`Vol. 3, No. 1-2: 273-279
`
`A Step, Stride and Heading Determination for the Pedestrian
`Navigation System
`
`Jeong Won Kim
`GNSS Technology Research Center, Chungnam National University, Korea
`e-mail: kimjw@cnu.ac.kr Tel: + 82-42-821-7709 ; Fax: +82-42-823-5436
`
`Han Jin Jang
`GNSS Technology Research Center, Chungnam National University, Korea
`e-mail: handoo01@cnu.ac.kr Tel: + 82-42-821-7709 ; Fax: +82-42-823-5436
`
`Dong-Hwan Hwang
`GNSS Technology Research Center, Chungnam National University, Korea
`e-mail: dhhwang@cnu.ac.kr Tel: + 82-42-821-5670 ; Fax: +82-42-823-5436
`
`Chansik Park
`School of Electrical and Computer Engineering, Chungbuk National University, Korea
`e-mail: chansp@cbucc.chungbuk.ac.kr Tel: + 82-43-261-3259 ; Fax: +82-43-268-2386
`
`Received: 15 Nov 2004 / Accepted: 3 Feb 2005
`
`Abstract. Recently, several simple and cost-effective
`pedestrian navigation
`systems
`(PNS) have been
`introduced. These systems utilized accelerometers and
`gyros in order to determine step, stride and heading. The
`performance of the PNS depends on not only the
`accuracy of the sensors but also the measurement
`processing methods. In most PNS, a vertical impact is
`measured to detect a step. A step is counted when the
`measured vertical impact is larger than the given
`threshold. The numbers of steps are miscounted
`sometimes since the vertical impacts are not correctly
`measured due to inclination of the foot. Because the
`stride is not constant and changes with speed, the step
`length parameter must be determined continuously during
`the walk in order to get the accurate travelled distance.
`Also, to get the accurate heading, it is required to
`overcome drawbacks of low grade gyro and magnetic
`compass. This paper proposes new step, stride and
`heading determination methods
`for
`the pedestrian
`navigation system: A new reliable step determination
`method based on pattern recognition is proposed from the
`analysis of the vertical and horizontal acceleration of the
`foot during one step of the walking. A simple and robust
`stride determination method is also obtained by analysing
`the
`relationship between stride, step period and
`acceleration. Furthermore, a new integration method of
`gyroscope and magnetic compass gives a reliable
`
`heading. The walking test is preformed using the
`implemented system consists of a 1-axis accelerometer, a
`1-axis gyroscope, a magnetic compass and 16-bit
`microprocessor. The results of walking test confirmed the
`proposed method.
`
`Key words: Pedestrian navigation
`system, Step
`detection, Stride determination, Heading determination
`
`1 Introduction
`
`Pedestrian navigation system(PNS) provides velocity and
`position of a person and can be applied to many other
`areas such as E-911 service, location based services
`(LBS), tourism, rescue, military infantry, medical studies,
`leisure, and navigation for the blind. In PNS, it is
`necessary to locate the position of the user in any time
`and any environment. Even GPS is useful personal
`navigation system, its availability is significantly reduced
`when a signal is blocked. Also ultra wide band (UWB)
`and radio frequency identification (RFID) techniques are
`introduced for personal navigation, but these systems
`
`Niantic's Exhibit No. 1019
`Page 001
`
`
`
`
`
`
`274
`
`Journal of Global Positioning Systems
`
`require dense infrastructure. For these reasons, a self-
`contained navigation system based on a dead reckoning
`(DR) principle is of interest (B. Merminod et al, 2002).
`To locate the position of the PNS user, distance and
`heading from a known origin have to be measured with
`an acceptable level of accuracy. In PNS, an accelerometer
`is used to count the number of steps, which is combined
`with the stride to obtain the travelled distance. In
`addition, a magnetic compass or gyroscope is used as a
`heading sensor.
