean squared errors areldquo0041651rdquo and ldquo0050928rdquo respectively (see Figure 2(b)) Forfurther evaluation on the Doppler Radar signals comparedto the reference respiration belt five data sets (a minute ofrecording for each data set) were collected from the subject(random breathing) where the mean square error (MSE) andcorrelation coefficient were computed Results are shown inTable 1 and we notice good correlations obtained between theDoppler signals and the respiratory belt signals

Table 1 Quantitative evaluation of Doppler Radar signal withreference respiration belt

Data set Mean square error Correlation coefficient1 0017 09682 0094 09383 0009 09654 0005 09425 0015 0975

Table 2 Polynomial modelling and DTW performance evaluation

(a) Polynomial order evaluation

Order Inhalation ExhalationRMSE Corr RMSE Corr

1 214119864 minus 03 09912 204119864 minus 03 099182 202119864 minus 03 09921 203119864 minus 03 099193 315119864 minus 04 09998 539119864 minus 05 099994 197119864 minus 17 1 102119864 minus 17 15 326119864 minus 17 1 136119864 minus 17 1

(b) Performance evaluation of random breathing component with selectedmodel

Breathingcomponent

Polynomialmodel MSE Corr Class

Fastinhalation

Normal 111119890 minus 04 0933 FastFast 428119890 minus 06 0989

Normalinhalation

Normal 223119890 minus 06 0999 NormalFast 837119890 minus 05 0954

Fastexhalation


Normalexhalation


For the decomposition of the breathing signal intoinhalation and exhalation components it is necessary tocalculate the transition time of each breathing componentindependent of the breathing amplitude In addition to thispreliminary measurements were obtained from a voluntarysubject (asthmatic) to understand if there were any detectablebreathing pattern differences in his breathing compared tonormal patterns see Figures 3 and 6 In the future as anextension to this current work more trials will be performedwith more subjects particularly with different breathingconditions for further analysis This paper is a preliminaryexercise to convey the correlation of Doppler Radar withclinically used chest strap devices

4 Results

The results were based on choosing the 119868119876 baseband signalclosest to the optimum point [37] and best matched with


0 5 10 15 20 25 30

0

01

(a)

minus01Am

plitu

de

t (s)

Raw Qres signal

0 50 100 150

0

005

(b)

minus005Am

plitu

de

t (s)

0 5 10 15 20 25 30

0002

(c)

minus002Am

plitu

de

t (s)

SG filteringFourier filtering

Qres signal after filtering

0 02 04 06 08 1 120

02

04

Frequency (Hz)(d)

Am

plitu

de

Single-sided amplitude spectrum of Qres

X 009155

Qres signal after piecewise fitting with window length 200ms

5 10 15 20 25 30

0

005

(a)

minus005Am

plitu

de

t (s)

Raw Qres signal

20 40 60 80 100 120 140

0

002

(b)

minus002Am

plitu

de

t (s)

5 10 15 20 25 30

0

002

(c)

minus002Am

plitu

det (s)

SG filteredFourier filtered


0 02 04 06 08 1 120

02

04

Frequency (Hz)(d)

Am

plitu

de


X 01221


Figure 3 Breathing pattern from voluntary asthmatic subject (data set 1 and data set 2)

the independent breathing measurement Only a portion ofobservations were displayed in this paper where the outputconsisted ofDoppler Radar-basedmeasurements for differenttypes of inhalation and exhalation patterns collected over aspecific period of time

41 Normal Breathing Normal adult breathing rates rangefrom 12 to 20 cycles (inhalation exhalation and pause) in aminute [4] Figure 6(a) represents the normal breathingpattern It can be seen that for a period of 20 seconds therewere 5 breaths which corresponded to 025Hz (asymp15 breathsper minute) and the FFT of the signal shows a constant peakat 02441Hz or 14646 breaths per minute The patterns andextracted rate correlated with the independent breathingcounts

42 Fast Breathing Rapid breathing is typically defined asabove 20 breaths per minute for resting adults and thisis called Tachypnea [38] In this experiment our aim wasto establish if breathing at different rates can be detectedrobustly and the feasibility of subsequent classification

Figure 6(b) represents the fast breathing pattern with differ-ent dynamics Here the subject was inhaling and exhalingat a faster rate resulting in a shorter breathing cycle Resultsshow the occurrence of 12 breathing cycles in a period of20 seconds (36 per minute) The FFT also shows a peak at06104Hz corresponding to 366 breaths per minute similarto the independent breathing cycle counts

43 Slow Inhalation-Fast Exhalation Wemimic another typeof breathing scenario where the inhalation is slower than theexhalation rate Data was collected for a period of 10 secondsand from Figure 6(c) a longer inhalation time (marked ingreen box) and a shorter exhalation time (marked in redbox) are evident This is as expected as the subject inhalesslowly and exhales at a faster pace Results show that therewere two clear breathing cycles in a period of 10 secondsObserved results show an average of 25 1 for the I E ratiowhere the FFT computation approximated the breathing rateto be 1465 (02441Hz) breaths per minute and the expectedbreathing rate was 12 breaths per minute from independentmeasurements For these particular experiments an average


of 25 seconds was required for inhalation compared to theone second needed for exhalation

44 Fast Inhalation-Slow Exhalation Figure 6(d) shows thesignal representation for fast inhalation and slow exhalationMeasurements clearly show that two breathing cycles with anaverage of 1 25 I E ratio occurred The breathing rate wasexpected to be 12 breaths per minute and from the FFTthe breathing rate was estimated as 1465 (02441Hz) breathsper minute Results from both observations clearly showthat the exhalation is longer than inhalation Both the casesdiscussed in Sections 43 and 44 further prove that therespiration rate alone is not adequate in describing therespiratory activities of the subjects A more descriptiveinformation could be obtained through the breathing cycledecomposition approach from the noncontactDoppler Radarmeasurement

5 Discussions

Results in Section 4 have demonstrated the feasibility ofDoppler Radar in capturing various types of breathingdynamics and this section further discusses the importance ofbreathing cycle analysis decomposition and identification

51 Possible Abnormal Breathing Patterns It is clear that sim-ply recording breathing frequencies measured as a angularfrequency using spectralmethods is inadequate for analysingasymmetric breathing patterns [23] albeit useful for extract-ing the fundamental cycle for breathing periodsThe evidenceso far is that decomposing the breathing cycle into its inhala-tion and exhalation components offers a more accurate andinsightful approach to detecting and interpreting breathingand can be performed reliably using Doppler Radar In thisparticular experiment the breathing pattern of a voluntarysubject (age 23 height 180 cm and weight 95 kg) who hasasthma was collected within the duration of 30 seconds butnot during an asthma attack Results are shown in Figure 3Notice the inhalation component (marked in the greencolour box) is of a shorter duration compared to the exha-lation component (marked in the red color box) where theapproximated I E ratio for that subject is 1 25 Both theresults showed a longer duration recorded for exhalationcompared to inhalation where the implications are such thatthe subject could be having difficulties in exhaling [39] andthis enforces the value in the analysis by decomposition

In future work experiments from Sections 41ndash51 will beextended with an increased number of subjects (normal andabnormal) in a clinical trial to further support the qualitativeand quantitative evaluations This can facilitate finding amore accurate and insightful way to describe the respiratoryfunctions using a noncontact form of measurements Fur-thermore additional analysis could be performed includingthe amplitude variation and the shape of each decomposedbreathing component pertaining to different types of subjectsFor instance amplitude variation in the voluntary subjectwith asthma was observed to be lesser than that of the subjectwith normal breathing Consideration on respiratory effort

breathing patterns and other related factors (eg respiratoryfunction such as tidal volume) would be an essential study inthe future in evaluating the potential use of Doppler Radar inrespiratory researchwhich includes sensing detections anal-ysis and qualitative assertions

52 Breathing Component Decomposition Although a com-plete breathing cycle comprises of inhalation and exhalationshort and even long pauses can also exist between these statesdepending on the regularity of breathing and other factorssuch as the need for oxygen surrounding environmentand so forth A long pause for instance of more than 10seconds [40] is defined as an abnormal event and is known asapnoea relevant for detecting sleep apnoea and even SIDSBreathing patterns can also potentially be used together withthe analysis of tidal volume [24] to diagnose other aspects ofbreathing problems such as shallow breathing and the capa-bility in detecting apnoea These have been reported in [15]using microwave Doppler Radar

The main purpose of decomposing the breathing cyclesis to gain useful information of the breathing activity Forinstance an abnormal breathing rate of 8 breathsmin couldbe analysed with more information such as inhalation andexhalation rates and so forth This can be particularly usefulwhen it could be used in the early diagnosis of specificbreathing conditions or in a pulmonary rehabilitation [41ndash43] especially if it could be performed in a noncontact form

Each of the inhalation and the exhalation componentswas extracted to obtain the polynomial coefficients fromnormal and fast breathing data respectively and resultsindicate that a fourth-order RMSE (root mean square error)and Corr (correlation coefficient) polynomial were sufficientto fit these components (eg randomly chosen inhalation andexhalation component) as shown in the Table 2(a) Subse-quently using the same approach the computed fourth-orderpolynomial model was used to characterise two differenttypes of inhalation and exhalation breathing components(normal and fast) This model was then used to identify theexperimental breathing scenario as discussed in Section 532

