Using Simple Harmonic Motion to Estimate Walking Distance for Waist-mounted PDR

Transcription

1 Usig Simple Harmoic Motio to Estimate Walkig Distace for Waist-mouted PDR Ku-Cha La Departmet of CSIE Natioal Cheg Kug Uiversity Taia, Taiwa (R.O.C.) We-Yuah Shih Departmet of CSIE Natioal Cheg Kug Uiversity Taia, Taiwa (R.O.C.) Abstract A huge body of work utilized sigal stregth of short rage sigal (such as WiFi, Bluetooth, ultra soud or Ifrared) to build a radio map for idoor localizatio, by deployig a great umber of beaco odes i the buildig. The drawback of such a ifrastructure-based approach is that the deploymet ad calibratio of the system is costly ad labor-itesive. To overcome that, some prior studies proposed the use of Pedestria Dead Reckoig (PDR) for idoor localizatio. The PDR system does ot require to build a beaco-based ifrastructure, i which a small umber of sesors are put o the pedestria. These sesors (such as G-sesor ad Gyro) are used to estimate the distace ad directio that the user traveled. The PDR approach ca be geerally categorized ito two types: foot-mouted ad waist-mouted. I geeral, the foot-mouted system ca get accurate step legth, but perform poorly i estimated headig directio. O the other had, the waist-mouted system ca estimate directio with high accuracy, but is hard to measure the step legth. I this work, we proposed a waist-mouted based PDR usig oe 3-axis accelerometer ad oe gyroscope sesor. We utilize vertical acceleratio to implemet double itegral for measurig the user s istat height chage ad use some physical features of vertical acceleratio durig the walkig to calibrate the measuremet. The based o the Pythagoras' Theorem, we ca estimate each step legth based o the user s height chage durig his/her walkig. Our experimet results show that the accuracy is about 98.6% i estimatig the user s walkig distace. I. INTRODUCTION Idoor locatio sesig systems have become very popular i recet years. There are may real-world applicatios depedig o such a system. For example, oe ca cosider the locatio detectio of medical persoel i a hospital, locatio detectio of fireme i a buildig o fire, moitorig of parolees, ad trackig of shopper mobility i a mall [1]. There were a large body of prior work i the area of idoor localizatio, may of them ivolvig the use of sigal stregth idicator (RSSI) [] ad a large umber of preistalled devices, called beaco odes, which periodically emit sigals. By listeig [3] ad moitorig sigal chage [4], the user s locatio ca be estimated through triagulatio. While the idea of these existig systems is attractive, they require itesive pre-deploymet efforts for every ew buildig. I additio, these approaches also suffer iterferece ad multi-path problems [5]. O the other had, the persoal dead reckoig or pedestria dead reckoig (PDR) system is a self-cotaied techique for idoor localizatio. It is used to measure the trajectory of walker without the eed to pre-istall beaco odes i the buildig. I other words, this techique oly requires a couple of sesors to be put o the user, so that it ca be used i ay buildig without pre-istallig beaco odes or pre-buildig RF maps/propagatio models based o surveys of the eviromet. Most of the PDR systems use iertial sesors (accelerometer, gyroscope, ad digital compass) to measure step legth ad headig directio. Geerally, the PDR system that uses the iertial sesors ca be classified ito two categories, depedig o the place where the sesors are placed. For the first type, the sesors are mouted o foot. The footmouted method uses double itegral to estimate distace ad use gyroscope or compass to measure the headig directio. The secod is waist-mouted, which usually detects the step evet to calculate total umber of steps; ad the multiplies it by a costat step legth which is set based o the pedestria s characteristic (weight, height, ad age) to estimate the movig distace. Some waist-mouted methods also use liear combiatio or liear regressio to fid the relatioship betwee the acceleratio, walkig speed, ad step legth. The PDR system has a well-kow problem, called sesor drift. Due to the costraits of hardware, the iertial sesors will costatly have some small amout of errors whe estimatig the distace. I additio, the oise sigal will further exacerbate this problem. Whe oe use the accelerometer sesor to do double itegral i estimatig the distace, these errors will be accumulated. Therefore, oe eeds a mechaism to calibrate such