Pedestrian Dead Reckoning as an Indoor Positioning System,Step detection and step estimation ,kims approach ,Smartphone-based Pedestrian Dead Reckoning

1. 1WILCO 2017
Smartphone-based Pedestrian
Dead Reckoning
as an Indoor Positioning System

2. 2WILCO 2017
Introduction to PDR
• Positioning is the technique to know object’s position in a frame of reference.
• Position can be done
 Global Positioning System (GPS)
 Base Transceiver Station (BTS)
 GPS satellite signal dependence  Can't Used in the Buildings
 BTS cell-Phone  Very small accuracy
• Indoor positioning is very important when the user needs to know its position in
a building.
 Example : Firefighters during rescuing effort.
• An alternative of indoor positioning is Pedestrian Dead Rocking (PDR).
• PDR determines the latest position of a pedestrian by adding estimated
displacement to starting known position .
• Displacement is represented by amount of steps and each steps has its various
step length.
• Detection number of steps and estimation of step length can be done using
accelerometer sensor.

3. 3WILCO 2017
PDR for Indoor Positioning

4. 4WILCO 2017
Pedestrian Dead Reckoning (PDR)
 PDR is a pedestrian position solution by adding distance travelled to the
known starting position.
 Pedestrian distance travelled can be determined by using accelerometer
sensor to detect steps and estimate displacement.
 Accelerometer sensor must be attached to the body to record acceleration.
 Special sensor modules is attached on the helmet
 Attached at the foot
 Low cost sensor integrated in smartphone and placed it to the trouser
pocket.
 The implementation of PDR techniques includes :
 Orientation Projection
 Filtering
 Step detection
 Step length estimation

5. 5WILCO 2017

6. 6WILCO 2017
Accelerometer output

7. 7WILCO 2017
Normalized output

8. 8WILCO 2017
Filtering
 The acceleration signal must be filtered to obtain the desired output signal
 Gravity-free signal  Low frequency signal component that causing offset shift up
the y-axis about 9.8 𝑚/𝑠2
 To eliminate the influence of gravity the signal is filtered with high pass filtering
𝑎𝑐𝑐_𝐻𝑃𝑎𝑣𝑔 = 𝑎𝑐𝑐_𝑛𝑒𝑤∗ 1 −∝ 𝑎𝑐𝑐_𝐻𝑃𝑎𝑣𝑔∗ ∝
𝑎𝑐𝑐 𝐻𝑃𝑓𝑖𝑙𝑡𝑒𝑟𝑒𝑑 = 𝑎𝑐𝑐 _𝑛𝑒𝑤 − 𝑎𝑐𝑐_𝐻𝑃𝑎𝑣𝑔∗
 Low frequency signal component is subtracted to remove DC component
 The output of high-pass filtering then processed by low pass filtering to smooth
the signal and reducing random noise.
 Low pass filtering has done by using a moving average filter
𝑦 𝑖 =
1
𝑀
𝑗=−(𝑀−1)/2
(𝑀−1)/2
𝑥(𝑖 + 𝑗)
𝑦[] and 𝑥[] are output average filtered and input non filtered signal . M is moving
window, the number of points used in the moving average

9. 9WILCO 2017

10. 10WILCO 2017
Step detection
 Two common step detection methods
1. Peak detection
2. Zero crossing detection
 The zero crossing method counts signal crossing zero level to determine the
occurrence of step.
 This method is not appropriate to detect user’s steps it requires certain time
interval threshold to make decision whether the zero crossing represents a valid
step or not. (Need calibration process)
 Peak detection scheme detects a step when valid maximum peak(maxima) and
valid minimum peak (minima) are detected in sequence in a certain interval.
 Maxima Maximum peaks exceeds upper threshold.
 Minima Minimum peaks lower than lower threshold.
 Upper threshold  Summed last valid minima with ∆ 𝒕𝒉𝒓𝒆𝒔𝒉𝒐𝒍𝒅 value
 Lower threshold Subtracted last valid maxima with ∆ 𝒕𝒉𝒓𝒆𝒔𝒉𝒐𝒍𝒅 value
 ∆ 𝒕𝒉𝒓𝒆𝒔𝒉𝒐𝒍𝒅 is a constant value that determined experimentally

11. 11WILCO 2017
Step peak detection

12. 12WILCO 2017
Step Length Estimation
 Total travelled distance can be calculated by estimating step length in every valid
detected step.
 Two methods:
1. Static Method
2. Dynamic method
 Static method assumes any valid steps having the same length
stepsize = Height . K
k=0.415 for men and 0.413 for women
 Dynamic method assumes any valid steps having their different step length which can be
estimated using certain approaches.
 Kim approach : propose an experimental equation which is representing a relation
between step length and average acceleration which is occur during a step .
𝑆𝑡𝑒𝑝𝑠𝑖𝑧𝑒 = 𝑘
3
𝑘=1
𝑁
𝑎 𝑘
𝑁

13. 13WILCO 2017
Step Length Estimation

14. 14WILCO 2017
Flow chart of overall the positioning system

15. 15WILCO 2017
Future work
 Heading Estimation
 Heading estimation plays a significant role in indoor environments.
 The most common method is the magnetometer-based heading solution.
 Heading Estimation Using a Gyroscope.
Comparison between compass and gyroscope

16. 16WILCO 2017
References
 A. R. Pratama, Widyawan and R. Hidayat, "Smartphone-based Pedestrian Dead
Reckoning as an indoor positioning system," 2012 International Conference on
System Engineering and Technology (ICSET), Bandung, 2012, pp. 1-6.
 I. Bylemans, M. Weyn, and M. Klepal, “Mobile Phone-based Displacement
Estimation for Opportunistic Localisation Systems,” in Third International
Conference on Mobile Ubiquitous Computing, Systems, Services and
Technologies, 2009.
 J. W. Kim, H. J. Jang, D-H. Hwang, and C. Park, “A Step, Stride and Heading
Determination for the Pedestrian Navigation System, "Journal of Global
Positioning Systems, pp. 273-279, 2004.
 Xuebing Yuan,Shuai Yu,Shengzhi Zhang,Guoping Wang and Sheng Liu “Quaternion
-Based Unscented Kalman Filter for Accurate Indoor Heading Estimation Using
Wearable Multi-Sensor System”