StudyonTrippingRisksinFastWalkingthrough Cadence...

Research ArticleStudy on Tripping Risks in Fast Walking throughCadence-Controlled Gait Analysis

Wen-Fong Wang1 Wei-Chih Lien 2 Che-Yu Liu1 and Ching-Yu Yang 3

1Department of Computer Science and Information Engineering National Yunlin University of Science and TechnologyDouliu Yunlin 640 Taiwan2Department of Physical Medicine and Rehabilitation National Cheng Kung University Hospital College of MedicineNational Cheng Kung University Tainan 600 Taiwan3Department of Computer Science and Information Engineering National Penghu University of Science and TechnologyMagong Penghu 880 Taiwan

Correspondence should be addressed to Wei-Chih Lien lwclwhabms8hinetnet

Received 20 October 2017 Accepted 5 April 2018 Published 24 May 2018

Academic Editor Giedrius Vanagas

Copyright copy 2018 Wen-Fong Wang et al +is is an open access article distributed under the Creative Commons AttributionLicense which permits unrestricted use distribution and reproduction in any medium provided the original work isproperly cited

Fast walking is a common exercise for most people to promote health However a higher cadence due to fast walking on ordinaryor uneven ground raises the risk of tripping To investigate the tripping issue research to observe the gait in fast walking is neededTo explore the relationship between fast gait and the risk of tripping a gait recording system with a specific synchronizationmechanism was developed in this work +e system can acquire gait signals from wearable sensors and action cameras at differentcadences Meanwhile algorithms for gait cycle segmentation and characteristic extraction were proposed for analyzing a fast gaitIn the gait analysis the correlations of low moderate and high cadence in cueing and no cueing gaits were computed and tworesults were obtained First the higher the cadence is the larger the motion strength in the terminal foot swing will be and thesmaller the motion strength at the starting foot swing Second the decreased distance of foot clearance becomesmore conspicuousas the cadence increased especially if one is walking more than 120 beats +e results indicate that fast walking with bigger stridesand lower cadence is the best way to maintain safety in moving over ordinary ground

1 Introduction

Walking is an important daily activity and many people dothat to maintain healthy condition Movements in walkingcan be affected by the rhythmic sounds which make themovements smoother and more stable [1 2] In walkingmany people use the music tempo to regulate their cadenceand breathing rate to keep their exercises more lastingAdditionally they also conduct fast walking to increase theirexercise intensity for calorie consumption In terms ofmedical treatment many patients that have Parkinsonrsquosdisease [3] children with cerebral palsy [4] and manypatients that have strokes [5] and gait rehabilitation [6] usemetronomes to help their exercises

In the literature rhythm on gait has two research topicswhich are in health promotion andmedical applications For

the health promotion some studies reveal that the musictempo affects gait movements and benefits human healthconditions In [7] Mendonca et al studied the music playedat the treadmill to regulate the walking pace and analyze thegait parameters +eir experiment included walking withand without music cueing synchronization and the resultsshowed that when there was music cueing synchronizationpeople would change their walking step with the musictempo When there was no music cueing synchronizationpeople would not change their walking step with the musictempo obviously In [8] Eikema et al found that old peoplersquosgait parameters such as the stride and step reduced dra-matically and their walking speed could be changed usingthe voice cue to regulate their pace on treadmills After thetraining on walking in cue their gait and strides were im-proved obviously to be closer to young people

HindawiJournal of Healthcare EngineeringVolume 2018 Article ID 2723178 11 pageshttpsdoiorg10115520182723178

In medical applications acoustic tempos are frequentlyused in gait rehabilitation with positive instantaneous andprolonged transfer eects on various gait characteristics In[9] rhythmic auditory cues including music andmetronomebeats were used to improve disordered gait in clinicalconditionseir study indicated that music andmetronomecues may not be used interchangeably and the cue type andfrequency (cadence) should be considered when evaluatingeects of rhythmic auditory cues on gait In [10] the gaitpattern in various rhythmic auditory stimulation temposwas investigated in stroke patients e results showed thatas the tempo increased the spatiotemporal gait parametersof the stroke patients changed signicantly After gaittraining with rhythmic auditory stimulation gait parametersof the aected and unaected sides improved signicantlycompared to baseline

In the previous studies treadmills were often used forstep and pace control and thus the conducted experimentswere often restricted in space [5 7 8 10] erefore theresearch results were biased due to few data from daily walkIn recent years the advancement of the wearable technologyhas made sensors miniaturized and widely applied in manyresearch and applications [11] As the wearable sensorsovercome space restrictions and improve the degree offreedom in use these devices can be used to replace xed andexpensive measuring apparatuses [12] In this study wedeveloped a gait recording system which equipped with a setof wearable triaxial acceleration sensors and action camerase sensors are in miniature size and can be easily worn bysubjects to acquire gait signals from daily walk e actioncameras can be used to record gait movements With thissystem gait signals for daily walking can be acquired withoutspace restrictions

In [13] Rubenstein described that the risks of falls andtrips are due to some identiable factors such as weaknessunsteady gait confusion and certain medications and si-multaneously indicated that attention to the factors cangreatly reduce the rates of falling He also described that oneof the most eective fall reduction programs involve envi-ronmental inspection and hazard reduction In [14] Lordet al indicated that the most challenging area to incorporateinto fall aetiology is that of environmental risk factors sincethe research base is smaller than that for other types of riskfactors At the same time research in this area has typicallybeen less rigorous or has been troubled by methodologicallimitations Since a gait recording system with the wearabletechnology has been developed in this work to improve themethodological limitations we are interested in exploringthe tripping risks in fast walking through an uneven groundwhich may exist with a few environmental risk factors bya cadence-controlled strategy With the application of actioncameras and a specic synchronization mechanism betweenthe signals from the wearable sensors and the images fromthe cameras we realize the gait analysis of fast walkingwithout space restrictions

is investigation is organized as follows Section 2presents research materials including gait description gaitsignal acquisition cadence control and a signal synchro-nization mechanism Section 3 describes research methods

including the experimental protocol signal preprocessingsignal segmentation and characteristic extraction andcorrelation analysis Section 4 shows the results which in-clude correlation analysis in cadence versus gait events andperiods with and without cueing In addition analyses of gaitphase variation and gait periods versus walking types are alsodone for exploring the tripping risks Concluding remarksand future work are given in Section 5

2 Research Materials

21 Gait Description Human walking consists of a numberof cyclically sequential gait events and one complete gaitcycle (Figure 1) can be measured from one toe o event tothe next one [15] As shown in Figure 1 a full gait cycleincludes a stance phase and a swing phase where 60 of thecycle is spent in stance and 40 in swing [16]

e gait cycle has seven gait events that is initial contact(IC) opposite toe o (OTO) heel rise (HR) opposite initialcontact (OIC) toe o (TO) feet adjacent (FA) and tibiavertical (TV) [17] It can also be subdivided into seven periodsthat is loading response midstance terminal stance preswinginitial swing midswing and terminal swing While nelysubdividing gait events and periods for every cycle dierentgait parameters such as the cycle duration stance time swingtime stride length step length cadence and walking speed canbe analyzed to learn about personal gait features

22 System of Signal Acquisition In this study a wearablesensor module with a triaxial accelerometer shown in Figure 2was developed to acquire motion signals related to humangait e size of this sensor is about 50times 30mm2 and itincludes amicroprocessor (TIMSP430trade series [18]) a triaxialaccelerometer (ST LIS3DH [19]) a 8GB micro-SD iexclashmemory card and a 33V lithium battery e accelerationmeasuring range of the sensor is dynamically user selectablefull scales of plusmn2 g plusmn4 g plusmn8 g plusmn16 g and the samplingdata rate is programmable from 1Hz to 5 kHz Here theacceleration measuring range is set to plusmn16 g and the sam-pling data rate to 120Hz For the acquired signal data theywould be stored in the micro-SD memory card temporarily