53 Analysis of the Breathing Component

531 I E Ratio Analysis The ratio between the inhalationor exhalation components was computed from the averagetime duration in considerations of the entire set Using thecollected data there were 15 fast and 7 normal componentsextracted from the data sets and the ratios of each ofthe components (in comparison with the average time ofrespective inhalationexhalation components) are shown inFigure 4 It was seen that there were two distinct groupscorresponding to two different breathing dynamics in twodifferent events where this could not be estimated from therespiration rate estimation (spectral analysis)

532 Dynamic Time Warping and Evaluation by CorrelationThe time duration for complete inhalation and exhala-tion components varies between individuals and situationsTherefore in order to summarise characterise compare and


35

3

25

2

15

1

05

0 5 10 15 20

Ratio

Inhalation component

Inhalation ratio between fast breathing and normal breathing

Normal inhalation componentFast inhalation componentAverage computed model for inhalation

Ratio

Exhalation component

Exhalation ratio between fast breathing and normal breathing

25

2

15

1

05

0 5 10 15 20

Normal exhalation componentFast exhalation componentAverage computed model for exhalation

Figure 4 Ratio of breathing component

interpret breathing patterns a number of alternatives can beconsidered In our experiments these include

(i) extraction of inhalation and exhalation componentsbased on normal and fast breathing criteria

(ii) computation of fourth-order polynomials model foreach breathing condition (normal and fast) from theextracted components respectively

(iii) using dynamic time warping to find the optimalalignment between the predefined model from (ii)and the randomly picked breathing component

(iv) using the correlationmethod to identify the similarityof the aligned results from (iii) for identification andcomputing the MSE between the curves

Two different polynomials for inhalation and exhalationin normal and fast breathingweremodelled from the data sets(procedure (i)-(ii)) For validation dynamic time warpingwas performed between randomly chosen components (anydata set) with the model based on polynomial representation(procedure (iii)-(iv))

The purpose of performing this experiment was to usethe derivedmodel as a reference and to classify each breathingcomponent based on two different classes In brief by deriv-ing a model based on the rate of breathing we can in fact


004

003

002

001

0

minus002

minus001

minus003

minus004

004

003

002

001

0

minus002

minus001

minus003

minus004

200 400 600 800 1000 200 400 600 800 1000 500 1000 1500500 1000 1500

Time Time Time Time

Am

plitu

deOriginal signals

Fast inhale polynomial modelRandom normal inhale component

(A) Normal inhale component with fast inhale model (B) Normal inhale component with normal inhale model

Warped signals Original signals Warped signals

DTW fast inhale polynomial modelDTW random normalinhale component

Normal inhale polynomial modelRandom normal inhale component

DTW normal inhalepolynomial modelDTW random normal inhale component

(a) DTW of normal inhalation component with respective inhalation model

0015

001

0005

0

minus001

minus0005

minus0015

minus002

200 600 1000 200 600 1000 1400

Time200 400 600 800 1000 500 1000 1500

Time TimeTime

Am

plitu

de

Original signals Warped signals Original signals Warped signals

003

002

001

0

minus002

minus001

minus003

Fast exhale polynomial modelRandom fast exhale component

(A) Fast exhale component with fast exhale model (B) Fast exhale component with normal exhale model

DTW fast exhale polynomial modelDTW random fastexhale component

Normal exhale polynomial modelRandom fast exhale component

DTW normal exhalepolynomial modelDTW random fast exhale component

(b) DTW of fast exhalation component with respective exhalation model

Figure 5 DTW evaluation

identify and correlate the extracted breathing componentswith the derived model to distinguish different respiratoryclasses For validation purposes the experiments were per-formed as follows

(a) fast inhalation component with normal and fastinhalation model

(b) normal inhalation component with normal and fastinhalation model

(c) fast exhalation component with normal and fastexhalation model

(d) normal exhalation component with normal and fastexhalation model

Each of the breathing components was randomly pickedfrom the data sets It was then evaluated and represented interms of mean square error (MSE) and correlation coefficient(Corr) as shown in Table 2(b) For graphical representationas an example we associate ldquonormal inhalation componentwith normal and fast inhalation modelrdquo and ldquofast exhalationcomponent with normal and fast exhalation modelrdquo and theresults were shown in ldquoFigures 5(a) and 5(b)rdquo respectively

6 Conclusions

In this paper we have demonstrated the feasibility of breath-ing detection under varying conditions using Doppler RadarWe have shown that noninvasive breathing detection using


0 2 4 6 8 10 12 14 16 18 20

0

01

Am

plitu

de

minus01

t (s)

0 10 20 30 40 50 60 70 80 90 100

0

005

Am

plitu

de

minus005

t (s)

0 2 4 6 8 10 12 14 16 18 20

0005

Am

plitu

de

SG filteredFouriern filtered

minus005

t (s)

Am

plitu

de

0 02 04 06 08 1 120

05

1

Frequency (Hz)

X 02441

(A) Raw Qres signal

(C) Qres signal after filtering

(D) Single-sided amplitude spectrum of Qres

(B) Qres signal after piecewise fitting with window length 200ms

(a) Normal breathing

0 02 04 06 08 1 12 14 16 18 20

05

1

Frequency (Hz)

Am

plitu

de

X 06104

0 2 4 6 8 10 12 14 16 18 20

0

05

Am

plitu

de

minus05

t (s)

0 10 20 30 40 50 60 70 80 90 100

0

005

Am

plitu

de

minus005

t (s)

0 2 4 6 8 10 12 14 16 18 20

001

Am

plitu

de


t (s)

minus01

(A) Raw Qres signal




(b) Fast breathing

0 1 2 3 4 5 6 7 8 9 10

0

02

Am

plitu

de

minus02

t (s)

0 5 10 15 20 25 30 35 40 45 50

0

005

Am

plitu

de

minus005

t (s)

Am

plitu

de

0 02 04 06 08 1 12 14 16 18 20

05

1

Frequency (Hz)

X 02441

0 1 2 3 4 5 6 7 8 9 10

0

005

Am

plitu

de

minus005

t (s)SG filteredFourier filtered

(B) Ires signal after piece wise fitting with window length 200ms

(C) Ires signal after filtering

Raw Ires signal(A)

Single-sided amplitude spectrum of Ires(D)

(c) Slow inhalation-fast exhalation

0 1 2 3 4 5 6 7 8 9 10

0

05

minus05Am

plitu

de

t (s)

0 5 10 15 20 25 30 35 40 45 50

0

01

minus01Am

plitu

de

t (s)

Am

plitu

de

0 02 04 06 08 1 12 14 16 18 20

1

2

Frequency (Hz)

X 02441

0 1 2 3 4 5 6 7 8 9 10

0

01

minus01Am

plitu

de




Raw Ires signal(A)


(d) Fast inhalation-slow exhalation

Figure 6 Doppler Radar signals from various type of breathing scenarios


Doppler Radar could potentially be used to detect differenttypes of breathing patterns such as rapid breathing and slowbreathing We have also demonstrated that by decomposingthe respiratory cycle into inhalation pause and exhalation itis possible to extract additional information on the breathingactivities For this purpose we proposed a fourth-orderpolynomial to represent each atomic component of breathingand demonstrated the use of DTW in classifying breathingcomponent independently into the corresponding class Inthe derived model each component is associated to a specificbreathing scenario which in particular is fast and normalbreathing Regarding future work experimental trials willbe extended with more subjects as well as improved signalprocessing techniques (eg isolation of motion artefacts andmore robust model based filtering techniques) breathingcomponent modelling and classification techniques

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

This work was supported by Australian Federal and VictoriaState Governments and the Australian Research Councilthrough the ICT Centre of Excellence program National ICTAustralia (NICTA)

References

[1] J Boyle N Bidargaddi A Sarela and M Karunanithi ldquoAuto-matic detection of respiration rate from ambulatory single-lead ECGrdquo IEEE Transactions on Information Technology inBiomedicine vol 13 no 6 pp 890ndash896 2009

[2] H Gibson ldquoA form of behaviour therapy for some states diag-nosed as affective disorderrdquo Behaviour Research and Therapyvol 16 no 3 pp 191ndash195 1978

[3] P Grossman ldquoRespiration stress and cardiovascular functionrdquoPsychophysiology vol 20 no 3 pp 284ndash300 1983

[4] G Yuan N A Drost and R A McIvor ldquoRespiratory rate andbreathing patternrdquoMcMasterUniversityMedical Journal vol 10pp 23ndash28 2013

[5] SMondini andCGuilleminault ldquoAbnormal breathing patternsduring sleep in diabetesrdquo Annals of Neurology vol 17 no 4 pp391ndash395 1985

[6] H Corning Mosbys PDQ for Respiratory CaremdashRevisedReprint ElsevierHealth Sciences 2012 httpbooksgooglecomaubooksid=hYgfvCdwa3sC