errors. I foot-mouted methods, a method, called zero velocity update (ZUPT), is used to fix this problem, which is based o the characteristic of how people walk. Whe a perso walks, oe foot will stay o the groud first util the other oe touches the groud. At that momet, the static oe (i.e. the oe o the groud) will have zero velocity i the horizotal directio. The system ca use this feature to reset the calculatio of velocity. However, the waist-mouted methods have o such a feature to be utilized. I other words, it will be difficult to defie a momet whe the sesor is static i the horizotal directio whe oe is walkig. Therefore, the accuracy of a aïve waist-mouted method is typically worse tha a foot-mouted method. O the other had, although the

2 foot-mouted methods are good at estimatig the step legth, they are limited i gettig accurate orietatio. Existig waist-mouted methods ormally estimate the distace by multiplyig a fixed step legth with the total umber of steps. Whe the user is differet, the system eeds to be re-adjusted or re-traied. I additio, the sesor drift problem will make the results uusable for ormal operatio. I this paper, we propose a ovel waist-mouted method that ca use acceleratio directly to estimate the distace ad does ot require ay modificatio of the system whe the user is differet. Our cotributio lies i usig the characteristic of simple harmoic motio (SHM) to solve the sesor drift problem. We estimate the height chage of the waist usig the double itegral of the vertical acceleratio which is obtaied through the accelerometer. We the estimate the horizotal walkig distace from the height chage of the waist usig Pythagoras' Theorem. Our experimet results show that the accuracy of our distace estimatio method is about 98.5%. Fially, from the practical poit of view, our approach ca be readily implemeted o some daily devices such as PDA or smart phoe sice it is waist-mouted. II. RELATED WORKS Our work has bee built o prior work i idoor localizatio, gait recogitio ad pedestria dead reckoig. A. Idoor localizatio Previous studies of idoor localizatio ca be maily divided ito two categories. Oe uses the sigal stregth to build a radio map []. The other oe employs beaco odes that emit beaco sigal periodically [3, 4]. Some recet work proposed the use of existig WiFi access poits (APs) i the buildig for localizatio [0, 1]. Fially, some visio-based studies [] used camera to capture ad aalyze the user s movemet for robot localizatio. Most of these prior work utilized sigal stregth to locate the target. O the had, our study makes use of the sesor mouted o the user to estimate his/her trajectory for localizatio. B. Pedometer ad gait recogitio Pedometer [6] is a popular device used to moitor the umber of steps durig the walkig. They ca be maily divided ito three types. The first type used a mechaical arm which will moves up ad dow i respose to the hip s vertical acceleratios durig the walkig. The movemet of this arm will tur o ad off the circuit, so the device ca detect the step. The secod type is i priciple similar to the first type: a plastic tube is setup i the circuit ad a magetic coil is put i the tube. Whe oe is walkig, the coil will also move up ad dow ad it will cause electromagetic iductio to produce electric curret. Whe the curret passes the circuit, it cosiders that a ew step occurred. The third type uses accelerometer to detect the vertical acceleratio ad use some gait recogitio algorithm to detect a ew step evet, which is similar to our work. Most recogitio algorithms [7] used accelerometer which is either foot-mouted or waist-mouted. They typically observe the vertical acceleratio durig the walkig ad use a threshold to detect the peak of the sigal to recogize a ew step evet. Some pedometers ca also estimate the walkig distace by multiplyig the total umber of steps with a costat step legth [6]. However, the walkig patter of each ca be differet ad a costat step legth might itroduce iaccuracy. I our work, we estimate the distace of each step usig the simple harmoic motio ad Pythagoras' Theorem. C. PDR system The PDR system is used to estimate the trajectory of a, icludig the walkig distace, headig directio, ad eve the chage of height. It is maily used to the locatio accuracy i a eviromet where GPS is ot available. Iertial sesors such as accelerometer, gyroscope, ad compass, are commoly used to estimate the trajectory of pedestria i a PDR system. Depedig