Oppositeinitial contact

10

Toe off

Feet adjacent

Initialswing

Terminalswing

LoadingresponseTerminal

stanceMidstance

Preswing

Midswing

Tibia vertical

Opposite toe off

Heel rise

Initial contact

14

13

10

20Stance phase (60)

20

13Swing phase (40)

Figure 1 Events periods and cycles of human gait [15]

2 Journal of Healthcare Engineering

and later the data are uploaded to server computers for signalprocessing In the procedure of motion signal acquisition thesensor would be attached steadily on one foot of a subject toa position 2sim3 cm above hisher ankle

To record the video of gait movements a camera (SonyAS100V Action Cam [20]) shown in Figure 2 was appliede highest image capture rate of the camera is 240 FPS(frames per second) and each captured image has theresolution of 1280times 720 pixels us the camera satises therequirement of our experiment for recording and watchingthe actions of gait events For the sake of recording syn-chronization between the sensors and cameras the imagecapture rate was set to 120 FPS to meet the requirement ofsynchronizing gait signals and foot locomotion images

23 Control of Cadence In this study the music tempo wasemployed to adjust the cadence of subjects for observing thegait variability in fast walking Normally the music tempo iscalculated by beats per minute (BPM) and the number ofbeats for walking means the number of steps to go ina minute Generally the cadence of a normal person inwalking is about 80sim120 BPM and for fast walking it isabout 120sim140 BPM For the purpose of comparing the gaitpatterns in slow normal and fast walking we focused on thetempo of 80 100 and 120 BPM for the controlled cadence

Usually music has many dierent styles such as popmusic like hip-hop rock and roll and jazz which havea strong tempo to ease the regulation of the walking rhythmHere the jazzrsquos tempo was considered to regulate thewalking rhythm [21] is idea has three considerationsFirstly it usually possesses the drum sounds of low fre-quency for beats so that the subjects can identify them clearlywhile walking Secondly the jazzrsquos tempo is almost xed andgood for gait control Lastly it can make subjects feel relaxedand walk with ease Although a metronome seems a littlebetter than the jazzrsquos music to provide a clear tempo we stillpreferred the jazzrsquos tempo due to the above considerations

24 Synchronization between Motion Signals and CapturedImages fromVideo In the aforementioned signal acquisition

system the sampling rate of the wearable sensors was set to120Hz and the video frame rate of the high-speed cameraswas also set to 120 FPS erefore a synchronizationmechanism is needed to relate the image of one gait eventrecorded by the camera to one motion signal acquired by thewearable sensor sequentially for identifying the gait char-acteristics To complete the synchronization a leadingclearance matching synchronization mechanism was pro-posed and we asked all participants to stand still for 10seconds after the setup of the sensors and cameras andbefore each walking test In this way a long and clear sectionof the temporal gap can be created for identifying thesynchronized origin of the gait images and motion signals

Under the supervision of onemedical expert in our teamwe discovered the gait events continuously in the recordedvideo through a frame-by-frame video player in slow mo-tion At the same time we marked the temporal position ofeach found image and used that position to extract thecorresponding accelerometrical signal of each gait event inthe acquired signal set In this study we call the corre-sponding accelerometrical signal of a gait event as a gaitcharacteristic As shown in Figure 3 a sample curve of gaitcycles acquired from the right foot of one participant isdepicted with a signicant valley (labeled by ldquoV-ICrdquo)

Sony AS100V Action Cam

Memory

Sensor

30 mm

50 m

m

MCU

Figure 2 Wearable sensor module with a triaxial accelerometer and a high-speed camera for gait signal acquisition

0

TO FA

TV OTO

IC-P

V-IC

IC HR Time

Swing phase Stance phase

Figure 3 Matching gait characteristics for the gait events of theright foot through image marking in one gait cycle

Journal of Healthcare Engineering 3

a signicant peak (ldquoIC-Prdquo) and six gait events For the gaitcharacteristics of the left foot they can be acquired in thesame way as above By locating several sets of the gait cyclesas a reference and using specic computer algorithms itbecomes easy and quick to nd out the gait events in therecorded video and the corresponding gait characteristics

rough the synchronization mechanism the appearedgait characteristics can be veried exactly by the actions ofthe gait events as shown in Figures 1 and 3 From thetemporal positions of the gait events and characteristicsrelevant gait parameters can be calculated accordingly

3 Research Methods

To obtain useful human gait information there are ve stagesthat is data acquisition signal preprocessing signal seg-mentation characteristic extraction and analysis for thesignal processing of gait characteristics (Figure 4) For the gaitsignal acquisition two wearable triaxial acceleration sensorswere used where each foot was put on one sensor Since therewere many other unrelated signals involved in the raw gaitsignals from the sensors the fast Fourier transform (FFT) [22]was used to analyze the noise and signal spectra and a But-terworth lter [23] to remove the noise Subsequently theltered signals on x- and y-axes were segmented according tothe gait cycles with the assistance of using the modulusmaxima wavelet (MMW) analysis [24] to identify the tem-poral separation positions A number of gait characteristicswere then extracted to compute the essential gait parametersFinally the gait parameters were fused together for the gaitanalysis of tripping risks in free and fast walking undervarious cadences regulated by dierent jazzrsquos tempos

31 Experimental Protocol In this study twenty-ve healthyparticipants were recruited for the experiment of walkingunder dierent cadences and each of them had completed

an auditory test to ensure that their hearing and temporecognition were normal In terms of performing all gaitmovements all participants had been veried to have nomedical restrictions on their activities tempo recognitionand walking Before the walking tests all participants onlyknew that they would take part in the experiment of normalwalking and they did not know that the experiment involvedmusical rhythme data obtained from the cases of withoutand with playingmusic tempos would be compared and thenbe analyzed for the eects of walking on playing tempos forall gait events and periods

In the walking tests the sensors were xed on the feet ofthe participants at the position 2sim3 cm above their ankle ofeach foot e sensors measure the gait signals in a 3Dcoordinate system where the longitudinal lateral andvertical axes are represented as x- y- and z-axes as shown inFigure 5 It is worthy to note that the signal component ofthe longitudinal axis (or x-axis) has the most obvious fea-tures of gait while walking and shows mainly the movementcharacteristics in the acceleration aspect erefore the

Gait signal inputData acquisition

Signal preprocessing

Signal segmentation

Characteristicextraction

Analysis

x-axis extraction

Gait parameter combination

Analytical results

y-axisextraction

Gait cyclecomputation

Characteristiccomputation

Correlationcomputation

MMW analysis and segmentation algorithm

FFT and Butterworth filter

Body swaycomputation

Figure 4 e iexclow diagram of gait signal processing

x-axis

y-axis

z-axis

Figure 5 e 3D coordinate system for signals from the wearabletriaxial acceleration sensors


x-axial component of the gait signals in the accelerometerwas mainly investigated and analyzed to extract the gaitparameters in our study For the signalrsquos lateral axis (ory-axis) the corresponding signal components reveal the leftand right accelerations of gait movements and they help tounderstand the walking balance of participants us they-axial signal components of gait movements would beincorporated into our study of gait analysis In the obser-vations of the signalrsquos vertical axis (or z-axis) it was foundthat the signalrsquos variations were greatly with personal habitin walking and it might be suitable to investigate personalwalking habit erefore the signal components of x- andy-axes were considered as the basis to study and analyze thegait characteristics of tripping risks in this study