[7] Y Munjal S Sharma M A K Agarwal and P Gupta Api Text-book ofMedicine SeriesG Reference Information and Interdis-ciplinary Subjects Series Jaypee Brothers Medical Publishers2012 httpbooksgooglecomaubooksid=L7pW3yGjj7kC

[8] L Stead and S Thomas Emergency Medicine Board ReviewSeries LippincottampWilliams 2000 httpbooksgooglecomaubooksid=lmTpnSGEYwwC

[9] B Aehlert and R Vroman Paramedic Practice Today Above andBeyond vol 2 Jones amp Bartlett Learning 2011 httpbooksgooglecomaubooksid=gA3mcImmXbAC

[10] KNakajima T Tamura andHMiike ldquoMonitoring of heart andrespiratory rates by photoplethysmography using a digitalfiltering techniquerdquoMedical Engineering and Physics vol 18 no5 pp 365ndash372 1996

[11] D Girbau A Lazaro A Ramos and R Villarino ldquoRemotesensing of vital signs using a doppler radar and diversity toovercome null detectionrdquo IEEE Sensors Journal vol 12 no 3pp 512ndash518 2012

[12] J H Oum D-W Kim and S Hong ldquoTwo frequency radar sen-sor for non-contact vital signal monitorrdquo in Proceedings of theIEEE MTT-S International Microwave Symposium Digest (MTTrsquo08) pp 919ndash922 June 2008

[13] W Xu C Gu C Li andM Sarrafzadeh ldquoRobust Doppler radardemodulation via compressed sensingrdquo Electronics Letters vol48 no 22 pp 1428ndash1430 2012

[14] N Birsan D-P Munteanu G Iubu and T Niculescu ldquoTime-frequency analysis in Doppler radar for noncontact cardiopul-monary monitoringrdquo in Proceedings of the E-Health and Bio-engineering Conference (EHB rsquo11) pp 1ndash4 November 2011

[15] Y S Lee P N Pathirana T Caelli and S Li ldquoFurther applica-tions of Doppler radar for non-contact respiratory assessmentrdquoin Proceedings of the 35th Annual International Conference of theIEEE Engineering in Medicine and Biology Society (EMBC rsquo13)pp 3833ndash3836 Osaka Japan July 2013

[16] Y S Lee P N Pathirana T Caelli and R Evans ldquoDopplerradar in respiratory monitoring detection and analysisrdquo inProceedings of the 2nd International Conference on ControlAutomation and Information Sciences (ICCAIS rsquo13) pp 224ndash228November 2013

[17] S Suzuki T Matsui H Kawahara et al ldquoA non-contact vitalsign monitoring system for ambulances using dual-frequencymicrowave radarsrdquo Medical and Biological Engineering andComputing vol 47 no 1 pp 101ndash105 2009

[18] S Suzuki T Matsui H Imuta et al ldquoA novel autonomicactivation measurement method for stress monitoring Non-contact measurement of heart rate variability using a compactmicrowave radarrdquoMedical and Biological Engineering and Com-puting vol 46 no 7 pp 709ndash714 2008

[19] O Boric-Lubecke V M Lubecke A Host-Madsen DSamardzija and K Cheung ldquoDoppler radar sensing of multiplesubjects in single and multiple antenna systemsrdquo in Proceedingsof the 7th International Conference on Telecommunications inModerm Satellite Cable and Broadcasting Services (TELSIKSrsquo05) vol 1 pp 7ndash11 September 2005

[20] A Tariq and H Ghafouri-Shiraz ldquoVital signs detection usingdoppler radar and continuous wavelet transformrdquo in Pro-ceedings of the 5th European Conference on Antennas andPropagation (EUCAP rsquo11) pp 285ndash288 April 2011

[21] A Abushakra M Faezipour and A Abumunshar ldquoEfficientfrequency-based classification of respiratory movementsrdquo inProceedings of the IEEE International Conference on Elec-troInformation Technology (EIT rsquo12) pp 1ndash5 May 2012

[22] D G E Criner and J Gerard Critical Care Study GuideSpringer New York NY USA 2002

[23] R P Dellinger and J E Parrillo Critical Care MedicinePrinciples of Diagnosis and Management in the Adult ElsevierHealth Sciences 2007

[24] W Massagram V M Lubecke and O Boric-LubeckeldquoMicrowave non-invasive sensing of respiratory tidal volumerdquoin Proceedings of the Annual International Conference of theIEEE Engineering in Medicine and Biology Society (EMBC rsquo09)pp 4832ndash4835 September 2009


[25] Y S Lee P N Pathirana C L Steinfort and T Caelli ldquoMon-itoring and analysis of respiratory patterns using microwaveDoppler radarrdquo IEEE Journal of Translational Engineering inHealth and Medicine vol 2 pp 1ndash12 2014

[26] P N Pathirana S C Herath and A V Savkin ldquoMultitargettracking via space transformations using a single frequencycontinuous wave radarrdquo IEEE Transactions on Signal Processingvol 60 no 10 pp 5217ndash5229 2012

[27] GW Stimson Introduction to Airborne Radar Scitech Publish-ing San Francisco 2nd edition 1998

[28] A D Droitcour Non-Contact Measurement of Heart and Res-piration Rates with a Single-Chip Microwave Doppler RadarStanford University 2006

[29] R V Lenth ldquoOn a form of piecewise linear regressionrdquo TheAmerican Statistician vol 29 no 3 pp 116ndash117 1975

[30] R W Schafer ldquoWhat is a savitzky-golay filter [Lecture Notes]rdquoIEEE Signal Processing Magazine vol 28 no 4 pp 111ndash117 2011

[31] T OHaver ldquoInteractive fourier filterrdquo September 2006 httpwwwmathworkscommatlabcentralfileexchange12377

[32] B-K Park S Yamada and V Lubecke ldquoMeasurement methodfor imbalance factors in direct-conversion quadrature radarsystemsrdquo IEEEMicrowave andWireless Components Letters vol17 no 5 pp 403ndash405 2007

[33] XHuang ldquoOn transmitter gainphase imbalance compensationat receiverrdquo IEEECommunications Letters vol 4 no 11 pp 363ndash365 2000

[34] O Steila ldquoAutomatic in-phase quadrature balancingrdquo 2006httpwwwqslnetik1xpvdsppdfaiqbenpdf

[35] S Salvador and P Chan ldquoToward accurate dynamic time warp-ing in linear time and spacerdquo Intelligent Data Analysis vol 11no 5 pp 561ndash580 2007 httpdlacmorgcitationcfm id=13679851367993

[36] E J Keogh and M J Pazzani ldquoDerivative dynamic time warp-ingrdquo in Proceedings of the 1st SIAM International Conference onData Mining (SDM rsquo01) 2001

[37] O Boric-Lubecke V M Lubecke I Mostafanezhad B-K ParkW Massagram and B Jokanovic ldquoDoppler radar architecturesand signal processing for heart rate extractionrdquo MikrotalasnaRevija pp 12ndash17 2009

[38] Mosby Mosbys Medical Dictionary Elsevier New York NYUSA 2009

[39] University of Virginia School of Medicine AsthmaattacksUniversity of Virginia School of Medicine 2011 httpwwwmedicinevirginiaeduclinicaldepartmentspediatricsclinical-servicestutorialsasthmaattacks

[40] T Al-Ani C K Karmakar A H Khandoker and MPalaniswami ldquoAutomatic recognition of obstructive sleepapnoea syndrome using power spectral analysis of electrocar-diogram and hidden markov modelsrdquo in Proceedings of theInternational Conference on Intelligent Sensors Sensor Networksand Information Processing (ISSNIP rsquo08) pp 285ndash290 Decem-ber 2008

[41] B M Cappo and D S Holmes ldquoThe utility of prolonged res-piratory exhalation for reducing physiologial and psychologicalarousal in non-threatening and threatening situationsrdquo Journalof Psychosomatic Research vol 28 no 4 pp 265ndash273 1984

[42] A Klintworth Z Ajtay A Paljunite S Szabados and L HejjelldquoHeart rate asymmetry follows the inspirationexpiration ratioin healthy volunteersrdquo Physiological Measurement vol 33 no10 article 1717 2012

[43] I M Lin L Y Tai and S Y Fan ldquoBreathing at a rate of 55breaths per minute with equal inhalation-to-exhalation ratioincreases heart rate variabilityrdquo International Journal of Psy-chophysiology vol 91 no 3 pp 206ndash211 2014

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014


Active and Passive Electronic Components

Control Scienceand Engineering

Journal of



RotatingMachinery


Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design



Shock and Vibration


Civil EngineeringAdvances in

Acoustics and VibrationAdvances in



Electrical and Computer Engineering

Journal of

Advances inOptoElectronics


Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of


Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014


Chemical EngineeringInternational Journal of Antennas and

Propagation




Navigation and Observation



DistributedSensor Networks



the chest wall displacement and the tidal volume Differenttypes of breathing can potentially be deduced from such chestwall or abdomendisplacement information during inhalationand exhalation This information can be used to identifydifferent types of breathing signatures such as shallow breath-ing deep breathing slow breathing fast breathing and othertypes of breathing patterns Indeed the displacement of thechest wall or abdomen in shallow breathing is expected tobe small and the complete breathing cycle would occur in ashorter time period compared to normal breathing