o the place where the sesor is mouted o the pedestria, we ca classify the PDR system ito two types: foot-mouted ad waist-mouted. Durig the walkig, whe oe foot is movig forward, the other oe must stad o the groud to support the weight of the body. The foot-mouted methods [8] use these movemet characteristics to detect the step evet. To estimate the stride legth, these methods typically perform a double itegral o the horizotal acceleratio. However, the accumulatio of distace estimatio [9] through double itegral, called sesor drift, could itroduce serious iaccuracy without further calibratio. Oe approach which was proposed to elimiate the cumulative error caused by sesor drift is called zero velocity update. Durig the walkig, oe foot eeds to stay o the groud ad remais static which the other foot swigig forward. At this momet, the agular velocity ad speed of this static foot will be close to zero. Usig this characteristic, whe sesor detects the foot touchig the groud, the system ca reset the horizotal velocity to zero to calibrate the accumulated. O the other had, for the waist-mouted PDR system [10], oe will ot be able to fid zero velocity i the horizotal directio, so the method used by foot-mouted PDR system ca ot be applied here. Oe commo approach for waistmouted PDR system is to use a step recogitio algorithm to record the umber of steps ad estimate the walkig distace by multiplyig the umber of steps with a costat step legth. Sice the differet users might have differet step legths, some prior studies [10] proposed a improvemet by first collectig empirical data from differet users ad the use such data ad liear regressio to fid the relatio betwee step legth, walkig frequecy, ad the variace of acceleratio. While such a method provides a higher accuracy for step legth, it requires the creatio of traiig data for every user first before the PDR system ca become useful. I geeral, the footmouted system ca provide a higher accuracy for step legth, but is ot suitable to estimate the headig directio. O the other had, the waist-mouted system ca estimate the orietatio of the user with a good accuracy. I additio, it ca be implemeted o a everyday device like smart-phoe or PDA. However, it is hard to estimate the walkig distace with a waist-mouted PDR system because it ca ot use the horizotal acceleratio directly [4]. I this work, we propose a ovel waist-mouted method which does ot require traiig for every user but the results ca still be as good as the footmouted system.

3 Measured axis: SesorReadig = α + Gravity Exteral force = α m/s Gravity = 9.8 m/s ExteralForce Readig axis Gravity = -9.8 m/s Figure 1. Gravity Elimiatio. III. METHODOLOGY I this sectio, we first describe the cocept of simple harmoic motio (SHM). We the discuss some pre-filterig mechaism to remove the effect of gravity o the sesor readigs. Next, we describe our system architecture ad algorithms i details. Fially, we discuss how we remove oise sigal durig walkig usig a low pass filter. A. The cocept of simple harmoic motio (SHM) Simple harmoic motio (SHM) [18] has bee commoly used to model various motios, such as the oscillatio of a sprig. Whe a object is i simple harmoic motio, the displacemet of object is proportioal to the exteral force placed o the object ad the force always poits to the positio of equilibrium. I other words, whe the object is displaced from its equilibrium positio, it experieces a et restorig force toward its equilibrium positio. Oe characteristic of SHM is that whe a object is at its largest displacemet from the equilibrium positio, the object s speed will become zero, ad at the same time, the object will reach its maximum acceleratio rate. For istace, let s put a pe o a sprig ad make the sprig oscillate i vertical way. The we put a log ribbo paper o the table ad let the pe draw o it to show the trajectory of sprig vibratio. At the same time, we pull this paper i the horizotal directio at a stable speed. The trajectory show o the paper will look like a siusoidal wave ad the vertical velocity of the highest ad the lowest poits of the sprig will be zero. Ispired by this, if we mout a sesor o the waist of a pedestria durig his walkig [19], the trajectory of sesor ca be approximated by a siusoidal wave. I additio, give that the velocity of the highest ad lowest poits of the sesor will be zero, we ca utilize these characteristics to detect whe the user starts a ew step. Fially, oce we