`The stride and step are important parameters for PNS
`dead reckoning algorithm. Many methods have been
`suggested to detect a step. One such method is to detect
`the peaks of vertical acceleration, which correspond to
`the step occurrences because the vertical acceleration is
`generated by vertical impact when the foot hits the
`ground. If the vertical impact is larger than given
`threshold, it is considered as a step. Since the pattern of
`impact signal depends on type of movement (going up or
`down stairs, crawling, running etc.) and type of ground
`over which the person walks (hard or soft surface, sand),
`the determination of threshold is not so easy for reliable
`step detection (Ladetto and Merminod, 2002). This paper
`proposes reliable step detection method based on pattern
`recognition. From the analysis of the vertical and
`horizontal acceleration of the foot during one step of the
`walking, the signal pattern of walking behaviours is
`obtained.
`The stride of the walker in PNS is a scale factor in a dead
`reckoning algorithm. Unlike a scale factor of an odometer
`in a car navigation system, the stride in PNS is a time-
`varying parameter
`(Mar and Leu, 1996).
` The
`predetermined stride cannot be used effectively for the
`distance measurement because the strides of the walker
`are different according to the human parameters. The
`stride depends on several factors such as walking
`velocity, step frequency and height of walker etc. As the
`stride is not a constant and can change with speed, the
`step length parameter must be determined continuously
`during the walk to increase the precision. It is suggested
`that the stride could be estimated online based on a linear
`relationship between the measured step frequency and the
`stride (Levi and Judd, 1996). A real-time step calibration
`algorithm using a Kalman filter with GPS positioning
`measurement was also proposed (Jirawimut et al,2003).
`In this paper, we analyse a relationship between stride,
`step period and acceleration to obtain simple and robust
`method of stride determination. A real time online
`estimation
`is possible by using only 1-aixis
`accelerometer.
`The combination of gyroscope and magnetic compass has
`already been applied in car navigation (Mar and Leu,
`1996) and it might be a very useful heading sensor for
`pedestrian navigation system. However low cost sensor
`has important drawbacks: A low cost gyro has large bias
`
`and drift error. The magnetic disturbances can be induced
`fatal compass error. Moreover the error is occurred by an
`oscillation of human body in a walking behaviour. In this
`paper, a gyro and a magnetic compass are integrated
`using Kalman filter for reliable heading of pedestrian.
`To evaluate the performance of the proposed methods,
`actual walking
`test
`in
`the
`indoor environment
`is
`conducted. The equipment of walking test is implemented
`using a 1-aixs accelerometer, a 1-axis gyroscope and a
`magnetic compass. It consists of two parts: a sensor
`module and a navigation computer module. The sensor
`module is attached on the ankle. The step number and
`stride is computed using the output of the accelerometer
`on the sensor module. And walking direction is obtained
`from the gyro and magnetic compass module. The
`experiments show the very promising results: less than
`1% step detection error, less than 5% travelled distance
`error and less than 5% heading error.
`
`2 Step detection
`
`2.1 Step behaviour analysis of pedestrian
`
`A cycle of human walking is composed of two phases:
`standing and walking phase. The step detection means a
`recognition of walking phase. The walking phase is
`divided into a swing phase and a heel-touch-down phase.
`Each phase is shown in figure 1.
`
`Swing phase
`
`1st Swing phase
`
`2nd Swing phase
`
`Heel- touch- down phase
`
`Ground
`
`
`
`Fig. 1 A walking behaviour
`In the 1st swing phase, the foot is located on behind of
`gravity centre of human body. And the foot is located on
`front of gravity centre of human body in the 2nd swing
`phase. The foot accelerated during swing phase. The
`acceleration is composed of vertical and horizontal
`components as shown in figure 2, where a , h , g
`means horizontal acceleration, vertical acceleration and
`gravity force, respectively.
`Figure 3 and 4 show motion of leg in the 1st swing phase
`and the 2nd swing phase respectively.
`
`
`Niantic's Exhibit No. 1019
`Page 002
`
`
`
`
`
`
`
`
`Kim et al: A Step, Stride and Heading Determination for the Pedestrian Navigation System
`
`275
`
`2.0g. The pattern of acceleration signal in figure 5 is
`obtained from 625 times simulations.
`
`
`Fig. 5 Pattern of horizontal and vertical acceleration signal
`Figure 5 shows the typical pattern of acceleration signal
`on the swing phase. The acceleration of horizontal
`direction has 1 positive peak and 1 negative peak in
`swing phase while the acceleration of vertical direction
`has 1 negative peak only.