Doppler Radar operates by transmitting a radio wavesignal and receiving the modulated version of the signaldue to the motion triggered by the target [24 26] Thereflected wave is in the modulated form where it undergoes afrequency shift proportional to the radial velocity that can bedescribed using theDoppler effectWhen a target has a quasi-periodic motion the time varying position of the target canbe represented as a phase modulated signal and the phaseshift is directly proportional to the objectrsquos movement Thusthe movement of the chest wallabdomen for respiration dueto the inhalation exhalation and the pause states can bedetected and modelled using the reflected Doppler shiftedsignal the main focus of this paper We provide a com-prehensive description of the noncontact respiratory mea-surement via Doppler Radar which was then validated withindependent measurements using a respiration belt andbreathing cycle counts We also demonstrate the differenttypes of inhaling and exhaling states from data collectedusing our Doppler Radar systemThe purpose of this paper issummarized as follows

(i) investigation of Doppler Radarrsquos feasibility in captur-ing different types of breathing patterns under variousbreathing scenarios

(ii) correction of 119868119876 signal imbalance and cross-validation of Doppler breathing signal with stan-dard respiration measurement the respiration belt(MLT1132 iezo-respiratory belt transducer)

(iii) decomposition of the breathing signal (from DopplerRadar) into its respective inhalation and exhalationcomponents representing each component modelusing 4th polynomial fitting (see Table 2(a)) andclassifying decomposed breathing components intoits respective breathing scenarios

2 Methods

21 Respiration Monitoring via Microwave Doppler RadarThe Doppler effect occurs when there is a shift in thefrequency of the wave either reflected or radiated received byan object in motion [27] Consider a transmitted sine wavesignal with an angular frequency 120596

0

119879119909= sin (120596

0119905 + 1206010) (1)

where 119879119909is the transmitted signal 119905 is the time and 120601

0is the

arbitrary phase shift Assume that the target is stationary at adistance of 119903

0from the radar and the transmission time from

radar to target is 1199030119888where 119888 is thewave propagation velocity

The target range at time 119905 is given by equation below 119903(119905) =

1199030+ 119903(119905 minus 119905

0) where 119903 is the range of the target from the radar

and 119903 (velocity) is the rate of change of 119903 and 1199050is the time at

119903 = 1199030 The received signal at the stationary target is the same

as the transmitted signal at the time 1199030119888 which can be given

as

119877target = sin(1205960119905 minus

12059601199030

119888+ 1206010) (2)

The received signal from the target at time 119905 would havebeen sent Δ119905 seconds prior to time 119905 This can be representedas Δ119905 = 2119903

0119888 Referring to (1) signal can be depicted in the

same formulation given as

119877119909= sin (120596

0(119905 minus Δ119905) + 120601

0) (3)

Substituting Δ119905 = 21199030119888 into (3) the received signal is further

represented as

119877119909= sin(120596

0119905 minus

212059601199030

119888+ 1206010) (4)

For a target moving (radially) with respect to the radar thedistance will vary and by using 119903(119905) = 119903

0+ 119903(119905 minus 119905

0) and 120596

119889=

21205960119903119888 the received signal can be further derived as

119877119909= sin(120596

0(119905 minus

2119903 (119905)

119888) + 1206010)

= sin(1199080(119905 minus

21199030

119888minus2 119903 (119905 minus 119905

0)

119888) + 120601

0)

= sin(1205960(1 minus

2 119903

119888) 119905 minus

21205960

119888(1199030minus 1199031199050) + 1206010)

= sin((1205960minus 120596119889) 119905 minus

212059601199030

119888+ 1205961198891199050+ 1206010)

(5)

where the frequency of the reflected signal is shifted by 120596119889

and the phase angle by 1205961198891199050 Therefore the Doppler shift 120596

119889

can also be denoted by 120596119889= 2120587119891

119889 where 119891

119889= 2 119903119891

0119888 is the

Doppler shift in Hertz and 1198910is the transmitted frequency

Using 120582 = 1198881198910 119891119889can be written as 119891

119889= minus2 119903120582 where

the negative sign accounts for the fact that if 119903 is negative(when the target is approaching) the Doppler frequency willbe positive or vice versa [27] From (5) the phase angle Φ ofthe received signal is given as120596

1198891199050Therefore the transmitted

wave from the radar to the target will be reflected to thereceiver with some phase shifting and can be represented asphase modulation given as

Φ =21205960119903

1198881199050=4120587 (119903)

120582 (6)

The measurement model for human respiration usingDoppler Radar can be derived as follows Generally theDoppler shift in frequency is given by

119891119889 (119905) =

2119891V (119905)119888

=2V (119905)120582

(7)

where V(119905) is the velocity of the target 120582 is the wavelengthof the transmitted signal and 119888 is velocity of the propagating



119903(119905) given by Φ

119903(119905) =


119879 (119905) = cos (1205960119905 + 1206010(119905)) (8)








represented as

119877 (119905) = cos [1205960(119905 minus 119905119889) + 120601 (119905 minus 119905

119889)]

= cos [1205960(119905 minus

21198890+ 2119909 (119905 minus 119889 (119905) 119888)

119888)

+ 120601(119905 minus21198890+ 2119909 (119905 minus 119889 (119905) 119888)

119888)]

(9)


119877 (119905) asymp cos(2120587119891119905 minus41205871198890

120582minus4120587119909 (119905)

120582+ 120601(119905 minus

21198890

119888)) (10)


119861 (119905) = cos(120579 + 4120587119909 (119905)

120582+ Δ120601 (119905)) (11)


119861(119905)) and a



119868119861(119905) = cos(120579 + 4120587119909 (119905)

120582+ Δ120601 (119905))

119876119861 (119905) = sin(120579 + 4120587119909 (119905)

120582+ Δ120601 (119905))

(12)






119909(119905) is defined by

119877119909 (119905) = 119883119868 cos (1199080119905) + 119883119876 sin (1199080119905) (13)

where 119883119868and 119883



1198771015840

119909(119905) = 119877

119909(119905) lowast cos (119908

0119905) + 119877

119909(119905) lowast 119860

119890lowast sin (119908

0119905 + 120601)

(14)



1198681015840= 119877119909(119905) lowast cos (119908

0119905)

1198761015840= 119877119909(119905) lowast 119860

119890lowast sin (119908

0119905 + 120601)

(15)



1198681015840= 119883119868cos (119908

0119905) cos (119908

0119905) + 119883

119876sin (119908

0119905) cos (119908

0119905)

1198761015840= 119883119868cos (119908

0119905)

lowast 119860119890(sin (119908

0119905) cos (120601) + cos (119908

0119905) sin (120601))

+ 119883119876sin (119908

0119905)

lowast 119860119890(sin (119908

0119905) cos (120601) + cos (119908

0119905) sin (120601))

(16)


119868and119883


[

119883119868

119883119876

] =[[

[

1 0


119860119890cos (120601)

]]

]

[

1198681015840

1198761015840] (17)













NI‐DAQ

AD8347

AD

MATLAB

I Q

Local oscillator

AD8347

(IQ demodulator)


Matlab environment

Tx and Rx

Tx

Rx




Fourier filter



Polynomial modeling


Filtering stage


Raw data (IQ)












20 30

(A)

0 10

05

10

Am

plitu

de


minus5

t (s)

(C)

0 10 20 30

05

10

Am

plitu

de


minus5

t (s)

(E)

I

Q

0 10 20 30

05

10

Am

plitu

de


minus5

t (s)

(B)

0 5 10

minus505



I

Q

(D)

0 5 10

05


minus10

minus5

minus5minus10

I

Q

(F)

0 5 10

05


minus10

minus5

minus10 minus5I

Q


Respiration signal3

2

1

0

0 5 10 15 20 25 30 35

minus1

minus2

Am

plitu

de

t (s)












Breathingcomponent


Fastinhalation


Normalinhalation


Fastexhalation


Normalexhalation



4 Results



0 5 10 15 20 25 30

0

01

(a)

minus01Am

plitu

de

t (s)

Raw Qres signal

0 50 100 150

0

005

(b)

minus005Am

plitu

de

t (s)

0 5 10 15 20 25 30

0002

(c)

minus002Am

plitu

de

t (s)



0 02 04 06 08 1 120

02

04

Frequency (Hz)(d)

Am

plitu

de


X 009155


5 10 15 20 25 30

0

005

(a)

minus005Am

plitu

de

t (s)

Raw Qres signal

20 40 60 80 100 120 140

0

002

(b)

minus002Am

plitu

de

t (s)

5 10 15 20 25 30

0

002

(c)

minus002Am

plitu

det (s)



0 02 04 06 08 1 120

02

04

Frequency (Hz)(d)