fid the poits with zero velocity i the vertical way, we ca use these to do the double itegral for calculatig the height chage of the user s waist. The height chage ca the be used to estimate step distace based o Pythagoras' Theorem. B. Remove the effect of the gravity Durig the walkig, a perso s body will move up ad dow, as show i Fig. 4. I other words, if we assume the legth of leg is L, the waist lie will move up-ad-dow betwee L ad (L-h) from the groud, where h is the chage of the height of the waist. Cosiderig the triagle show i Fig. 4, ExteralForce Gravity (-9.8m/s ) Figure.1. Gravity Elimiatio i differet θ ExteralForce cosθ Figure.. Force Diagram. Gravity cosθ formed by two feet of a perso ad his step legth D. Give that L is kow, usig Pythagoras' Theorem, we ca estimate D if we kow the height of this triagle, i.e. (L-h). To obtai (L-h), we eed to first calculate h which is the height chage of the waist durig the walkig. Therefore, if we mout a accelerometer o the user s waist, the readigs of the accelerometer ca be used to estimate the height chage h which ca the be used to calculate the step legth D based o Pythagoras' Theorem. However, there is oe problem with this approach. The readigs from the accelerometer will be affected by the gravity. The accelerometer is typically built o a silico wafer i circuit to detect the acceleratio. There is a poly-silico sprig o the wafer surface to provide a force agaist the exteral force. Ay exteral force could cause the deflectio of the sprig. Accordig to differet levels ad directio of the deflectio, the values of resistace or capacitor i the circuit will chage proportioally, ad the the device will output the correspodig voltages which are the readigs that we ca measure from the sesor. However, eve i the static state, the sesor ca still detect the acceleratio due to that the gravity costatly gives the wafer a dowward force which is always 1g o earth, except whe the axis (we use a three-axis accelerometer) from which we collect the readig is perpedicular the to gravity force. Therefore, to obtai the vertical acceleratio of the waist durig the walkig for estimatig height chage, we eed to first remove the effect of the gravity from the sesor readigs. Whe the sesor is movig, there might be a agle betwee the directio of measured axis ad the gravity. This agle ca be from 0 to 180. For the sake of discussio, we cosider two cases: whe the axis is parallel to the gravity force ad whe it is ot. 1) If parallel This meas that the agle betwee the measured axis ad the directio of gravity is 0 or 180, as show i Fig. 1. I this case, we ca estimate the vertical acceleratio caused by the up-ad-dow movemet of the waist durig the walkig based

4 o the (1). Here SesorReadig is the readig we collect from the sesor for oe particular axis while α is the exteral force that causes sesor to move up or dow. α = Sesor Re adig Gravity( = 9.8m / s Note that sice the gravity is a dowward force, accordig to the law of iertia, whe the sesor moves up, the silico wafer will move dow. Therefore, the readyigα from the sesor will be more tha 1g (-9.8m/s ). ) If ot parallel If the agle betwee the measured axis ad gravity either 0 or 180, as show i Fig..1, the measuremet collected from the sesor will be cotributed by the compoets of two differet forces; oe is from the gravity ad the other oe is from the exteral force, as show i Fig... Therefore, to estimate the exteral force, we eed to first compute the compoet of the exteral force o the measured axis, which ca be achieved by subtractig the compoet of the gravity from the sesor readig. M = gravity cosθ N = ExteralForce Sesor Re adig = N cosθ + M ) (1) (.1) (.) (.3) I the above equatios, SesorReadig is the value read from the measured axis of the sesor which is the sum of the compoet of ExteralForce ad the compoet of the gravity. M is the compoet of the gravity compoet which ca be measured whe the sesor is static. N = ( Sesor Re adig M ) cosθ θ = cos 1 ( M / gravity) (.4) (.5) From the (.3), we ca calculate the exteral force with the (.4), ad θ ca be calculated usig Iverse Trigoometric fuctios, as show i the equatio.5. VerticalAc c = ( Sesor Re adig M ) csc(cos 1 ( M / 9.8)) (.6) Combiig (.4) ad (.5), we ca obtai the vertical acceleratio, as show i the (.6). Note that here we assume the exteral force comes from the up-ad-dow movemet of the waist i the vertical directio. C. The algorithm After the effect