After installing the sensors the participants were askedto walk barefoot to and fro the experimental pathway for3 minutes so that they could get accustomed to walkingnaturally wearing the sensors In the tests all participantswalked barefoot on a 30-meter straight pathway of a ceramictile iexcloor in 7 dierent walking tests which followed theprocedure as shown in Figure 6 For the purpose of syn-chronizing the motion signals and captured images fromvideo one camera installed at the starting point of thepathway was employed In addition there were six camerasset on both sides of the pathway at the positions of 10 15and 20 meters from the starting point ese cameras wereused to record the gait movements in the tests

For the rst test the participants were asked to walk intheir most natural and comfortable way which would betreated as the normal gait (Test 1) After that they wereasked to take a 3-minute rest and meanwhile jazz music at

80BPM was played to prepare the participants for the next teste volume of the music that the participants listened to wasaround 65sim70db since the volume of conversation betweennormal people is usually about 60db After the rest the par-ticipantswere asked towalkwhile the jazz at 80BPMwas playing(Test 2) and then they were asked to take the second 3-minuterest with the jazz at 100BPM being played After that they wereasked towalkwhile the jazz at 100BPMwas playing (Test 3) andthen they took the third 3-minute rest with the jazz at 120BPMbeing played After that they walked again while the jazz at120BPM was playing (Test 4) and then they took the fourth3-minute rest with the jazz at 80BPM being played Up to thispoint the participants completed the tests with no cue

For the subsequent tests the participants were asked tolisten to the played jazzrsquos tempo and synchronize theirwalking cadence with the tempo e fourth 3-minute restcould enable them to adapt their walking cadence to followthe music tempo e next test made the participants walkto the beats of the jazz at 80 BPM (Test 5) After that theytook the fth 3-minute rest with the jazz at 100 BPM beingplayed After the rest they were asked to walk to the beats ofthe jazz at 100 BPM (Test 6) en after the sixth 3-minuterest with the jazz at 120 BPM being played they were askedto walk to the beats of the jazz at 120 BPM (Test 7)

32 Signal Preprocessing Since the sensors have very highsensitivity slight environmental vibrations may cause

100

50

0

ndash500 2 4 6 8

Signal time (s) (120 Hz)10 12 14

Acce

lera

tion

(ms

2 )

15

1

05

0

times10ndash3

10 15 20 25 30 35Frequency (Hz)

45 5040 55 60

Pow

er sp

ectr

um

(a)

0 200 400 600 800Signal point (120 Hz)

1000 1200 16001400 1800

2010

0ndash10ndash20ndash30Ac

cele

ratio

n (m

s2 )

(b)

Figure 7 Gait noise removal (a) the raw gait signals and their FFTspectrum and (b) the ltered gait signals

e experiment preparation

End

Rest 3 minutes

Rest 3 minutes

Rest 3 minutes

Rest 3 minutes

Rest 3 minutes

Rest 3 minutes

Test 1 comfortable walk

Test 2 walk at 80 BPM (no cue)






Figure 6 e iexclow diagram of the walking tests


signicant and abrupt changes in the acquired signals As thegait signals were captured and recorded by the sensors theoriginal signals obtained were often mixed up with manyother irrelevant signals which are collectively called noisesAs a consequence the FFT was employed to nd out themajor spectrum of the gait movements of common people asshown in Figure 7(a) It can be seen that the frequencydistribution is roughly from 0Hz to 50Hz Since the gaitmovements of common people are usually at a lower fre-quency band the Butterworth lter [25] which is a low-passlter and has high computational eordfciency was used toremove the frequency higher than 20Hz for the gait analysisin this study After removing the noise signals the lteredgait signals can be seen as shown in Figure 7(b)

33 Gait Cycle Segmentation and Characteristic ExtractionTo investigate gait cycles the gait characteristics of both leftand right feet were synthesized as shown in Figure 8 based onthe aforementioned leading clearance matching synchroni-zation mechanism By observing Figures 1 3 and 8 a gaitcycle of the right foot starts from a TO event (TOR) and thenpasses through the characteristics FAR TVR and ICR tocomplete the swing phase of the right foot Subsequently theright foot turns into the stance phase which has the char-acteristic sequence of an OTO event (ie TOL) HRR and anOIC event (ICL) For a gait cycle of the left foot the sequenceof gait characteristics is the same as that of the right foot

In Figure 3 the easiest temporal separation positions to berecognized for the two gait phases are the TO event (a signal

peak) and the IC event (a zero-crossing point) between theV-IC point (a signal valley) and the IC-P point (a signal peak)erefore if the signal characteristics of TO and IC wereidentied exactly then gait phases and cycles could be seg-mented accordingly e temporal positions of the above twocharacteristics could then be used to compute the durations ofa stance phase a swing phase and the associated gait cycle

Since it is a tedious work to nd out all gait eventsmanually from the recorded video computer algorithmsshould be considered to extract the gait characteristics fromthe triaxial accelerometrical signals Since the TO event andthe zero-crossing point for IC are the most importantcharacteristics to segment and extract all gait characteristicsthe detailed working steps for the gait segmentation andcharacteristic extraction are described in Algorithm 1

In the GCS algorithm there are ve major steps to locatethe temporal positions of V-IC IC-P and IC and then a gaitcycle can be determined To extract other gait characteristicsa gait eventsrsquo neighboring relationship shown in Figures 1 3and 8 can be found and exploited For example the temporalinterval between the IC and TO points is about 60 of a gaitcycle length thus the IC temporal position to the TO shouldnormally be 60 plusmn 5 is neighboring relationship canbe utilized to locate the TO temporal position from the ICe rest characteristics can also be located in the same way asthe working steps described in Algorithm 2

In Algorithms 1 and 2 no working procedure forextracting the HR characteristic was describedis is becauseno signicant signal feature around the HR gait event can be

Step 1 Signal extrema depiction apply an MMW algorithm to the accelerometrical signals (Figure 9(a)) to produce signalsrsquo peaks andvalleys (Figure 9(b))

Step 2 ldquoV-ICrdquo extraction apply a threshold THVminusIC to lter out those illegal extrema as shown in Figure 9(c) For extracting ldquoV-ICrdquo

(1) retain vi with |vi|ltTHVminusIC for the V-IC of the ith cycle where vi is the signal of the ith V-IC(2) for all consecutive vi and vi+1 compute temporal_length (t(vi+1)minus t(vi)) in the unit of signal points where t(vi)

indicates the temporal position of vi(3) if 51le temporal_lengthle 120 signal points (since the normal cadence in walking is about 60sim140 BPM) retain both of the

vi and vi+1(4) if not discard the bigger one

Step 3 ldquoIC-Prdquo extraction do the same way as the last step to ldquoV-ICrdquo and then produce piStep 4 IC extraction for the ith IC search a xi point between t(vi) and t(pi)with |xi| 0 mark xi as the IC characteristic as shown in

Figure 9(d)Step 5 Gait cycle segmentation for each neighboring pair (ICiminus1 ICi) in the gait signals of one foot compute Gait_Cycle_lengthi

(t(ICi)minus t(ICiminus1)) in the unit of signal points

ALGORITHM 1 Gait cycle segmentation (GCS)

TOR FAR TVR ICR OTO HRR OIC TOR hellip0 13 27 40 70 100

Doublesupport

50 63 77 90 120 150TOL FAL TVL ICL OTO HRL OIC TOL hellip

Right footsequence

Le footsequence

Figure 8 Gait signal synthesis of the left and right feet the cyclic sequence of gait events


Step 1 ldquoTOrdquo extraction extract the ldquoTOrdquo of the ith gait cycle by nding the rst TO peak which appears in the interval [t(ICi) +Gait_Cycle_lengthi+1lowast(60minus5) t(ICi) + Gait_Cycle_lengthi+1lowast(60 + 5)] as shown in Figure 9(e)