Am

plitu

de


X 01221











5 Discussions












35

3

25

2

15

1

05

0 5 10 15 20

Ratio




Ratio



25

2

15

1

05

0 5 10 15 20











004

003

002

001

0

minus002

minus001

minus003

minus004

004

003

002

001

0

minus002

minus001

minus003

minus004

200 400 600 800 1000 200 400 600 800 1000 500 1000 1500500 1000 1500

Time Time Time Time

Am

plitu

deOriginal signals








0015

001

0005

0

minus001

minus0005

minus0015

minus002

200 600 1000 200 600 1000 1400

Time200 400 600 800 1000 500 1000 1500

Time TimeTime

Am

plitu

de


003

002

001

0

minus002

minus001

minus003














6 Conclusions



0 2 4 6 8 10 12 14 16 18 20

0

01

Am

plitu

de

minus01

t (s)

0 10 20 30 40 50 60 70 80 90 100

0

005

Am

plitu

de

minus005

t (s)

0 2 4 6 8 10 12 14 16 18 20

0005

Am

plitu

de


minus005

t (s)

Am

plitu

de

0 02 04 06 08 1 120

05

1

Frequency (Hz)

X 02441

(A) Raw Qres signal





0 02 04 06 08 1 12 14 16 18 20

05

1

Frequency (Hz)

Am

plitu

de

X 06104

0 2 4 6 8 10 12 14 16 18 20

0

05

Am

plitu

de

minus05

t (s)

0 10 20 30 40 50 60 70 80 90 100

0

005

Am

plitu

de

minus005

t (s)

0 2 4 6 8 10 12 14 16 18 20

001

Am

plitu

de


t (s)

minus01

(A) Raw Qres signal




(b) Fast breathing

0 1 2 3 4 5 6 7 8 9 10

0

02

Am

plitu

de

minus02

t (s)

0 5 10 15 20 25 30 35 40 45 50

0

005

Am

plitu

de

minus005

t (s)

Am

plitu

de

0 02 04 06 08 1 12 14 16 18 20

05

1

Frequency (Hz)

X 02441

0 1 2 3 4 5 6 7 8 9 10

0

005

Am

plitu

de

minus005




Raw Ires signal(A)



0 1 2 3 4 5 6 7 8 9 10

0

05

minus05Am

plitu

de

t (s)

0 5 10 15 20 25 30 35 40 45 50

0

01

minus01Am

plitu

de

t (s)

Am

plitu

de

0 02 04 06 08 1 12 14 16 18 20

1

2

Frequency (Hz)

X 02441

0 1 2 3 4 5 6 7 8 9 10

0

01

minus01Am

plitu

de




Raw Ires signal(A)








Acknowledgments


References















































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation











119903(119905) given by Φ

119903(119905) =


119879 (119905) = cos (1205960119905 + 1206010(119905)) (8)








represented as

119877 (119905) = cos [1205960(119905 minus 119905119889) + 120601 (119905 minus 119905

119889)]

= cos [1205960(119905 minus

21198890+ 2119909 (119905 minus 119889 (119905) 119888)

119888)

+ 120601(119905 minus21198890+ 2119909 (119905 minus 119889 (119905) 119888)

119888)]

(9)


119877 (119905) asymp cos(2120587119891119905 minus41205871198890

120582minus4120587119909 (119905)

120582+ 120601(119905 minus

21198890

119888)) (10)


119861 (119905) = cos(120579 + 4120587119909 (119905)

120582+ Δ120601 (119905)) (11)


119861(119905)) and a



119868119861(119905) = cos(120579 + 4120587119909 (119905)

120582+ Δ120601 (119905))

119876119861 (119905) = sin(120579 + 4120587119909 (119905)

120582+ Δ120601 (119905))

(12)






119909(119905) is defined by

119877119909 (119905) = 119883119868 cos (1199080119905) + 119883119876 sin (1199080119905) (13)

where 119883119868and 119883



1198771015840

119909(119905) = 119877

119909(119905) lowast cos (119908

0119905) + 119877

119909(119905) lowast 119860

119890lowast sin (119908

0119905 + 120601)

(14)



1198681015840= 119877119909(119905) lowast cos (119908

0119905)

1198761015840= 119877119909(119905) lowast 119860

119890lowast sin (119908

0119905 + 120601)

(15)



1198681015840= 119883119868cos (119908

0119905) cos (119908

0119905) + 119883

119876sin (119908

0119905) cos (119908

0119905)

1198761015840= 119883119868cos (119908

0119905)

lowast 119860119890(sin (119908

0119905) cos (120601) + cos (119908

0119905) sin (120601))

+ 119883119876sin (119908

0119905)

lowast 119860119890(sin (119908

0119905) cos (120601) + cos (119908

0119905) sin (120601))

(16)


119868and119883


[

119883119868

119883119876

] =[[

[

1 0


119860119890cos (120601)

]]

]

[

1198681015840

1198761015840] (17)













NI‐DAQ

AD8347

AD

MATLAB

I Q

Local oscillator

AD8347

(IQ demodulator)


Matlab environment

Tx and Rx

Tx

Rx




Fourier filter



Polynomial modeling


Filtering stage


Raw data (IQ)












20 30

(A)

0 10

05

10

Am

plitu

de


minus5

t (s)

(C)

0 10 20 30

05

10

Am

plitu

de


minus5

t (s)

(E)

I

Q

0 10 20 30

05

10

Am

plitu

de


minus5

t (s)

(B)

0 5 10

minus505



I

Q

(D)

0 5 10

05


minus10

minus5

minus5minus10

I

Q

(F)

0 5 10

05


minus10

minus5

minus10 minus5I

Q


Respiration signal3

2

1

0

0 5 10 15 20 25 30 35

minus1

minus2

Am

plitu

de

t (s)












Breathingcomponent


Fastinhalation


Normalinhalation


Fastexhalation


Normalexhalation



4 Results



0 5 10 15 20 25 30

0

01

(a)

minus01Am

plitu

de

t (s)

Raw Qres signal

0 50 100 150

0

005

(b)

minus005Am

plitu

de

t (s)

0 5 10 15 20 25 30

0002

(c)

minus002Am

plitu

de

t (s)



0 02 04 06 08 1 120

02

04

Frequency (Hz)(d)

Am

plitu

de


X 009155


5 10 15 20 25 30

0

005

(a)

minus005Am

plitu

de

t (s)

Raw Qres signal

20 40 60 80 100 120 140

0

002

(b)

minus002Am

plitu

de

t (s)

5 10 15 20 25 30

0

002

(c)

minus002Am

plitu

det (s)



0 02 04 06 08 1 120

02

04

Frequency (Hz)(d)

Am

plitu

de


X 01221











5 Discussions












35

3

25

2

15

1

05

0 5 10 15 20

Ratio




Ratio



25

2

15

1

05

0 5 10 15 20











004

003

002

001

0

minus002

minus001

minus003

minus004

004

003

002

001

0

minus002

minus001

minus003

minus004

200 400 600 800 1000 200 400 600 800 1000 500 1000 1500500 1000 1500

Time Time Time Time

Am

plitu

deOriginal signals








0015

001

0005

0

minus001

minus0005

minus0015

minus002

200 600 1000 200 600 1000 1400

Time200 400 600 800 1000 500 1000 1500

Time TimeTime

Am

plitu

de


003

002

001

0

minus002

minus001

minus003














6 Conclusions



0 2 4 6 8 10 12 14 16 18 20

0

01

Am

plitu

de

minus01

t (s)

0 10 20 30 40 50 60 70 80 90 100

0

005

Am

plitu

de

minus005

t (s)

0 2 4 6 8 10 12 14 16 18 20

0005

Am

plitu

de


minus005

t (s)

Am

plitu

de

0 02 04 06 08 1 120

05

1

Frequency (Hz)

X 02441

(A) Raw Qres signal





0 02 04 06 08 1 12 14 16 18 20

05

1

Frequency (Hz)

Am

plitu

de

X 06104

0 2 4 6 8 10 12 14 16 18 20

0

05

Am

plitu

de

minus05

t (s)

0 10 20 30 40 50 60 70 80 90 100

0

005

Am

plitu

de

minus005

t (s)

0 2 4 6 8 10 12 14 16 18 20

001

Am

plitu

de


t (s)

minus01

(A) Raw Qres signal




(b) Fast breathing

0 1 2 3 4 5 6 7 8 9 10

0

02

Am

plitu

de

minus02

t (s)

0 5 10 15 20 25 30 35 40 45 50

0

005

Am

plitu

de

minus005

t (s)

Am

plitu

de

0 02 04 06 08 1 12 14 16 18 20

05

1

Frequency (Hz)

X 02441

0 1 2 3 4 5 6 7 8 9 10

0

005

Am

plitu

de

minus005




Raw Ires signal(A)



0 1 2 3 4 5 6 7 8 9 10

0

05

minus05Am

plitu

de

t (s)

0 5 10 15 20 25 30 35 40 45 50

0

01

minus01Am

plitu

de

t (s)