of gravity is removed, we ca obtai the vertical acceleratio due to the up-ad-dow movemet of the waist, which ca be the used to detect each ew step. Oce we ca distiguish every differet step, we ca do double itegral to calculate the height chage of the waist ad the use that iformatio to estimate the legth of each stride based o Pythagoras' Theorem. After the walkig distace is determied, combiig with the orietatio iformatio from the gyro sesor, we ca obtai the -D coordiate of the user usig (6). Before we describe our algorithm, we should first discuss how we iterpret the measuremets of vertical acceleratios durig the walkig. There are three major evets i oe step, icludig heel-off-groud, stace, ad heel-touchig-groud. Acceleratio (m/s^) Oe step Figure 3. The vertical acceleratio of walkig. As show i Fig. 3, the lowest valley of the wave idicates whe the heel is touchig groud, the first peak occurs whe the walker i the stace state, ad the secod valley of the wave idicate whe the walker just moves his heel off the groud. Durig the walkig, whe a perso steps forward ad the heel is touchig the groud, the positio of the waist is i its lowest positio. Accordig to the characteristic of SHM, the vertical velocity at this positio will be zero ad the vertical acceleratio will reach its maximum. Whe the body is i its stace, oe foot is o the groud ad the other foot is swigig forward. At this momet, the waist has the smallest displacemet from the equilibrium positio, so we reset the vertical velocity at this poit. After that, the body starts to lea forward ad the foot which was o the groud is ow o its tiptoe, which will give a force to push the body up, so that the vertical acceleratio will chage to the opposite directio ad cause aother valley i Fig. 3, due to the law of iertia. 1) Stride legth estimator Whe a ew step evet is detected, the system ca compute the step legth usig the double itegral: the itegral of acceleratio will give us the velocity ad we ca get the distace of the step from the itegral of the velocity. However, to avoid that the errors from the results of the double itegral are accumulated, we eed to do zero velocity update (ZUPT) to avoid the error accumulatio. Give that the sesor mouted o waist is i a simple harmoic motio, we ca fid it highest (the first peak i Fig. 3.) ad lowest (the first valley i Fig. 3) positios ad reset the vertical velocities of these two poits to be zero. Nevertheless, sice we cosider the height chage of the waist as a simple harmoic motio from its highest positio to the same positio agai, if we simply do double itegral, the displacemet will be zero or close to zero. Therefore, we calculate the absolute value of curret velocity ad use it to do the ext itegral. I (3) ad (4), the is the umber of acceleratio sample, V is the vertical velocity, a is the vertical acceleratio, ad h is the height chage. We ca get the h Stace Heel off groud Heel touchig groud L L L L-h D Time (40ms) Figure 4. Walkig Diagram.

5 velocity by doig itegral of vertical acceleratio. I (4), we use absolute value of velocity to do other itegral ad cosider the height chage due to ascedig ad descedig of the body movemet. The obtaied result is the divided by to get a average of height chage, here we assume that the height chage i the directios of ascedig ad descedig are approximately the same. Fially, we use the obtaied height chage to estimate stride legth based o the Pythagoras' Theorem. As show i Fig. 4, if we assumig that the kow leg legth of is L ad the height chagig is h, we ca use the (5) to calculate the step distace D. V = h = ( 1 1 D = a V dt dt) / L ( L h) (3) (4) (5) I our system, we cosider each step ca be divided ito two parts, from stace to stride ad the from stride back to stace. So if we cosider a user move from poit A to poit B, the the first step ad the last step are actually oly half a step, because the first step is oly stride-stace ad the last step is oly stace-stride. Therefore, we cosider the first step ad the last step as the special cases. If the system detects that the user is curretly i his first or last step, the height chage will ot be divided by two. I additio, there is oly oe poit where the velocity eeds to be reset. The iitial velocity of first step is already zero ad the last step does ot eed to reset the velocity whe the body is i the stace state. Except the above methods, durig pedestria walkig, there are ot oly regular walkig evets, but