Step 2 ldquoFArdquo extraction extract the ldquoFArdquo of the ith gait cycle by nding the rst FA peak which appears in the interval [t(TOi) +Gait_Cycle_lengthi+1lowast(13minus3) t(TOi) + Gait_Cycle_lengthi+1lowast(13 + 3)]

Step 3 ldquoTVrdquo extraction extract the ldquoTVrdquo of the ith gait cycle by nding the rst TV peak which appears in the interval [t(TOi) +Gait_Cycle_lengthi+1lowast(27minus 3) t(TOi) + Gait_Cycle_lengthi+1lowast(27 + 3)]

ALGORITHM 2 Gait characteristic extraction (GCE)

Signal points (120s)

Acce

lera

tion

(ms

2 ) 2010

0ndash10ndash20ndash30

0 100 200 300 400 500 600

(a)

reshold line

0

2010

ndash10ndash20ndash30

0 100 200 300 400 500 600

Acce

lera

tion

(ms

2 )


(b)

Discard

2010


0 100 200 300 400 500 600

Acce

lera

tion

(ms

2 )


(c)

IC point

0

2010


0 100 200 300 400 500 600

Acce

lera

tion

(ms

2 )


(d)

TO point

0

2010


0 100 200 300 400 500 600

Acce

lera

tion

(ms

2 )


(e)

Figure 9 Gait signals processed by the gait cycle segmentation algorithm (a) ltered signals (b) setting up a threshold line (c) unqualiedvalley points that are discarded (d) ldquoICrdquo nding and (e) ldquoTOrdquo nding


found+erefore the temporal positions of the HR event wereextractedmanually by observing the recorded video under thesupervision of the medical doctor in this study

34 CorrelationAnalysis In fast walking we are interested inexploring the relationship of tripping risks to the gaitcyclesphaseseventsperiods under a chosen cadence+rough a correlation analysis we could examine the re-lationship between the cadence and the gait characteristicswhich were obtained by the aforementioned algorithmsSubsequently the correlation coefficients could be calculatedto express the trend of positive negative or no correlationbetween two data sets Let x1 x2 xn1113864 1113865 andy1 y2 yn1113864 1113865 be two data sets for the gait characteristics+en Pearsonrsquos correlation coefficient rxy is computed by

rxy n 1113936 xiyi minus 1113936 xi 1113936 yi

n 1113936 x2i minus 1113936 xi( 1113857

21113969

middot

n 1113936 y2i minus 1113936 yi( 1113857

21113969 (1)

For the interpretation of the correlation coefficient thepositivenegative sign means the trend instead of the degree+e correlation degree between 0simplusmn04 04simplusmn07 and07simplusmn10 indicates that the two data sets have low moderateand high correlation respectively +e correlation degree isplusmn10 which means that the two sets are totally correlated

4 Results

In the walking experiment there were totally 2600 gait cyclesacquired and stored into 175 recordsWith the segmentationand extraction algorithms the characteristics listed inTable 1 were obtained by locating their temporal positions+en all gait periods and cycles belonging to each partic-ipant were normalized by the ratio in percentage corre-sponding to the beginning of each gait cycle

41 Cadence versus Gait Events +e correlation analysis forthe cadence and relevant characteristics were calculated asshown in Table 1 In this table we can find that the mostimportant characteristics correlated to the cadence are theTO and V-IC characteristics (Figure 3) which can be used todelineate the phases of a gait cycle Note that the V-IC pointis closely corresponding to the IC event +e correlationcoefficients of the TO and V-IC to the cadence are minus074 and+086 respectively+us the higher the cadence is the largerthe motion strength in the terminal swing period (betweenthe TV and IC events) will be applied and also thesmaller the motion strength in the TO event due to thenegative correlation

In Table 1 and Figure 8 the FA and OHR (opposite HR)characteristics which are corresponding to the FA andopposite HR events show the relevant correlation with thecadence +ese two characteristics can be corresponded tothe gait periods of the midswing and the opposite terminalstance with the correlation coefficients of minus064 and minus072respectively It indicates that the higher the cadence is theshorter the two periods will be

Basically the normal walking speed in daily life ofa person is equal to the product of hisher step length and the

cadence Since the average step length in walking of a personis highly related to hisher height the walking speed ismostly affected by the cadence In the order of absolutevalues of the correlation coefficients the cadence variation inwalking is mainly affected by the characteristics of V-IC(related to the TV and IC events or the terminal swingperiod) TO and OHR (or the opposite terminal stanceperiod) For example the signal curve around V-IC changesprominently in the terminal swing period because theswinging foot conducts greater acceleration and decelerationin fast walking When walking slowly the situation becomescontrapositive

From the above analysis it can be easily found that highercadences trigger larger motion strength of the swinging gaitevents and shorten the action time in the terminal swing andopposite terminal stance periods +is situation can likelyraise the tripping risks provided that an accident hinders theprogression of continuous gait events in the swing phase

42 Cadence versus Gait Periods In Figure 9 the V-ICcharacteristic acts as an anchor point for segmenting each gaitcycle We analyzed the correlation of the cadence and gaitperiods for each gait cycle and obtained the average correlationcoefficients as listed in Table 2 In this table the correlationcoefficient of ldquoNo_cuerdquo was obtained based on playing a musictempo without reminding the participants to notice the tempoand for ldquoCuerdquo the participants were informed to follow themusic tempo Obviously the music tempo influences slightlythe swing phase of every gait cycle in the ldquoNo_cuerdquo case sincethe beats can still influence the participants by hearing InldquoCuerdquo every gait period was influenced greatly by differentmusical rhythms so that the tempo changed the gait cycles andproduced different walking speeds

Table 2 Correlation analysis of the cadence and gait periods

PeriodType

No_cue CuePreswing 010 082Initial swing minus010 minus083Midswing minus022 minus079Terminal swing 020 082Loading response minus002 082Midstance 005 minus086Terminal stance minus003 084

Table 1 Correlation analysis of the cadence and relevant gaitcharacteristics

Cadence TO FA OHR TV V-IC IC-PCadence 1 minus074 minus064 minus072 minus032 086 003TO minus074 1 075 059 028 minus072 027FA minus064 075 1 076 018 minus052 004OHR minus072 059 076 1 054 minus073 025TV minus032 028 018 054 1 minus052 019V-IC 086 minus072 minus052 minus073 minus052 1 minus029IC-P 003 027 004 025 019 minus029 1


43 Gait Phase Variation versusWalk Types Since the musictempos could greatly aect gait cycles the gait phaseseparation on both feet was done with Algorithms 1 and 2 asshown in Table 3 where ldquoNormalrdquo means walking com-fortably and naturally NC_80 NC_100 and NC_120 meanwalking in the ldquoNo_cuerdquo cases with tempo 80 100 and120 BPM and C_80 C_100 and C_120 mean walking in theldquoCuerdquo cases with tempo 80 100 and 120 BPM

As walking in ldquoNormalrdquo and ldquoNo_cuerdquo the gait phases ofboth feet dier only around the range of less than 122 Forwalking in ldquoCuerdquo the dierence in C_80 and C_100 is small(le033) but it becomes large (167) in C_120 Inaddition the gait phase variation increases as the cadenceincreases in ldquoCuerdquo

In Table 3 it can be seen that the left foot spends longertime in the swing phase and shorter time in the stance phasein comparison with the right foot It reveals that the par-ticipants would like to match the beat with the left IC gaitevent so that the gait cycles of the right foot tend to varygreatly to minimize the speculation of the next beat Inaddition there is no obvious tendency of the gait phases inthe cases of ldquoNormalrdquo and ldquoNo_cuerdquo However in ldquoCuerdquoa greater swing time variation happens between C_100 andC_120 It shows an increasing tendency that the left footspends longer time in swing