`The heel-touch-down phase follows the swing phase. A
`heel-touch-down is impact motion which hits the ground.
`In heel-touch-down phase, a heel hits the ground at first.
`And then a sole of foot and toe contact with the ground.
`When the foot hits the ground, the ground repulses the
`foot. At this time, impact force acts on the foot. The
`figure 6 shows the heel-touch-down phase.
`
`
`
`
`
`a
`
`h
`
`g
`
`Ground
`
`
`
`Fig. 2 Leg of walker
`
`VtA
`−)(
`
`direction
`
`θsina
`
`θ
`a
`sin
`
`
`θ+
`
`a
`
`
`( gh
`−
`
`)
`
`cos
`θ
`
`gh −
`(
`
`)
`
`θcos
`
`θ
`
`gh −
`
`Acceleration
`of
`Vertical direction
`
`HtA
`−)(
`
`direction
`
`a
`
`θcos
`
`θ
`a
`
`a
`
`cos
`θ
`
`−
`
`
`( gh
`−
`
`sin)
`
`θ
`
`θ
`
`gh −
`
`(
`
`gh −
`
`Acceleration
`θsin)
`of
`Horizontal direction
`
`Fig. 3 1st Swing phase
`
`VtA
`−)(
`
`direction
`
`HtA
`−)(
`
`direction
`
`aa
`
`θ
`
`θcos
`
`θ
`
`(
`
`gh
`−
`
`a
`
`
`
`
`sin) θθ cos+
`
`gh −
`(
`
`θsin)
`
`Acceleration
`of
`Horizontal direc tion
`
`gh −
`(
`
`)
`
`θcos
`
`a
`θ
`θsina
`
`θ
`
`gh −
`
`
`( gh
`−
`
`)
`
`a
`
`
`
`
`
`cos − θθ sin
`
`Acceleration
`of
`Vertical direction
`
`Fig. 4 2nd Swing phase
`The horizontal direction acceleration and vertical
`direction acceleration during the swing phase is denoted
`)(tθ is inclination angle of the leg
`in equation 1, where
`at time t.
`
` )(tA
`tA
`)(
`
`=
`=
`
`
`)(t
`gh
`(
`sin)
`θ
`−
`gh
`t
`(
`)(
`cos
`)
`θ
`−
`
`+
`−
`
`a
`a
`
`
`)(t
`cos
`θ
`t
`sin
`)(
`θ
`
`H
`
`direction
`
`−−
`
`Impact
`force
`
`Ground
`Repulsive
`Power
`
`
`
`Fig. 6 Heel-touch-down phase
`
`Ground Repulsive Power
`
`Impact force
`
`Vertical Axis
`
`Horizontal Axis
`
`
`Heel- touch- down
`Heel- touch- down
`Fig. 7 The typical pattern of signal in heel-touch-down phase
`
` (1)
`
`direction
`V
`In many researches, a step is declared when the measured
`HtA
`VtA
` or
` is
`larger
`than
`the
`−)(
`−)(
`direction
`direction
`)(tθ depend on
`threshold. However
`since
`the
`characteristics of walking which is different from each
`person, it is hard to determine the exact value of threshold
`HtA
`VtA
`of
`or
`. The step number is
`−)(
`−)(
`direction
`direction
`miscounted when wrongly predetermined threshold is
`applied. By using the signal pattern of acceleration, this
`problem can be solved. Typical signal pattern of
`acceleration is obtained from the computer simulation.
`We adopted common assumptions
`that a
`typical
`inclination of leg was within the limit of 30 degree ~ 50
`degree and a , h have a range of 0.8 ~ 2.3g and 0.6 ~
`
`Niantic's Exhibit No. 1019
`Page 003
`
`
`
`
`
`
`276
`
`Journal of Global Positioning Systems
`
`Figure 7 shows typical repulsive and impact force
`patterns during the heel-touch-down phase.
`By combining the swing phase and heel-touch-down
`phase in the figures 5 and 7, we obtain the signal pattern
`of one walking cycle. Figure 8 and 9 show entire signal
`pattern of the walking phase.