also may uexpected ad upredictable vibratios to iterfere the sesor value. Thus, we deped o low pass filter to get rid of these oises, like [3]. Fially, we ca combie the distace ad the orietatio iformatio (obtaied from gyroscope) usig (6) to calculate the user s coordiate ad trajectory i a D space. IV. EXPERIMENTS AND RESULTS I this sectio, we discuss the results of our experimets for the proposed method. We performed two experimets. For the first experimet, the user walked i a straight lie. We performed two set of tests. Oe is less tha 10 meter ad the other is loger tha 10 meter. I the secod experimet, we combie the orietatio iformatio (from gyro) to estimate the locatio of the user i a -D space. A. Hardware The sesor platform we used is Taroko which is a modificatio versio of TelosB mote [11] which is desiged origially by UC Berkley. Taroko is a programmable, lowpower wireless sesor platform. The microcotroller uit of Taroko is TI MSP430 F1611 with 16 bit RISC [1]. MSP430 has 48K bytes flash memory ad 10K bytes RAM that supports the serial commuicatios such as UART, IC, SPI, ad Digital I/O. Beside, Taroko is equipped with the CC40 RF trasceiver [13] which is a low cost device for wireless commuicatio i.4ghz based o IEEE [14]. The maximum radio distace is aroud 100m. It also supports the USB iterface usig FTDI chip [15]. Taroko ca use the USB iterface to coect to a computer for powerig, program upload ad data collectio. Taroko ca accommodate differet sesor compoets, e.g., temperature, humidity, ifrared, accelerometer, gyroscope, ad magetic. I this work, we use a exteded board to combie the accelerometer (ADXL330 [16]) ad gyroscope (IDG500 [17]). The ADXL330 is a small, low power, 3-axis accelerometer with sigal coditioed voltage outputs, all o a sigle moolithic IC. It ca measure acceleratio with a miimum full-scale rage of ±3g. The IDG- 500 is a agular rate sesor ad its full scale rage is ±500 /s. It uses the MEMS techology with vertically drive, vibratig masses to make a fuctioally complete, low-cost agular rate sesor. All required electroics are itegrated oto a sigle chip with the sesor. Our PDR algorithm is implemeted o the sesor to reduce the chace of packet loss whe sedig sesor ifo back to a computer for remote processig. B. Straight distace measuremet I this experimet, we performed two set of tests. Oe is i the room (about 6~8m) ad the other is at the corridor of buildig (10~15m). The former is repeated for 0 times ad the latter is repeated for 5 times. To obtai the groud truth, we used a laser distace meter to measure the distace. Durig each trial, the user put the sesor o his waist ad walked i differet speeds. The estimated distace is based o the results computed by the sesor, ad the groud truth is obtaied from the laser distace meter. Our results show that the average of accuracy is about 98.6% ad the stadard deviatio is 1.09%. The average accuracy for the loger-distace case is similar to that of the short-distace case. The average of accuracy is 98.5% ad the stadard deviatio is 1.9%. We also compare our results with the existig foot-mouted methods. Our accuracy is close to the best results from the foot-mouted method [8] (about more tha 98%) ad better tha other waistmouted method [10] (about 96%), as show i TABLE I. Note that the foot-mouted method is geerally ot good for estimatig the directio of headig, as discussed previously. C. Coordiate test with groud truth I this experimet, we combie the directio iformatio from gyro to calculate the -D coordiate usig (6). X Y = X = Y Dis + Dis Cos θ Si θ (6) Where X ad Y are the th coordiate, Dis is the estimated distace, ad θ is the agle by sesor detected. We multiply the Dis with Cos (θ) ad Si (θ) to calculate how far the user has moved i the directio of X-axis ad Y-axis, which is the add to the last coordiate to estimate the curret positio. The actual walkig distace is 10.75m, ad the estimated distace based o our method is about 10.80m. Note that although the effect of gyro drift is ot sigificat i this experimet, it is a importat issue to cosider for idoor localizatio. However, i this paper, we oly focus o calibratig the distace drift ad leave the calibratio of gyro measuremets as the future work.

