WalkCompass:
I am a Smartphone and I can Tell my User’s Walking Direction
Nirupam Roy
!
He Wang Romit Roy Choudhury
!
MobiSys 2014, Bretton Woods, NH
WalkCompass: I am a Smartphone and I can Tell my Users Walking - - PowerPoint PPT Presentation
WalkCompass: I am a Smartphone and I can Tell my Users Walking Direction MobiSys 2014, Bretton Woods, NH Nirupam Roy He Wang Romit Roy Choudhury ! ! Indoor Localization: Chasing Accuracy Location Accuracy 2 Meter 50 Centimeter 3 Meter
Nirupam Roy
!
He Wang Romit Roy Choudhury
!
MobiSys 2014, Bretton Woods, NH
Location Accuracy
2000 2007 2010 2013 2004
3 Meter 2 Meter 1 Meter 50 Centimeter
Indoor Localization: Chasing Accuracy
Mean
Location Accuracy
2000 2007 2010 2013 2004
3 Meter 2 Meter 1 Meter 50 Centimeter
Indoor Localization: Chasing Accuracy
Mean 95th Percentile
Step Length Walking Direction Device Orientation Hardware
Homogeneity
Map/ Floor plan
Walking Direction
Walking Direction: The Definition
Walking Direction
8!
9!
Walking Direction: The Definition
10!
Why Not Compass?
Y axis
60° NE 60° NE
N
Why Not Compass?
Y axis
60° NE 60° NE
12!
30° NE
Y axis
N N
Why Not Compass?
Y axis
60° NE 60° NE
13!
30° NE
Y axis
N N
Walking Direction Walking Direction
14!
Compass Direction Walking Direction Why Not Compass?
Walking Direction
15!
Force Force
16!
Challenge 1: Finding the Right Force
17!
Complex combination of forces
Different walking patterns Different phone locations
18!
Challenge 2: Finding Common Solution
Raw accelerometer signal Phone Location: Pants pocket
19!
Understanding a Walk Signal
Raw accelerometer signal Phone Location: Pants pocket
20!
Understanding a Walk Signal
Foot!step! Leg!swing! Body!movement! Leg!touches! ground!
Segmentation of the Signal: Key Intuition
21!
22!
Anatomy of a Human Walk
Swing Phase (40%) Stance Phase (60%)
Heel-touch (Right) Heel-touch (Left) Heel-touch (Right)
23!
Raw Accelerometer Signal
Swing Phase (40%) Stance Phase (60%)
24!
High Frequency Components
Anatomy of a Human Walk
Heel-touch (Right) Heel-touch (Left) Heel-touch (Right)
Heel-touch Detection Signal
Swing Phase (40%) Stance Phase (60%)
25!
Anatomy of a Human Walk
Heel-touch (Right) Heel-touch (Left) Heel-touch (Right)
Raw Accelerometer Signal
Swing Phase (40%) Stance Phase (60%)
26!
Anatomy of a Human Walk
Low Frequency Components
Swing Phase (40%) Stance Phase (60%)
27!
Anatomy of a Human Walk
Swing Phase (40%) Stance Phase (60%)
28!
Anatomy of a Human Walk
Swing Phase (40%) Stance Phase (60%)
29!
Anatomy of a Human Walk
Swing Phase (40%) Stance Phase (60%)
30!
Anatomy of a Human Walk
W3D
31!
Walking Direction Vector
Y
W2D
32!
W3D Walking Direction Vector
33!
Global Walking Direction
60° North-East
W2D
34!
35!
Global Walking Direction W2D
Compass: How Reliable is this?
36!
37!
Compass: How It Works
38!
Compass: Effect of Interference
39!
Ferromagnetic Material
Compass: Effect of Interference
40!
Ferromagnetic Material
Resultant
Geo-mag G I R = + Compass: Effect of Interference G I R =
Iterative Magnetic Triangulation (IMT)
42!
43!
Gcorrect R I Iterative Magnetic Triangulation (IMT) Gcorrect I R =
44!
Gincorrect R I Iterative Magnetic Triangulation (IMT) Gincorrect I R =
45!
Methodology
Error in Walking Direction
10 20 30 50 100 150 Number of Steps Absolute Error (degree) Median 75th Percentile
47!
Error in Walking Direction
10 20 30 50 100 150 Number of Steps Absolute Error (degree) Median 75th Percentile
48!
Error!converges!to!8°!! within!6!steps!
30 210 60 240 90 270 120 300 150 330 180 30 210 60 240 90 270 120 300 150 330 180
Better Compass
Today’s Compass Corrected Compass (IMT Algorithm)
49!
30 210 60 240 90 270 120 300 150 330 180 30 210 60 240 90 270 120 300 150 330 180
Better Compass
Today’s Compass Corrected Compass (IMT Algorithm)
50!
Error!±40°! Error!±10°!
Discussion and Future Work
51!
Conclusion
52!
hCp://synrg.csl.illinois.edu!
hCp://synrg.csl.illinois.edu/projects/localizaSon/ walkcompass!