Am

plitu

de

0 02 04 06 08 1 12 14 16 18 20

1

2

Frequency (Hz)

X 02441

0 1 2 3 4 5 6 7 8 9 10

0

01

minus01Am

plitu

de




Raw Ires signal(A)








Acknowledgments


References















































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation











1198681015840= 119883119868cos (119908

0119905) cos (119908

0119905) + 119883

119876sin (119908

0119905) cos (119908

0119905)

1198761015840= 119883119868cos (119908

0119905)

lowast 119860119890(sin (119908

0119905) cos (120601) + cos (119908

0119905) sin (120601))

+ 119883119876sin (119908

0119905)

lowast 119860119890(sin (119908

0119905) cos (120601) + cos (119908

0119905) sin (120601))

(16)


119868and119883


[

119883119868

119883119876

] =[[

[

1 0


119860119890cos (120601)

]]

]

[

1198681015840

1198761015840] (17)













NI‐DAQ

AD8347

AD

MATLAB

I Q

Local oscillator

AD8347

(IQ demodulator)


Matlab environment

Tx and Rx

Tx

Rx




Fourier filter



Polynomial modeling


Filtering stage


Raw data (IQ)












20 30

(A)

0 10

05

10

Am

plitu

de


minus5

t (s)

(C)

0 10 20 30

05

10

Am

plitu

de


minus5

t (s)

(E)

I

Q

0 10 20 30

05

10

Am

plitu

de


minus5

t (s)

(B)

0 5 10

minus505



I

Q

(D)

0 5 10

05


minus10

minus5

minus5minus10

I

Q

(F)

0 5 10

05


minus10

minus5

minus10 minus5I

Q


Respiration signal3

2

1

0

0 5 10 15 20 25 30 35

minus1

minus2

Am

plitu

de

t (s)












Breathingcomponent


Fastinhalation


Normalinhalation


Fastexhalation


Normalexhalation



4 Results



0 5 10 15 20 25 30

0

01

(a)

minus01Am

plitu

de

t (s)

Raw Qres signal

0 50 100 150

0

005

(b)

minus005Am

plitu

de

t (s)

0 5 10 15 20 25 30

0002

(c)

minus002Am

plitu

de

t (s)



0 02 04 06 08 1 120

02

04

Frequency (Hz)(d)

Am

plitu

de


X 009155


5 10 15 20 25 30

0

005

(a)

minus005Am

plitu

de

t (s)

Raw Qres signal

20 40 60 80 100 120 140

0

002

(b)

minus002Am

plitu

de

t (s)

5 10 15 20 25 30

0

002

(c)

minus002Am

plitu

det (s)



0 02 04 06 08 1 120

02

04

Frequency (Hz)(d)

Am

plitu

de


X 01221











5 Discussions












35

3

25

2

15

1

05

0 5 10 15 20

Ratio




Ratio



25

2

15

1

05

0 5 10 15 20











004

003

002

001

0

minus002

minus001

minus003

minus004

004

003

002

001

0

minus002

minus001

minus003

minus004

200 400 600 800 1000 200 400 600 800 1000 500 1000 1500500 1000 1500

Time Time Time Time

Am

plitu

deOriginal signals








0015

001

0005

0

minus001

minus0005

minus0015

minus002

200 600 1000 200 600 1000 1400

Time200 400 600 800 1000 500 1000 1500

Time TimeTime

Am

plitu

de


003

002

001

0

minus002

minus001

minus003














6 Conclusions



0 2 4 6 8 10 12 14 16 18 20

0

01

Am

plitu

de

minus01

t (s)

0 10 20 30 40 50 60 70 80 90 100

0

005

Am

plitu

de

minus005

t (s)

0 2 4 6 8 10 12 14 16 18 20

0005

Am

plitu

de


minus005

t (s)

Am

plitu

de

0 02 04 06 08 1 120

05

1

Frequency (Hz)

X 02441

(A) Raw Qres signal





0 02 04 06 08 1 12 14 16 18 20

05

1

Frequency (Hz)

Am

plitu

de

X 06104

0 2 4 6 8 10 12 14 16 18 20

0

05

Am

plitu

de

minus05

t (s)

0 10 20 30 40 50 60 70 80 90 100

0

005

Am

plitu

de

minus005

t (s)

0 2 4 6 8 10 12 14 16 18 20

001

Am

plitu

de


t (s)

minus01

(A) Raw Qres signal




(b) Fast breathing

0 1 2 3 4 5 6 7 8 9 10

0

02

Am

plitu

de

minus02

t (s)

0 5 10 15 20 25 30 35 40 45 50

0

005

Am

plitu

de

minus005

t (s)

Am

plitu

de

0 02 04 06 08 1 12 14 16 18 20

05

1

Frequency (Hz)

X 02441

0 1 2 3 4 5 6 7 8 9 10

0

005

Am

plitu

de

minus005




Raw Ires signal(A)



0 1 2 3 4 5 6 7 8 9 10

0

05

minus05Am

plitu

de

t (s)

0 5 10 15 20 25 30 35 40 45 50

0

01

minus01Am

plitu

de

t (s)

Am

plitu

de

0 02 04 06 08 1 12 14 16 18 20

1

2

Frequency (Hz)

X 02441

0 1 2 3 4 5 6 7 8 9 10

0

01

minus01Am

plitu

de




Raw Ires signal(A)








Acknowledgments


References















































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










NI‐DAQ

AD8347

AD

MATLAB

I Q

Local oscillator

AD8347

(IQ demodulator)


Matlab environment

Tx and Rx

Tx

Rx




Fourier filter



Polynomial modeling


Filtering stage


Raw data (IQ)












20 30

(A)

0 10

05

10

Am

plitu

de


minus5

t (s)

(C)

0 10 20 30

05

10

Am

plitu

de


minus5

t (s)

(E)

I

Q

0 10 20 30

05

10

Am

plitu

de


minus5

t (s)

(B)

0 5 10

minus505



I

Q

(D)

0 5 10

05


minus10

minus5

minus5minus10

I

Q

(F)

0 5 10

05


minus10

minus5

minus10 minus5I

Q


Respiration signal3

2

1

0

0 5 10 15 20 25 30 35

minus1

minus2

Am

plitu

de

t (s)












Breathingcomponent


Fastinhalation


Normalinhalation


Fastexhalation


Normalexhalation



4 Results



0 5 10 15 20 25 30

0

01

(a)

minus01Am

plitu

de

t (s)

Raw Qres signal

0 50 100 150

0

005

(b)

minus005Am

plitu

de

t (s)

0 5 10 15 20 25 30

0002

(c)

minus002Am

plitu

de

t (s)



0 02 04 06 08 1 120

02

04

Frequency (Hz)(d)

Am

plitu

de


X 009155


5 10 15 20 25 30

0

005

(a)

minus005Am

plitu

de

t (s)

Raw Qres signal

20 40 60 80 100 120 140

0

002

(b)

minus002Am

plitu

de

t (s)

5 10 15 20 25 30

0

002

(c)

minus002Am

plitu

det (s)



0 02 04 06 08 1 120

02

04

Frequency (Hz)(d)

Am

plitu

de


X 01221











5 Discussions












35

3

25

2

15

1

05

0 5 10 15 20

Ratio




Ratio



25

2

15

1

05

0 5 10 15 20











004

003

002

001

0

minus002

minus001

minus003

minus004

004

003

002

001

0

minus002

minus001

minus003

minus004

200 400 600 800 1000 200 400 600 800 1000 500 1000 1500500 1000 1500

Time Time Time Time

Am

plitu

deOriginal signals








0015

001

0005

0

minus001

minus0005

minus0015

minus002

200 600 1000 200 600 1000 1400

Time200 400 600 800 1000 500 1000 1500

Time TimeTime

Am

plitu

de


003

002

001

0

minus002

minus001

minus003














6 Conclusions



0 2 4 6 8 10 12 14 16 18 20

0

01

Am

plitu

de

minus01

t (s)

0 10 20 30 40 50 60 70 80 90 100

0

005

Am

plitu

de

minus005

t (s)

0 2 4 6 8 10 12 14 16 18 20

0005

Am

plitu

de


minus005

t (s)

Am

plitu

de

0 02 04 06 08 1 120

05

1

Frequency (Hz)

X 02441

(A) Raw Qres signal





0 02 04 06 08 1 12 14 16 18 20

05

1

Frequency (Hz)

Am

plitu

de

X 06104

0 2 4 6 8 10 12 14 16 18 20

0

05

Am

plitu

de

minus05

t (s)

0 10 20 30 40 50 60 70 80 90 100

0

005

Am

plitu

de

minus005

t (s)

0 2 4 6 8 10 12 14 16 18 20

001

Am

plitu

de


t (s)

minus01

(A) Raw Qres signal




(b) Fast breathing

0 1 2 3 4 5 6 7 8 9 10

0

02

Am

plitu

de

minus02

t (s)

0 5 10 15 20 25 30 35 40 45 50

0

005

Am

plitu

de

minus005

t (s)

Am

plitu

de

0 02 04 06 08 1 12 14 16 18 20

05

1

Frequency (Hz)