`
`F irst
`Swing
`Phase
`
`Second
`Swing
`Phase
`
`Heel
`Touch
`Down
`
`Time
`
`Acceleration
`
`
`
`
`
`Fig. 8 Vertical acceleration signal pattern in walking phase
`
`First
`Swing
`Phase
`
`Second
`Swing
`Phase
`
`Heel
`Touch
`Down
`
`Time
`
`Acceleration
`
`Fig. 9 Horizontal acceleration signal pattern walking phase
`
`
`It is expected intuitively that the period of heel-touch-
`down phase is much shorter than the period of swing
`phase. The figure 10 shows a real horizontal acceleration
`signal in one step. It coincides with the signal pattern
`model in figure 9.
`
`2.2 Step detection method
`
`To discriminate one cycle of walking behaviour, the
`signal pattern of swing phase and heel-touch-down phase
`in figure 8 and 9 is adopted. The accelerometer measures
`the signal which is caused by walking behaviour. The
`step number is counted when all three phases (1st swing,
`2nd swing and heel-touch-down phase) are detected. This
`method
`reduces step misdetection probability and
`increase reliability. Recognizing swing and heel-touch-
`down pattern using sequential multi-threshold gives a
`robust and reliable step detection. Also the method can
`reduce misdetection
`probability
`of
`non-walking
`behaviour such as sitting, turning, kicking and jumping
`etc. The detail detection algorithm is given in figure 11.
`START
`
`
`Input
`
`acceleration
`
`No
`
`No
`
`No
`
`No
`
`1st
`
`1st
`
`2nd
`
`Input >
`Upper Threshold
`Yes
`Input >
`Lower Threshold
`Yes
`Input >
`Upper Threshold
`Yes
`Input >
`Lower Threshold
`Yes
`Detection
`
`2nd
`
`Step
`
`END
`
`
`Fig. 11 Flow chart of step detection
`
`First
`Swing
`Phase
`
`Second
`Swing
`Phase
`
`Heel
`Touch
`Down
`
`3 Stride determination
`
`Because the stride is not a constant value and changes
`with speed, the stride parameter must be determined
`continuously during the walk to increase its precision.
`The stride relates on walking speed, walking frequency
`and acceleration magnitude. In typical human walking
`behaviour, a period of one step becomes shorter, a stride
`becomes larger and the vertical impact becomes bigger as
`the walking speed increases. The relation between stride,
`period of one step and acceleration is established thru the
`
`Fig. 10 Real horizontal acceleration signal
`
`
`
`Niantic's Exhibit No. 1019
`Page 004
`
`
`
`
`
`
`
`
`Kim et al: A Step, Stride and Heading Determination for the Pedestrian Navigation System
`
`277
`
`actual walking test. Figure 12 show test result of two type
`strides: 60 cm and 80 cm stride.
`
`
`Fig. 12 The acceleration signal of 60 cm and 80 cm stride
`The tester walks with the fixed stride using ground
`marks. In the figure, the relation between the acceleration
`and stride is clearly shown. The tables 1 and 2 show
`relation between acceleration and one step time in this
`test. The longer stride induces the bigger acceleration.
`However a difference of one step time is hard to apply
`stride determination because of small difference in
`measurements.
`Tab. 1 The mean of acceleration absolute value
`
`Stride
`
`60cm
`
`80cm
`
`Mean value (g)
`
`0.2882
`
`0.5549
`
`4 Heading determination
`
`The gyroscope and magnetic compass is widely used to
`determine heading. The characteristics of two sensors are
`summarized in Table 3, where advantage of one sensor is
`disadvantage of the other.
`Tab. 3 Comparison between compass and gyroscope
`
`
`
`Advantage
`
`Disadvantage
`
`Magnetic
`compass
`
`-absolute azimuth
`-long term stable accuracy
`
`unpredictable
`disturbances
`
`external
`
`Gyro
`scope
`
`-no external disturbances
`-short term accuracy
`
`relative azimuth drift
`
`
`From table 3, an optimal and reliable system might be
`expected by integrating the gyroscopes with the magnetic
`compass. In the integrated system, the gyroscope can
`correct the magnetic disturbances, at the same time the
`compass can determine and compensate the bias of the
`gyros and the initial orientation. The combination of
`gyroscope and magnetic compass has already been
`applied in the car navigation system. The integration
`method of the gyroscope and the magnetic compass used
`in this paper is given in Figure 13.