To disclose more messages from the above observationswe examine the gait signals on the y-axis which can revealthe situations of left and right body sways as shown inFigure 10 Note that the gait signals in acceleration on y-axisin Figure 10 are corresponding to the gait signals on x-axis inTable 3 Since the gait cycle (particularly in stance) and bodysway have a close relationship to the balance the extremevalues of acceleration on y-axis can be used as a balanceindicator for the fast walk us the walking balance in gaitis investigated according to dierent walking speeds basedon with and without musical cue In Figure 10 a box plotshows an interval between a pair of accelerationrsquos maximumand minimum and an interquartile range of acceleration forone walk type

According to the walk types we can see that the box plotsfor the types in ldquoNo_cuerdquo are almost similar to each otherand for ldquoNormalrdquo its box plot is a little smaller to the cases ofldquoNo_cuerdquo is means that walking comfortably and nat-urally and walking without caring about the cue may havestable and balanced movements in gait with the lower bodysway As walking in ldquoCuerdquo the increasing tendency of thebody sway matches the increase of cadence Especially thecase of C_120 is the most unstable and unbalanced in all

cases due to the large range of acceleration from the left andright body sways In addition the case of C_80 is the moststable and balanced in all cases since the low cadence reg-ulated by the slow tempo benets the balance control ofbodyrsquos gait movements

44 Gait Periods versus Walk Types In Table 4 the per-centage of each leftright gait period is calculated withrespect to each gait cycle occurring in the seven dierentwalk types Noteworthily the gait periods of all participantsare consistent after the normalization of various gait cyclesIn this table we can see that the walk types of ldquoNormalrdquo andldquoNo_cuerdquo have no big dierence in the seven gait periodssince the walking cadence was determined by the partici-pants For the walk types of ldquoCuerdquo it can be observed thatthe gait periods of the leftright preswing and loading re-sponse are not sensitive to the higher cadence variation(ie 100 and 120 BPM) but is sensitive to the slow temposuch as 80 BPM

For the gait periods of the initial swing midswing andmidstance of both feet their period lengths in percentageincrease as the cadence increases For the terminal swing andterminal stance their period lengths in percentage decreaseas the cadence increases us we can nd that the musicalrhythm can intervene with the walking speed by regulatingthe cadence which will aect the increase or decrease of gaitperiod lengths

In addition we nd that the increments of the gait periodlengths for the midstance midswing and initial swing aremore than those in all other cases and the same situationhappens for the decrements of the period lengths for theterminal swing and terminal stance is indicates that thereexists a tripping risk the longer time for doing the midswing

Table 3 Percentage of the leftright gait phase with respect to the normalized gait cycle in seven dierent walk types

Phase ()Type

Normal NC_80 NC_100 NC_120 C_80 C_100 C_120Left swing 4222 4200 4167 4200 3800 4122 4300Right swing 4100 4178 4067 4133 3789 4089 4133Left ndash right 122 022 100 067 011 033 167Left stance 5778 5800 5833 5800 6200 5878 5700Right stance 5900 5822 5933 5867 6211 5911 5867Left ndash right minus122 minus022 minus100 minus067 minus011 minus033 minus167

15

10

5

0

ndash10

ndash5

Normal NC_80 NC_100 NC_120 C_80 C_100 C_120Ac

cele

ratio

n (m

s2 )

Figure 10 e box plot of the gait cycle variations in y-axis


means the stride becomes bigger and the shorter terminalswing means the movements in the tibia vertical to kneeextension (before the initial contact) become faster

If the walking paths were not flat and there were bumpsthat might block the movements in the tibia vertical to kneeextension a tripping risk would raise greatly due to thelonger time in midstance blocking the next stance supportwhile the center of body mass is moving ahead of the nextstance support point In summary larger steps or strides inwalking maintain larger gait speed and consume morecalories In this way cadence can be kept slower to preventtripping risks but we still have higher gait speed to consumecalories and promote health

5 Conclusion

Due to technological and spatial limitations fast walking forlong distance in daily life is difficult to be observed byscientific equipment In the literature most of the gait ex-periments were done by walking on treadmills [5 7 8 10]Actually gait signals acquired on treadmills are quite dif-ferent from the signals on the ground +is hinders theobservation of a fast gait on ordinary ground In this studywe conducted all the walking tests in a flat and straightpathway intended to reveal important clues for trippingrisks in fast walking

In the walking tests music cueing in three tempos wereemployed to regulate the walking speed of participants andseven walk types which include one ldquoNormalrdquo threeldquoNo_cuerdquo and three ldquoCuerdquo tests were devised for theacquisition of gait signals From the extracted gait charac-teristics we found that ldquoV-ICrdquo is an important sign that hasthe largest signal variation in one gait cycle and is highlycorrelated with cadences+us it can be used to compute thecadence of one specific walk

In the ldquoNormalrdquo and ldquoNo_cuerdquo tests the difference incadences gait phases and gait periods in both types is minorsince the background music tempos in ldquoNo_cuerdquo are mostlyignored by the participants Meanwhile the left and rightbody sways in the frontal plane are insignificant It meansthat the gaits are stable and balanced in body movements

In the ldquoCuerdquo tests every gait period is influenced greatlyby different music tempos which may cause various walkingspeeds Firstly a larger time variation in foot swing existsbetween the cadences in 100 and 120 beats By observing thegait periodrsquos variation in the midswing and terminal swingthe longer midswing time means the bigger stride and theshorter terminal swing time indicates the faster movementsin the tibia vertical to knee extension Secondly by theobservation to a higher cadence (eg 120 beats) anotherfactor of tripping risks may exist due to the high negativecorrelation of midstance and midswing to the higher ca-dence +is condition causes smaller foot clearance over theground and increases greatly the probability of tripping Aswalking through ordinary or uneven ground the trippingrisk of passing over bumps becomes substantial in fastermovements

A walking speed is determined from a cadence multi-plied by a stride length and it indicates that a bigger stridelength needs only a lower cadence to keep the same walkingspeed +erefore in the clinical practice the principle of fastwalking safely in the outdoors should be based on a lowercadence with bigger strides for health promotion Re-markably the aforementioned results show that the cadencebelow 120 beats is preferred From another point of viewa completely flat ground becomes a necessary and sufficientcondition for fast walking +e principle can also apply toindoor activities or locomotion since our experimentalground is close to a home environment

Ethical Approval

+is study was approved by the Institutional Review Boardof the National Cheng Kung University Hospital College ofMedicine National Cheng Kung University (Approval noA-ER-104-268)

Consent

All participants gave written informed consent to take partin this research

Table 4 Percentage of the leftright gait periods with respect to each gait cycle occurring in the seven walk types

Period ()Type

Normal NC_80 NC_100 NC_120 C_80 C_100 C_120Left preswing 951 909 997 886 1121 957 935Left initial swing 1300 1322 1278 1322 1133 1356 1411Left midswing 1289 1222 1378 1378 1100 1378 1500Left terminal swing 1633 1656 1511 1500 1567 1489 1389Left loading response 927 913 969 880 1090 932 932Left midstance 1989 1956 2011 1978 1811 1922 2133Left terminal stance 1911 2022 1856 2056 2178 1967 1700Right preswing 919 921 986 940 1116 937 930Right initial swing 1378 1333 1300 1300 1078 1244 1500Right midswing 1444 1389 1389 1444 1122 1333 1456Right terminal swing 1278 1456 1378 1389 1589 1511 1178Right loading response 959 901 980 926 1095 952 936Right midstance 1989 1933 1911 2000 1878 1944 2178Right terminal stance 2033 2067 2056 2000 2122 2078 1822