X 02441

0 1 2 3 4 5 6 7 8 9 10

0

005

Am

plitu

de

minus005




Raw Ires signal(A)



0 1 2 3 4 5 6 7 8 9 10

0

05

minus05Am

plitu

de

t (s)

0 5 10 15 20 25 30 35 40 45 50

0

01

minus01Am

plitu

de

t (s)

Am

plitu

de

0 02 04 06 08 1 12 14 16 18 20

1

2

Frequency (Hz)

X 02441

0 1 2 3 4 5 6 7 8 9 10

0

01

minus01Am

plitu

de




Raw Ires signal(A)








Acknowledgments


References















































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










20 30

(A)

0 10

05

10

Am

plitu

de


minus5

t (s)

(C)

0 10 20 30

05

10

Am

plitu

de


minus5

t (s)

(E)

I

Q

0 10 20 30

05

10

Am

plitu

de


minus5

t (s)

(B)

0 5 10

minus505



I

Q

(D)

0 5 10

05


minus10

minus5

minus5minus10

I

Q

(F)

0 5 10

05


minus10

minus5

minus10 minus5I

Q


Respiration signal3

2

1

0

0 5 10 15 20 25 30 35

minus1

minus2

Am

plitu

de

t (s)












Breathingcomponent


Fastinhalation


Normalinhalation


Fastexhalation


Normalexhalation



4 Results



0 5 10 15 20 25 30

0

01

(a)

minus01Am

plitu

de

t (s)

Raw Qres signal

0 50 100 150

0

005

(b)

minus005Am

plitu

de

t (s)

0 5 10 15 20 25 30

0002

(c)

minus002Am

plitu

de

t (s)



0 02 04 06 08 1 120

02

04

Frequency (Hz)(d)

Am

plitu

de


X 009155


5 10 15 20 25 30

0

005

(a)

minus005Am

plitu

de

t (s)

Raw Qres signal

20 40 60 80 100 120 140

0

002

(b)

minus002Am

plitu

de

t (s)

5 10 15 20 25 30

0

002

(c)

minus002Am

plitu

det (s)



0 02 04 06 08 1 120

02

04

Frequency (Hz)(d)

Am

plitu

de


X 01221











5 Discussions












35

3

25

2

15

1

05

0 5 10 15 20

Ratio




Ratio



25

2

15

1

05

0 5 10 15 20











004

003

002

001

0

minus002

minus001

minus003

minus004

004

003

002

001

0

minus002

minus001

minus003

minus004

200 400 600 800 1000 200 400 600 800 1000 500 1000 1500500 1000 1500

Time Time Time Time

Am

plitu

deOriginal signals








0015

001

0005

0

minus001

minus0005

minus0015

minus002

200 600 1000 200 600 1000 1400

Time200 400 600 800 1000 500 1000 1500

Time TimeTime

Am

plitu

de


003

002

001

0

minus002

minus001

minus003














6 Conclusions



0 2 4 6 8 10 12 14 16 18 20

0

01

Am

plitu

de

minus01

t (s)

0 10 20 30 40 50 60 70 80 90 100

0

005

Am

plitu

de

minus005

t (s)

0 2 4 6 8 10 12 14 16 18 20

0005

Am

plitu

de


minus005

t (s)

Am

plitu

de

0 02 04 06 08 1 120

05

1

Frequency (Hz)

X 02441

(A) Raw Qres signal





0 02 04 06 08 1 12 14 16 18 20

05

1

Frequency (Hz)

Am

plitu

de

X 06104

0 2 4 6 8 10 12 14 16 18 20

0

05

Am

plitu

de

minus05

t (s)

0 10 20 30 40 50 60 70 80 90 100

0

005

Am

plitu

de

minus005

t (s)

0 2 4 6 8 10 12 14 16 18 20

001

Am

plitu

de


t (s)

minus01

(A) Raw Qres signal




(b) Fast breathing

0 1 2 3 4 5 6 7 8 9 10

0

02

Am

plitu

de

minus02

t (s)

0 5 10 15 20 25 30 35 40 45 50

0

005

Am

plitu

de

minus005

t (s)

Am

plitu

de

0 02 04 06 08 1 12 14 16 18 20

05

1

Frequency (Hz)

X 02441

0 1 2 3 4 5 6 7 8 9 10

0

005

Am

plitu

de

minus005




Raw Ires signal(A)



0 1 2 3 4 5 6 7 8 9 10

0

05

minus05Am

plitu

de

t (s)

0 5 10 15 20 25 30 35 40 45 50

0

01

minus01Am

plitu

de

t (s)

Am

plitu

de

0 02 04 06 08 1 12 14 16 18 20

1

2

Frequency (Hz)

X 02441

0 1 2 3 4 5 6 7 8 9 10

0

01

minus01Am

plitu

de




Raw Ires signal(A)








Acknowledgments


References















































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










0 5 10 15 20 25 30

0

01

(a)

minus01Am

plitu

de

t (s)

Raw Qres signal

0 50 100 150

0

005

(b)

minus005Am

plitu

de

t (s)

0 5 10 15 20 25 30

0002

(c)

minus002Am

plitu

de

t (s)



0 02 04 06 08 1 120

02

04

Frequency (Hz)(d)

Am

plitu

de


X 009155


5 10 15 20 25 30

0

005

(a)

minus005Am

plitu

de

t (s)

Raw Qres signal

20 40 60 80 100 120 140

0

002

(b)

minus002Am

plitu

de

t (s)

5 10 15 20 25 30

0

002

(c)

minus002Am

plitu

det (s)



0 02 04 06 08 1 120

02

04

Frequency (Hz)(d)

Am

plitu

de


X 01221











5 Discussions












35

3

25

2

15

1

05

0 5 10 15 20

Ratio




Ratio



25

2

15

1

05

0 5 10 15 20











004

003

002

001

0

minus002

minus001

minus003

minus004

004

003

002

001

0

minus002

minus001

minus003

minus004

200 400 600 800 1000 200 400 600 800 1000 500 1000 1500500 1000 1500

Time Time Time Time

Am

plitu

deOriginal signals








0015

001

0005

0

minus001

minus0005

minus0015

minus002

200 600 1000 200 600 1000 1400

Time200 400 600 800 1000 500 1000 1500

Time TimeTime

Am

plitu

de


003

002

001

0

minus002

minus001

minus003














6 Conclusions



0 2 4 6 8 10 12 14 16 18 20

0

01

Am

plitu

de

minus01

t (s)

0 10 20 30 40 50 60 70 80 90 100

0

005

Am

plitu

de

minus005

t (s)

0 2 4 6 8 10 12 14 16 18 20

0005

Am

plitu

de


minus005

t (s)

Am

plitu

de

0 02 04 06 08 1 120

05

1

Frequency (Hz)

X 02441

(A) Raw Qres signal





0 02 04 06 08 1 12 14 16 18 20

05

1

Frequency (Hz)

Am

plitu

de

X 06104

0 2 4 6 8 10 12 14 16 18 20

0

05

Am

plitu

de

minus05

t (s)

0 10 20 30 40 50 60 70 80 90 100

0

005

Am

plitu

de

minus005

t (s)

0 2 4 6 8 10 12 14 16 18 20

001

Am

plitu

de


t (s)

minus01

(A) Raw Qres signal




(b) Fast breathing

0 1 2 3 4 5 6 7 8 9 10

0

02

Am

plitu

de

minus02

t (s)

0 5 10 15 20 25 30 35 40 45 50

0

005

Am

plitu

de

minus005

t (s)

Am

plitu

de

0 02 04 06 08 1 12 14 16 18 20

05

1

Frequency (Hz)

X 02441

0 1 2 3 4 5 6 7 8 9 10

0

005

Am

plitu

de

minus005




Raw Ires signal(A)



0 1 2 3 4 5 6 7 8 9 10

0

05

minus05Am

plitu

de

t (s)

0 5 10 15 20 25 30 35 40 45 50

0

01

minus01Am

plitu

de

t (s)

Am

plitu

de

0 02 04 06 08 1 12 14 16 18 20

1

2

Frequency (Hz)

X 02441

0 1 2 3 4 5 6 7 8 9 10

0

01

minus01Am

plitu

de




Raw Ires signal(A)








Acknowledgments


References















































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation












5 Discussions












35

3

25

2

15

1

05

0 5 10 15 20

Ratio




Ratio



25

2

15

1

05

0 5 10 15 20











004

003

002

001

0

minus002

minus001

minus003

minus004

004

003

002

001

0

minus002

minus001

minus003

minus004

200 400 600 800 1000 200 400 600 800 1000 500 1000 1500500 1000 1500

Time Time Time Time

Am

plitu

deOriginal signals








0015

001

0005

0

minus001

minus0005

minus0015

minus002

200 600 1000 200 600 1000 1400

Time200 400 600 800 1000 500 1000 1500

Time TimeTime

Am

plitu

de


003

002

001

0

minus002

minus001

minus003














6 Conclusions



0 2 4 6 8 10 12 14 16 18 20

0

01

Am

plitu

de

minus01

t (s)