`
`Gyro bias
`
`Heading error
`
`Angular rate
`
`Gyroscope
`
`∫
`
`Heading
`
`Tab. 1 The period of one step
`
`Initial Heading
`
`Compass
`rate
`
`Disturbance
`detection
`
`Kalman
`Filter
`
`Magnetic
`compass
`
`Heading
`
`Stride
`
`60cm
`
`80cm
`
`Mean of time (sec.)
`
`0.675
`
`0.662
`
`
`Equation 2 is the experimental equation obtained from
`several walking tests, where means the measured
`acceleration and represents the number of sample in one
`cycle of walking. The equation represents the relation
`between measured acceleration and stride. It is used for
`online estimation of the stride.
`
`Stride
`
`(
`
`m
`
`)
`
`=
`
`98.0
`
`×
`
`3
`
`N
`
`k∑
`A
`1
`=
`N
`
`k
`
`
`
`(2)
`
`
`Fig. 13 Scheme for an integration of gyroscope and magnetic compass
`When the pedestrian is walking, the influence of
`magnetic disturbance sources changes unpredictably,
`creating a error in the compass heading. This error
`degrades the performance of integration system. The
`impact of error can be reduced by detecting the
`disturbance. The error can be observed via the angular
`rate of compass heading:
`t
`t
`)
`(
`ψ
`
`compass
`t
`∆
`t∆ is the
`where ω is angular rate, ψ is heading and
`time interval. The disturbance can be detected when a
`difference of compass angular rate
` and
`compassω
`gyroω is larger than given
`gyroscope angular rate
`threshold. The compass measurement is ignored. The
`states of Kalman filter are heading error and sensor error
`(gyro bias).
`
`ω
`compass
`
`=
`
`−∆+ ψ
`compass
`
`k
`
`t
`(
`
`k
`
`)
`
`
`
`(3)
`
`Niantic's Exhibit No. 1019
`Page 005
`
`
`
`
`
`
`278
`
`5 Experiments
`
`Journal of Global Positioning Systems
`
`Fig. 15 The output signal of accelerometer
`
`Fig. 16 Estimated stride in 1st test
`
`
`
`
`
`
`
`Fig. 17 Estimated stride in 2nd test
`In the 1st test, the proposed method count step number
`without loss, while the 2 step detection is lost in 2nd test.
`The 2 step loss is happened in the last 199th and 200th
`step where the walking pattern is abruptly changing. The
`walking distance error
`is obtained 2.5m, 6.1m
`respectively. The travelled distance with less than 5%
`error is obtained. These results verify that the proposed
`method can measure accurate step numbers and distance.
`
`
`
`In order to evaluate the performance of the proposed
`method, the actual walking test is done. The tester is a
`male aged 26 with 175cm height. The experiments are
`done at the 4th floor hallway of the engineering building,
`Chungnam National University, Daejeon, Korea. In the
`experiments, walking distance determination and heading
`determination are carried out separately.
`
`5.1 Experimental setup
`
`Figure 14 shows the experimental equipments.
`
`Navigation Computer
`
`Sensor Module
`
`Data Acquisition
`
`Bluetooth
`
`
`
`C ommunication
`
`C ompass
`
`16- bit
`Microcontroller
`
`MEMORY
`
`POWER
`
`Accelermeter
`
`Gyro
`
`Fig. 14 Experimental equipment
`The experimental equipments consist of the sensing
`module, the navigation computer and a data acquisition
`system (notebook computer). The body-worn sensing
`module consists of a 16-bit microcontroller, a MEMS
`accelerometer
`(ADXL105, Analog device
`Inc.), a
`gyroscope (MEMS DMU, Crossbow Inc.), a low-cost
`digital magnetic compass sensor (CMPS03, ROBOT
`Electronics Inc.) and other electrical parts (RS-232
`converter, DC-DC converter, 9V battery, Bluetooth
`modem). The sensor module is attached on the ankle with
`horizontal direction as shown figure 14.
`
`5.2 Experiment of walking distance determination
`
`To evaluate performance of the step detection and stride
`determination algorithm, the tester was asked to walk for
`pre-determined path (74.2m and 145.6m straight path).
`Figure 15 shows the output of accelerometer.