Conflicts of Interest

+e authors declare that there are no conflicts of interestregarding the publication of this paper

Acknowledgments

+e authors are grateful to the MOST Taiwan for sup-porting this study under Contract nos MOST105-2622-E-224-003-CC3 and MOST106-2314-B-006-017-MY3

References

[1] G Abbud K Li and R DeMont ldquoAttentional requirementsof walking according to the gait phase and onset of auditorystimulirdquo Gait and Posture vol 30 no 2 pp 227ndash232 2009

[2] M Tsuruoka R Shibasaki and Y Tsuruoka ldquoAnalysis of 1ffluctuations of walking listening to Mozartrsquos musicrdquo inProceedings of the International Symposium on Computer-Based Medical Systems (CBMS) pp 62ndash65 Perth AustraliaOctober 2010

[3] A Delval C Moreau S Bleuse et al ldquoAuditory cueing of gaitinitiation in Parkinsonrsquos disease patients with freezing of gaitrdquoClinical Neurophysiology vol 125 no 8 pp 1675ndash1681 2014

[4] A B Bourgeois B Mariani K Aminian P Zambelli andC Newman ldquoSpatio-temporal gait analysis in children withcerebral palsy using foot-worn inertial sensorsrdquo Gait andPosture vol 39 no 1 pp 436ndash442 2014

[5] L R Nascimento C Q de Oliveira L Ada S M Michaelsenand L F Teixeira-Salmela ldquoWalking training with cueing ofcadence improves walking speed and stride length after strokemore than walking training alone a systematic reviewrdquoJournal of Physiotherapy vol 61 no 1 pp 10ndash15 2015

[6] M Roerdink P J Bank C L E Peper and P J BeekldquoWalking to the beat of different drums practical implicationsfor the use of acoustic rhythms in gait rehabilitationrdquo Gaitand Posture vol 33 no 4 pp 690ndash694 2011

[7] C Mendonca M Oliveira L Fontes and J Santos ldquo+e effectof instruction to synchronize over step frequency whilewalking with auditory cues on a treadmillrdquoHumanMovementScience vol 33 pp 33ndash42 2014

[8] D Eikema L Forrester and J Whitall ldquoManipulating thestride lengthstride velocity relationship of walking usinga treadmill and rhythmic auditory cueing in non-disabledolder individuals A short-term feasibility studyrdquo Gait andPosture vol 40 no 4 pp 712ndash714 2014

[9] J E Wittwer K E Webster and K Hill ldquoMusic and met-ronome cues produce different effects on gait spatiotemporalmeasures but not gait variability in healthy older adultsrdquo Gaitand Posture vol 37 no 2 pp 219ndash222 2013

[10] K Byung-Woo H-Y Lee and W-K Song ldquoRhythmic au-ditory stimulation using a portable smart device short-termeffects on gait in chronic hemiplegic stroke patientsrdquo Journalof Physicalgterapy Science vol 28 no 5 pp 1538ndash1543 2016

[11] B Mariani H Rouhani X Crevoisier and K AminianldquoQuantitative estimation of foot-flat and stance phase of gaitusing foot-worn inertial sensorsrdquo Gait and Posture vol 37no 2 pp 229ndash234 2013

[12] T Khurelbaatar K Kim S Lee and Y H Kim ldquoConsistentaccuracy in whole-body joint kinetics during gait usingwearable inertial motion sensors and in-shoe pressure sen-sorsrdquo Gait and Posture vol 42 no 1 pp 65ndash69 2015

[13] L Z Rubenstein ldquoFalls in older people epidemiology riskfactors and strategies for preventionrdquo Age and Ageing vol 35no 2 pp ii37ndashii41 2006

[14] S Lord C Sherrington H Menz and J Close Falls in OlderPeople Risk Factors and Strategies for Prevention CambridgeUniversity Press Cambridge UK 2007

[15] M W Whittle Gait Analysis An Introduction Butterworth-Heinemann Edinburgh Scotland 4th edition 2007

[16] M Nordin and V H Frankel Basic Biomechanics of theMusculoskeletal System Lippincott Williams amp WilkinsPhiladelphia PA USA 2001

[17] D J Magee Orthopedic Physical Assessment Elsevier HealthSciences Oxford UK 2014

[18] Texas Instruments MSP430F552x MSP430F551x Mixed-Signal Microcontrollers 2015 httpwwwticomlitdsslas590mslas590mpdf

[19] STMicroelectronics LIS3DH MEMS Digital Output MotionSensor 2011 httpwwwstcomresourceenapplication_notecd00290365pdf

[20] Sony Digital HD Video Camera Recorder Handbook 2014httpwwwsonyeesupportencontentcnt-manHDR-AS100Vlist

[21] I Monson Saying Something Jazz Improvisation and In-teraction University of Chicago Press Chicago IL USA2009

[22] K R Rao D N Kim and J J Hwang Fast FourierTransformndashAlgorithms and Applications Springer Netherlands2010

[23] Electronics Hub Butterworth Filter 2015 httpwwwelectronicshuborgbutterworth-filter

[24] L M Bruce and R R Adhami ldquoClassifying mammographicmass shapes using the wavelet transform modulus-maximamethodrdquo IEEE Transactions on Medical Imaging vol 18no 12 pp 1170ndash1177 1999

[25] K S Erer ldquoAdaptive usage of the Butterworth digital filterrdquoJournal of Biomechanics vol 40 no 13 pp 2934ndash2943 2007


International Journal of

AerospaceEngineeringHindawiwwwhindawicom Volume 2018

RoboticsJournal of

Hindawiwwwhindawicom Volume 2018


Active and Passive Electronic Components

VLSI Design



Shock and Vibration


Civil EngineeringAdvances in

Acoustics and VibrationAdvances in



Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawiwwwhindawicom

Volume 2018

Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom

The Scientific World Journal

Volume 2018

Control Scienceand Engineering

Journal of



Journal ofEngineeringVolume 2018

SensorsJournal of



RotatingMachinery


Modelling ampSimulationin EngineeringHindawiwwwhindawicom Volume 2018


Chemical EngineeringInternational Journal of Antennas and

Propagation




Navigation and Observation


Hindawi

wwwhindawicom Volume 2018

Advances in

Multimedia

Submit your manuscripts atwwwhindawicom

In medical applications acoustic tempos are frequentlyused in gait rehabilitation with positive instantaneous andprolonged transfer eects on various gait characteristics In[9] rhythmic auditory cues including music andmetronomebeats were used to improve disordered gait in clinicalconditionseir study indicated that music andmetronomecues may not be used interchangeably and the cue type andfrequency (cadence) should be considered when evaluatingeects of rhythmic auditory cues on gait In [10] the gaitpattern in various rhythmic auditory stimulation temposwas investigated in stroke patients e results showed thatas the tempo increased the spatiotemporal gait parametersof the stroke patients changed signicantly After gaittraining with rhythmic auditory stimulation gait parametersof the aected and unaected sides improved signicantlycompared to baseline

In the previous studies treadmills were often used forstep and pace control and thus the conducted experimentswere often restricted in space [5 7 8 10] erefore theresearch results were biased due to few data from daily walkIn recent years the advancement of the wearable technologyhas made sensors miniaturized and widely applied in manyresearch and applications [11] As the wearable sensorsovercome space restrictions and improve the degree offreedom in use these devices can be used to replace xed andexpensive measuring apparatuses [12] In this study wedeveloped a gait recording system which equipped with a setof wearable triaxial acceleration sensors and action camerase sensors are in miniature size and can be easily worn bysubjects to acquire gait signals from daily walk e actioncameras can be used to record gait movements With thissystem gait signals for daily walking can be acquired withoutspace restrictions