6 TABLE I. V. CONCLUSION AND FUTURE WORK I this paper, we develop a idoor localizatio scheme usig waist-mouted PDR. We use the iertial sesors like accelerometer ad gyroscope to desig a waist-mouted system. Our algorithm is based o the characteristic of huma walkig ad to utilize simple harmoic motio ad Pythagoras' Theorem to estimate the user s locatio. The ovelty i our approach is that, i additio to usig the vertical acceleratio to detect the step evet (for coutig the umber of steps like i the previous studies), we make use of the vertical acceleratio directly to estimate the step legth. As compared to footmouted method, our scheme provides a better platform for estimatig the user s orietatio. However, this system still has some problems. First, we do ot cosider the calibratio of the user directio. I the future work, we pla to use the buildig map to calibrate the gyro drift. Secod, by usig this algorithm, the device orietatio must set i some particular places, such as o the belt, to prevet the mistake of data. There are some works about adjustig this error, so we do t focus o this problem. The third problem is about the usig i 3-D localizatio. To estimate the height chage, the system ca combie with the pressure sesor or aother accelerometer to deal with altitude estimatio. REFERENCES COMPARATION Our approach [8] Hardware 3-axies accelerometer & 6-DoF iertial sesor gyroscope (accelerometer + Gyro + Magetic) Sample rate 5Hz 00Hz Placemet Waist Foot ZUPT Use the cocept of SHM Calculate the agular velocity to detect Accuracy 98.6 ± 1.09% (total >98% (total distace) distace) >99% (coordiate) [1] F. Carl, ad G. Has. Locatio ad Navigatio Support for Emergecy Respoders: A Survey. IEEE Pervasive Computig, vol. 9, pp [] P. Bahl, ad V.N. Padmaabha. RADAR: a i-buildig RFbased user locatio ad trackig system. I Nieteeth Aual Joit Coferece of the IEEE Computer ad Commuicatios Societies 000 (INFOCOM 000), 000, pp vol.. [3] W. Roy, H. Ady, F. Veroica, ad G. Joatha. The active badge locatio system. ACM Tras. If. Syst., vol. 10, 199, pp [4] C. Ho-li, T. Jr-be, L. Tsug-Te, C. Hao-Hua, ad H. Polly. Spiig beacos for precise idoor localizatio. I Proceedigs of the 6th ACM coferece o Embedded etwork sesor systems, Raleigh, NC, USA, 008. [5] L. Tsug-Ha, I.-H. Ng, L. Seg-Yog, K.-M. Che, ad H. Polly. A microscopic examiatio of a RSSI-sigature-Based idoor localizatio system. I Fifth Workshop o Embedded Networked Sesors 008, -6. [6] C.-C. Shih. The Implemetatio of a G-Sesdor-based Pedometer. M.S. Thesis, Dept. C.S.I.E., Natioal Cetral Uiv., Taoyua Couty, Taiwa 010. [7] H.J. Ailisto, M. Lidholm, J. Matyjarvi, E. Vildjiouaite, ad S.-M. Makela. Idetifyig people from gait patter with accelerometers. I Biometric Techology for Huma Idetificatio II, Orlado, FL, USA, Moday 8 March 005. [8] L. Ojeda, ad J. Borestei. Persoal Dead-reckoig System for GPS-deied Eviromets. I 007 IEEE Iteratioal Workshop o Safety, Security ad Rescue Robotics (SSRR 007), 007, pp 1-6. [9] D.H. Titterto, ad J.L. Westo. Strapdow Iertial Navigatio Techology. Peter Peregrius Ltd, [10] S.H. Shi, C.G. Park, H.S. Hog, ad J.M. Lee. MEMS-Based Persoal Navigator Equipped o the User's Body. I ION GNSS 18th Iteratioal Techical Meetig of the Satellite Divisio, Log Beach, CA, september 005. [11] TelosB, Available: [1] MSP430 F1611, Available: [13] CC40, Available: [14] , Available: [15] FTDI, Available: [16] ADXL 330, Available: [17] IDG 500, Available: [18] Simple Harmoic Motio, Available: [19] E. Ayyappa. Normal Huma Locomotio, Part 1: Basic Cocepts ad Termiology. I JPO: Joural of Prosthetics ad Orthotics, vol. 9, 1997, pp [0] C. Krisha, I. Aad Padmaabha, ad N.P. Vekata. Idoor localizatio without the pai. I Proceedigs of the sixteeth aual iteratioal coferece o Mobile computig ad etworkig, Chicago, Illiois, USA, 010. [1] Y. Ji, S. Biaz, S. Padey, ad P. Agrawal. ARIADNE: A Dyamic Idoor Sigal Map Costructio ad Localizatio System. I MobiSys, 006. [] R. Ballagas, M. Rohs, J. G. Sherida, ad J. Borchers. Sweep ad poit & shoot: Phoecam-based iteractios for large public displays. I CHI 05: Exteded abstracts o Huma factors ad computig systems, 005. [3] M. Marti, ad M. Michael, A step couter service for Javaeabled devices usig a built-i accelerometer, preseted at Proceedigs of the 1st Iteratioal Workshop o Cotext- Aware Middleware ad Services: affiliated with the 4th Iteratioal Coferece o Commuicatio System Software ad Middleware (COMSWARE 009), Dubli, Irelad, 009. [4] R. Stirlig, J. Colli, K. Fyfe, ad F. Lachapelle. A iovative shoe-mouted pedestria avigatio system. I Proceedigs of Europea Navigatio Coferece GNSS 003, Graz, Austria, -5 April 003