`The true step number of first test is 100 steps and second
`test 200 steps. The stride is determined using equation 2.
`The figure 16 and 17 show the strides of left leg.
`The mean of estimated stride is obtained as 76.1 cm and
`75.9 cm respectively. Table 4 shows result of walking test
`in detail.
`
`Niantic's Exhibit No. 1019
`Page 006
`
`
`
`
`
`
`
`
`
`
`Kim et al: A Step, Stride and Heading Determination for the Pedestrian Navigation System
`
`279
`
`Tab. 4 The measured walking distance
`
`Actual walking behavior
`
`1st test
`
`2nd test
`
`Step
`number
`
`Walking
`distance
`
`Step
`number
`
`Walking
`distance
`
`100 step
`
`74.2m
`
`200 step
`
`145.6m
`
`Measured
`step
`number
`
`Measured
`walking
`distance
`
`100 step
`
`76.728 m
`
`198 step
`
`151.674 m
`
`the vertical and horizontal
`detection, we analyse
`acceleration of the foot during one step of the walking.
`With this analysis, a new step determination based on the
`pattern recognition is proposed and the step number can
`be counted accurately. The relationship between stride
`and acceleration is derived from actual test. An efficient
`stride determination method where the stride can be
`estimated online, so that the user does not need to specify
`his/her stride, is proposed. The integration scheme of the
`gyro and magnetic compass is proposed for error
`compensation of gyro and disturbance rejection of
`magnetic compass. The experiments using the actual
`walking tests in indoor shows that the proposed method
`gives less than 1% step, 5% travelled distance and 5%
`heading errors. It is expected that the proposed PNS will
`be very useful navigation system
`for pedestrian
`navigation.
`
`5.3 Experiment of heading determination
`
`References
`
`For heading determination test, the tester walks a straight
`path of north direction. Figure 18 is result of heading
`determination test.
`
`
`
`Fig. 18 Estimated heading by gyro and by integration
`In the figure, the heading of stand-alone gyro shows
`oscillatory errors due to the body motion. Kalman filter in
`the
`integrated system
`reduces
`these errors. The
`experiments show that the heading of pedestrian can be
`determined with accuracy of 5 degree.
`
`6 Conclusions and outlook
`
`This paper proposes methods to estimate the PNS DR
`parameters: step, stride and heading. For accurate step
`
`
`Gabaglio, V., (2003): GPS/INS Integration for Pedestrian
`Navigation Ph. D. dissertation. Institute of Geomatics of
`the Swiss Federal Institute of Technologye in Lausanne.
`Mar, J., (1996): Simulations of the positioning accuracy of
`integrated vehicular navigation systems.
`In: J.-H.
`Leu(Eds.): Proc. Inst. Elect. Eng. Radar, Sonar Navigation,
`vol. 143, Apr., 121–128.
`Ladetto, Q., (2002): In Step with INS. In: B. Merminod
`(Eds.):GPS WORLD magazine, 30-38
`Quentin Ladetto (2000): On foot navigation: continuous step
`calibration using
`both
`complementary
`recursive
`prediction and adaptive Kalman filtering. Proceedings of
`ION GPS 2000, 1735~1740.
`Jirawimut, R., (2003): A Method for Dead Reckoning
`Parameter Correction in Pedestrian Navigation System.
`In: P. Ptasinski; V. Garaj; F. Cecelja; W.Balachandran
`(Eds.):IEEE Transactions
`on
`Instrumentation
`and
`Measurement , Vol. 52, No.1.
`Levi, R. W., (1996): Dead Reckoning Navigational System
`using Accelerometer to Measure Foot Impacts. In: T.
`Judd,(Eds.): United States Patent No. 5,583,776
`Lee, S.-W., (2001): Recognition of Walking Behaviours for
`Pedestrian Navigation. In: K. Mase,(Eds.): Proc. 2001
`IEEE Int’l Conf. Control Applications (CCA 01), IEEE
`Control Systems Soc., Piscataway, N.J., 1152–1155.
`Gabaglio, V., (1999): Real-time calibration of length of steps
`with GPS and accelerometers.
`In: B. Merminod
`(Eds.):Proceeding of GNSS 99, 599–605.
`
`Niantic's Exhibit No. 1019
`Page 007
`
`