In [13] Rubenstein described that the risks of falls andtrips are due to some identiable factors such as weaknessunsteady gait confusion and certain medications and si-multaneously indicated that attention to the factors cangreatly reduce the rates of falling He also described that oneof the most eective fall reduction programs involve envi-ronmental inspection and hazard reduction In [14] Lordet al indicated that the most challenging area to incorporateinto fall aetiology is that of environmental risk factors sincethe research base is smaller than that for other types of riskfactors At the same time research in this area has typicallybeen less rigorous or has been troubled by methodologicallimitations Since a gait recording system with the wearabletechnology has been developed in this work to improve themethodological limitations we are interested in exploringthe tripping risks in fast walking through an uneven groundwhich may exist with a few environmental risk factors bya cadence-controlled strategy With the application of actioncameras and a specic synchronization mechanism betweenthe signals from the wearable sensors and the images fromthe cameras we realize the gait analysis of fast walkingwithout space restrictions

is investigation is organized as follows Section 2presents research materials including gait description gaitsignal acquisition cadence control and a signal synchro-nization mechanism Section 3 describes research methods

including the experimental protocol signal preprocessingsignal segmentation and characteristic extraction andcorrelation analysis Section 4 shows the results which in-clude correlation analysis in cadence versus gait events andperiods with and without cueing In addition analyses of gaitphase variation and gait periods versus walking types are alsodone for exploring the tripping risks Concluding remarksand future work are given in Section 5

2 Research Materials

21 Gait Description Human walking consists of a numberof cyclically sequential gait events and one complete gaitcycle (Figure 1) can be measured from one toe o event tothe next one [15] As shown in Figure 1 a full gait cycleincludes a stance phase and a swing phase where 60 of thecycle is spent in stance and 40 in swing [16]

e gait cycle has seven gait events that is initial contact(IC) opposite toe o (OTO) heel rise (HR) opposite initialcontact (OIC) toe o (TO) feet adjacent (FA) and tibiavertical (TV) [17] It can also be subdivided into seven periodsthat is loading response midstance terminal stance preswinginitial swing midswing and terminal swing While nelysubdividing gait events and periods for every cycle dierentgait parameters such as the cycle duration stance time swingtime stride length step length cadence and walking speed canbe analyzed to learn about personal gait features

22 System of Signal Acquisition In this study a wearablesensor module with a triaxial accelerometer shown in Figure 2was developed to acquire motion signals related to humangait e size of this sensor is about 50times 30mm2 and itincludes amicroprocessor (TIMSP430trade series [18]) a triaxialaccelerometer (ST LIS3DH [19]) a 8GB micro-SD iexclashmemory card and a 33V lithium battery e accelerationmeasuring range of the sensor is dynamically user selectablefull scales of plusmn2 g plusmn4 g plusmn8 g plusmn16 g and the samplingdata rate is programmable from 1Hz to 5 kHz Here theacceleration measuring range is set to plusmn16 g and the sam-pling data rate to 120Hz For the acquired signal data theywould be stored in the micro-SD memory card temporarily

Oppositeinitial contact

10

Toe off

Feet adjacent

Initialswing

Terminalswing

LoadingresponseTerminal

stanceMidstance

Preswing

Midswing

Tibia vertical

Opposite toe off

Heel rise

Initial contact

14

13

10

20Stance phase (60)

20

13Swing phase (40)

Figure 1 Events periods and cycles of human gait [15]










Memory

Sensor

30 mm

50 m

m

MCU


0

TO FA

TV OTO

IC-P

V-IC

IC HR Time






3 Research Methods







Signal segmentation


Analysis

x-axis extraction


Analytical results

y-axisextraction








x-axis

y-axis

z-axis









100

50

0

ndash500 2 4 6 8


Acce

lera

tion

(ms

2 )

15

1

05

0

times10ndash3

10 15 20 25 30 35Frequency (Hz)

45 5040 55 60

Pow

er sp

ectr

um

(a)

0 200 400 600 800Signal point (120 Hz)

1000 1200 16001400 1800

2010


cele

ratio

n (m

s2 )

(b)



End

Rest 3 minutes

Rest 3 minutes

Rest 3 minutes

Rest 3 minutes

Rest 3 minutes

Rest 3 minutes



























Doublesupport


Right footsequence

Le footsequence








Acce

lera

tion

(ms

2 ) 2010


0 100 200 300 400 500 600

(a)

reshold line

0

2010


0 100 200 300 400 500 600

Acce

lera

tion

(ms

2 )


(b)

Discard

2010


0 100 200 300 400 500 600

Acce

lera

tion

(ms

2 )


(c)

IC point

0

2010


0 100 200 300 400 500 600

Acce

lera

tion

(ms

2 )


(d)

TO point

0

2010


0 100 200 300 400 500 600

Acce

lera

tion

(ms

2 )


(e)






n 1113936 x2i minus 1113936 xi( 1113857

21113969

middot

n 1113936 y2i minus 1113936 yi( 1113857

21113969 (1)


4 Results









PeriodType















Phase ()Type


15

10

5

0

ndash10

ndash5


cele

ratio

n (m

s2 )





5 Conclusion






Ethical Approval


Consent



Period ()Type





Acknowledgments


References





























RoboticsJournal of




VLSI Design



Shock and Vibration







Journal of



Volume 2018



Volume 2018


Journal of




SensorsJournal of



RotatingMachinery





Propagation






Hindawi


Advances in

Multimedia










Memory

Sensor

30 mm

50 m

m

MCU


0

TO FA

TV OTO

IC-P

V-IC

IC HR Time






3 Research Methods







Signal segmentation


Analysis

x-axis extraction


Analytical results

y-axisextraction








x-axis

y-axis

z-axis









100

50

0

ndash500 2 4 6 8


Acce

lera

tion

(ms

2 )

15

1

05

0

times10ndash3

10 15 20 25 30 35Frequency (Hz)

45 5040 55 60

Pow

er sp

ectr

um

(a)

0 200 400 600 800Signal point (120 Hz)

1000 1200 16001400 1800

2010


cele

ratio

n (m

s2 )

(b)



End

Rest 3 minutes

Rest 3 minutes

Rest 3 minutes

Rest 3 minutes

Rest 3 minutes

Rest 3 minutes



























Doublesupport


Right footsequence

Le footsequence








Acce

lera

tion

(ms

2 ) 2010


0 100 200 300 400 500 600

(a)

reshold line

0

2010


0 100 200 300 400 500 600

Acce

lera

tion

(ms

2 )


(b)

Discard

2010


0 100 200 300 400 500 600

Acce

lera

tion

(ms

2 )


(c)

IC point

0

2010


0 100 200 300 400 500 600

Acce

lera

tion

(ms

2 )


(d)

TO point

0

2010


0 100 200 300 400 500 600

Acce

lera

tion

(ms

2 )


(e)






n 1113936 x2i minus 1113936 xi( 1113857

21113969

middot

n 1113936 y2i minus 1113936 yi( 1113857

21113969 (1)


4 Results









PeriodType















Phase ()Type


15

10

5

0

ndash10

ndash5


cele

ratio

n (m

s2 )





5 Conclusion






Ethical Approval


Consent



Period ()Type





Acknowledgments


References





























RoboticsJournal of




VLSI Design



Shock and Vibration







Journal of



Volume 2018



Volume 2018


Journal of




SensorsJournal of



RotatingMachinery





Propagation






Hindawi


Advances in

Multimedia




3 Research Methods







Signal segmentation


Analysis

x-axis extraction


Analytical results

y-axisextraction








x-axis

y-axis

z-axis









100

50

0

ndash500 2 4 6 8


Acce

lera

tion

(ms

2 )

15

1

05

0

times10ndash3

10 15 20 25 30 35Frequency (Hz)