0 10 20 30 40 50 60 70 80 90 100

0

005

Am

plitu

de

minus005

t (s)

0 2 4 6 8 10 12 14 16 18 20

0005

Am

plitu

de


minus005

t (s)

Am

plitu

de

0 02 04 06 08 1 120

05

1

Frequency (Hz)

X 02441

(A) Raw Qres signal





0 02 04 06 08 1 12 14 16 18 20

05

1

Frequency (Hz)

Am

plitu

de

X 06104

0 2 4 6 8 10 12 14 16 18 20

0

05

Am

plitu

de

minus05

t (s)

0 10 20 30 40 50 60 70 80 90 100

0

005

Am

plitu

de

minus005

t (s)

0 2 4 6 8 10 12 14 16 18 20

001

Am

plitu

de


t (s)

minus01

(A) Raw Qres signal




(b) Fast breathing

0 1 2 3 4 5 6 7 8 9 10

0

02

Am

plitu

de

minus02

t (s)

0 5 10 15 20 25 30 35 40 45 50

0

005

Am

plitu

de

minus005

t (s)

Am

plitu

de

0 02 04 06 08 1 12 14 16 18 20

05

1

Frequency (Hz)

X 02441

0 1 2 3 4 5 6 7 8 9 10

0

005

Am

plitu

de

minus005




Raw Ires signal(A)



0 1 2 3 4 5 6 7 8 9 10

0

05

minus05Am

plitu

de

t (s)

0 5 10 15 20 25 30 35 40 45 50

0

01

minus01Am

plitu

de

t (s)

Am

plitu

de

0 02 04 06 08 1 12 14 16 18 20

1

2

Frequency (Hz)

X 02441

0 1 2 3 4 5 6 7 8 9 10

0

01

minus01Am

plitu

de




Raw Ires signal(A)








Acknowledgments


References















































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










35

3

25

2

15

1

05

0 5 10 15 20

Ratio




Ratio



25

2

15

1

05

0 5 10 15 20











004

003

002

001

0

minus002

minus001

minus003

minus004

004

003

002

001

0

minus002

minus001

minus003

minus004

200 400 600 800 1000 200 400 600 800 1000 500 1000 1500500 1000 1500

Time Time Time Time

Am

plitu

deOriginal signals








0015

001

0005

0

minus001

minus0005

minus0015

minus002

200 600 1000 200 600 1000 1400

Time200 400 600 800 1000 500 1000 1500

Time TimeTime

Am

plitu

de


003

002

001

0

minus002

minus001

minus003














6 Conclusions



0 2 4 6 8 10 12 14 16 18 20

0

01

Am

plitu

de

minus01

t (s)

0 10 20 30 40 50 60 70 80 90 100

0

005

Am

plitu

de

minus005

t (s)

0 2 4 6 8 10 12 14 16 18 20

0005

Am

plitu

de


minus005

t (s)

Am

plitu

de

0 02 04 06 08 1 120

05

1

Frequency (Hz)

X 02441

(A) Raw Qres signal





0 02 04 06 08 1 12 14 16 18 20

05

1

Frequency (Hz)

Am

plitu

de

X 06104

0 2 4 6 8 10 12 14 16 18 20

0

05

Am

plitu

de

minus05

t (s)

0 10 20 30 40 50 60 70 80 90 100

0

005

Am

plitu

de

minus005

t (s)

0 2 4 6 8 10 12 14 16 18 20

001

Am

plitu

de


t (s)

minus01

(A) Raw Qres signal




(b) Fast breathing

0 1 2 3 4 5 6 7 8 9 10

0

02

Am

plitu

de

minus02

t (s)

0 5 10 15 20 25 30 35 40 45 50

0

005

Am

plitu

de

minus005

t (s)

Am

plitu

de

0 02 04 06 08 1 12 14 16 18 20

05

1

Frequency (Hz)

X 02441

0 1 2 3 4 5 6 7 8 9 10

0

005

Am

plitu

de

minus005




Raw Ires signal(A)



0 1 2 3 4 5 6 7 8 9 10

0

05

minus05Am

plitu

de

t (s)

0 5 10 15 20 25 30 35 40 45 50

0

01

minus01Am

plitu

de

t (s)

Am

plitu

de

0 02 04 06 08 1 12 14 16 18 20

1

2

Frequency (Hz)

X 02441

0 1 2 3 4 5 6 7 8 9 10

0

01

minus01Am

plitu

de




Raw Ires signal(A)








Acknowledgments


References















































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










004

003

002

001

0

minus002

minus001

minus003

minus004

004

003

002

001

0

minus002

minus001

minus003

minus004

200 400 600 800 1000 200 400 600 800 1000 500 1000 1500500 1000 1500

Time Time Time Time

Am

plitu

deOriginal signals








0015

001

0005

0

minus001

minus0005

minus0015

minus002

200 600 1000 200 600 1000 1400

Time200 400 600 800 1000 500 1000 1500

Time TimeTime

Am

plitu

de


003

002

001

0

minus002

minus001

minus003














6 Conclusions



0 2 4 6 8 10 12 14 16 18 20

0

01

Am

plitu

de

minus01

t (s)

0 10 20 30 40 50 60 70 80 90 100

0

005

Am

plitu

de

minus005

t (s)

0 2 4 6 8 10 12 14 16 18 20

0005

Am

plitu

de


minus005

t (s)

Am

plitu

de

0 02 04 06 08 1 120

05

1

Frequency (Hz)

X 02441

(A) Raw Qres signal





0 02 04 06 08 1 12 14 16 18 20

05

1

Frequency (Hz)

Am

plitu

de

X 06104

0 2 4 6 8 10 12 14 16 18 20

0

05

Am

plitu

de

minus05

t (s)

0 10 20 30 40 50 60 70 80 90 100

0

005

Am

plitu

de

minus005

t (s)

0 2 4 6 8 10 12 14 16 18 20

001

Am

plitu

de


t (s)

minus01

(A) Raw Qres signal




(b) Fast breathing

0 1 2 3 4 5 6 7 8 9 10

0

02

Am

plitu

de

minus02

t (s)

0 5 10 15 20 25 30 35 40 45 50

0

005

Am

plitu

de

minus005

t (s)

Am

plitu

de

0 02 04 06 08 1 12 14 16 18 20

05

1

Frequency (Hz)

X 02441

0 1 2 3 4 5 6 7 8 9 10

0

005

Am

plitu

de

minus005




Raw Ires signal(A)



0 1 2 3 4 5 6 7 8 9 10

0

05

minus05Am

plitu

de

t (s)

0 5 10 15 20 25 30 35 40 45 50

0

01

minus01Am

plitu

de

t (s)

Am

plitu

de

0 02 04 06 08 1 12 14 16 18 20

1

2

Frequency (Hz)

X 02441

0 1 2 3 4 5 6 7 8 9 10

0

01

minus01Am

plitu

de




Raw Ires signal(A)








Acknowledgments


References















































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










0 2 4 6 8 10 12 14 16 18 20

0

01

Am

plitu

de

minus01

t (s)

0 10 20 30 40 50 60 70 80 90 100

0

005

Am

plitu

de

minus005

t (s)

0 2 4 6 8 10 12 14 16 18 20

0005

Am

plitu

de


minus005

t (s)

Am

plitu

de

0 02 04 06 08 1 120

05

1

Frequency (Hz)

X 02441

(A) Raw Qres signal





0 02 04 06 08 1 12 14 16 18 20

05

1

Frequency (Hz)

Am

plitu

de

X 06104

0 2 4 6 8 10 12 14 16 18 20

0

05

Am

plitu

de

minus05

t (s)

0 10 20 30 40 50 60 70 80 90 100

0

005

Am

plitu

de

minus005

t (s)

0 2 4 6 8 10 12 14 16 18 20

001

Am

plitu

de


t (s)

minus01

(A) Raw Qres signal




(b) Fast breathing

0 1 2 3 4 5 6 7 8 9 10

0

02

Am

plitu

de

minus02

t (s)

0 5 10 15 20 25 30 35 40 45 50

0

005

Am

plitu

de

minus005

t (s)

Am

plitu

de

0 02 04 06 08 1 12 14 16 18 20

05

1

Frequency (Hz)

X 02441

0 1 2 3 4 5 6 7 8 9 10

0

005

Am

plitu

de

minus005




Raw Ires signal(A)



0 1 2 3 4 5 6 7 8 9 10

0

05

minus05Am

plitu

de

t (s)

0 5 10 15 20 25 30 35 40 45 50

0

01

minus01Am

plitu

de

t (s)

Am

plitu

de

0 02 04 06 08 1 12 14 16 18 20

1

2

Frequency (Hz)

X 02441

0 1 2 3 4 5 6 7 8 9 10

0

01

minus01Am

plitu

de




Raw Ires signal(A)








Acknowledgments


References















































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation













Acknowledgments


References















































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation









Research Article Noncontact Detection and Analysis...

Documents

Transcript of Research Article Noncontact Detection and Analysis...