45 5040 55 60

Pow

er sp

ectr

um

(a)

0 200 400 600 800Signal point (120 Hz)

1000 1200 16001400 1800

2010


cele

ratio

n (m

s2 )

(b)



End

Rest 3 minutes

Rest 3 minutes

Rest 3 minutes

Rest 3 minutes

Rest 3 minutes

Rest 3 minutes



























Doublesupport


Right footsequence

Le footsequence








Acce

lera

tion

(ms

2 ) 2010


0 100 200 300 400 500 600

(a)

reshold line

0

2010


0 100 200 300 400 500 600

Acce

lera

tion

(ms

2 )


(b)

Discard

2010


0 100 200 300 400 500 600

Acce

lera

tion

(ms

2 )


(c)

IC point

0

2010


0 100 200 300 400 500 600

Acce

lera

tion

(ms

2 )


(d)

TO point

0

2010


0 100 200 300 400 500 600

Acce

lera

tion

(ms

2 )


(e)






n 1113936 x2i minus 1113936 xi( 1113857

21113969

middot

n 1113936 y2i minus 1113936 yi( 1113857

21113969 (1)


4 Results









PeriodType















Phase ()Type


15

10

5

0

ndash10

ndash5


cele

ratio

n (m

s2 )





5 Conclusion






Ethical Approval


Consent



Period ()Type





Acknowledgments


References





























RoboticsJournal of




VLSI Design



Shock and Vibration







Journal of



Volume 2018



Volume 2018


Journal of




SensorsJournal of



RotatingMachinery





Propagation






Hindawi


Advances in

Multimedia








100

50

0

ndash500 2 4 6 8


Acce

lera

tion

(ms

2 )

15

1

05

0

times10ndash3

10 15 20 25 30 35Frequency (Hz)

45 5040 55 60

Pow

er sp

ectr

um

(a)

0 200 400 600 800Signal point (120 Hz)

1000 1200 16001400 1800

2010


cele

ratio

n (m

s2 )

(b)



End

Rest 3 minutes

Rest 3 minutes

Rest 3 minutes

Rest 3 minutes

Rest 3 minutes

Rest 3 minutes



























Doublesupport


Right footsequence

Le footsequence








Acce

lera

tion

(ms

2 ) 2010


0 100 200 300 400 500 600

(a)

reshold line

0

2010


0 100 200 300 400 500 600

Acce

lera

tion

(ms

2 )


(b)

Discard

2010


0 100 200 300 400 500 600

Acce

lera

tion

(ms

2 )


(c)

IC point

0

2010


0 100 200 300 400 500 600

Acce

lera

tion

(ms

2 )


(d)

TO point

0

2010


0 100 200 300 400 500 600

Acce

lera

tion

(ms

2 )


(e)






n 1113936 x2i minus 1113936 xi( 1113857

21113969

middot

n 1113936 y2i minus 1113936 yi( 1113857

21113969 (1)


4 Results









PeriodType















Phase ()Type


15

10

5

0

ndash10

ndash5


cele

ratio

n (m

s2 )





5 Conclusion






Ethical Approval


Consent



Period ()Type





Acknowledgments


References





























RoboticsJournal of




VLSI Design



Shock and Vibration







Journal of



Volume 2018



Volume 2018


Journal of




SensorsJournal of



RotatingMachinery





Propagation






Hindawi


Advances in

Multimedia



















Doublesupport


Right footsequence

Le footsequence








Acce

lera

tion

(ms

2 ) 2010


0 100 200 300 400 500 600

(a)

reshold line

0

2010


0 100 200 300 400 500 600

Acce

lera

tion

(ms

2 )


(b)

Discard

2010


0 100 200 300 400 500 600

Acce

lera

tion

(ms

2 )


(c)

IC point

0

2010


0 100 200 300 400 500 600

Acce

lera

tion

(ms

2 )


(d)

TO point

0

2010


0 100 200 300 400 500 600

Acce

lera

tion

(ms

2 )


(e)






n 1113936 x2i minus 1113936 xi( 1113857

21113969

middot

n 1113936 y2i minus 1113936 yi( 1113857

21113969 (1)


4 Results









PeriodType















Phase ()Type


15

10

5

0

ndash10

ndash5


cele

ratio

n (m

s2 )





5 Conclusion






Ethical Approval


Consent



Period ()Type





Acknowledgments


References





























RoboticsJournal of




VLSI Design



Shock and Vibration







Journal of



Volume 2018



Volume 2018


Journal of




SensorsJournal of



RotatingMachinery





Propagation






Hindawi


Advances in

Multimedia







Acce

lera

tion

(ms

2 ) 2010


0 100 200 300 400 500 600

(a)

reshold line

0

2010


0 100 200 300 400 500 600

Acce

lera

tion

(ms

2 )


(b)

Discard

2010


0 100 200 300 400 500 600

Acce

lera

tion

(ms

2 )


(c)

IC point

0

2010


0 100 200 300 400 500 600

Acce

lera

tion

(ms

2 )


(d)

TO point

0

2010


0 100 200 300 400 500 600

Acce

lera

tion

(ms

2 )


(e)






n 1113936 x2i minus 1113936 xi( 1113857

21113969

middot

n 1113936 y2i minus 1113936 yi( 1113857

21113969 (1)


4 Results









PeriodType















Phase ()Type


15

10

5

0

ndash10

ndash5


cele

ratio

n (m

s2 )





5 Conclusion






Ethical Approval


Consent



Period ()Type





Acknowledgments


References





























RoboticsJournal of




VLSI Design



Shock and Vibration







Journal of



Volume 2018



Volume 2018


Journal of




SensorsJournal of



RotatingMachinery





Propagation






Hindawi


Advances in

Multimedia





n 1113936 x2i minus 1113936 xi( 1113857

21113969

middot

n 1113936 y2i minus 1113936 yi( 1113857

21113969 (1)


4 Results









PeriodType















Phase ()Type


15

10

5

0

ndash10

ndash5


cele

ratio

n (m

s2 )





5 Conclusion






Ethical Approval


Consent



Period ()Type





Acknowledgments


References





























RoboticsJournal of




VLSI Design



Shock and Vibration







Journal of



Volume 2018



Volume 2018


Journal of




SensorsJournal of



RotatingMachinery





Propagation






Hindawi


Advances in

Multimedia












Phase ()Type


15

10

5

0

ndash10

ndash5


cele

ratio

n (m

s2 )





5 Conclusion






Ethical Approval


Consent



Period ()Type





Acknowledgments


References





























RoboticsJournal of




VLSI Design



Shock and Vibration







Journal of



Volume 2018



Volume 2018


Journal of




SensorsJournal of



RotatingMachinery





Propagation






Hindawi


Advances in

Multimedia




5 Conclusion






Ethical Approval


Consent



Period ()Type





Acknowledgments


References





























RoboticsJournal of




VLSI Design



Shock and Vibration







Journal of



Volume 2018



Volume 2018


Journal of




SensorsJournal of



RotatingMachinery





Propagation






Hindawi


Advances in

Multimedia




Acknowledgments


References





























RoboticsJournal of




VLSI Design



Shock and Vibration







Journal of



Volume 2018



Volume 2018


Journal of




SensorsJournal of



RotatingMachinery





Propagation






Hindawi


Advances in

Multimedia




RoboticsJournal of




VLSI Design



Shock and Vibration







Journal of



Volume 2018



Volume 2018


Journal of




SensorsJournal of



RotatingMachinery





Propagation






Hindawi


Advances in

Multimedia


StudyonTrippingRisksinFastWalkingthrough Cadence...

Documents

Transcript of StudyonTrippingRisksinFastWalkingthrough Cadence...