In Cluttered Indoor Environments Through Probabilistic Classification Methods Rutgers University Chenren Xu Joint work with Bernhard Firner Yanyong Zhang Richard Howard Jun Li ID: 798902
Download The PPT/PDF document "Improving RF-Based Device-Free Passive ..." is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Slide1
Improving RF-Based Device-Free Passive Localization In Cluttered Indoor Environments Through Probabilistic Classification Methods
Rutgers University
Chenren
Xu
Joint work with
Bernhard
Firner
,
Yanyong
Zhang
Richard Howard, Jun Li,
Xiaodong
Lin
Slide2Passive LocalizationMotivationIndoor challenge
Proposed solution
Experimental methodology
Performance evaluationConclusion and future work
2
Slide3RF-Based Localization 3
Active Localization
Slide4RF-Based Localization 4
Slide5RF-Based Localization 5
Passive Localization
Slide6Passive LocalizationMotivationIndoor challenge
Proposed solution
Experimental methodology
Performance
e
valuation
Conclusion and future work
6
Slide7Why Passive Localization? Monitor indoor human mobility
Elder/health care
7
Slide8Why Passive Localization? Monitor indoor human mobility
Detect traffic flow
8
Slide9Why Passive Localization? Monitor indoor human mobility
Health/elder care, safety
Detect traffic flow
Provides
privacy
protection
No identification
Use existing wireless infrastructure
9
Slide10Passive LocalizationMotivationIndoor challenge
Proposed solution
Experimental methodology
Performance
e
valuation
Conclusion and future work
10
Slide11Multipath Effect11
Slide12Multipath Effect12
Slide13Multipath Effect13
Slide14Cluttered Indoor Scenario14
Slide15Cluttered Indoor Scenario
15
A
u
ser steps
across one
Line-of-Sight
Slide16Cluttered Indoor Scenario
16
A
u
ser steps
across one
Line-of-Sight
RSS fluctuates in a unpredictable fashion
Slide17Cluttered Indoor Scenario
17
The RSS change can either go up to 12
dBm
Slide18Cluttered Indoor Scenario
18
Or go down to -12
dBm
Slide19Cluttered Indoor Scenario
19
These two
peak points
can have 24 dB difference
in energy within
only 2 meters
.
Slide20Cluttered Indoor Scenario
20
We
also observe that these two points within
0.2
m can have 15
dB
difference
.
Deep fade
Slide21Cluttered Indoor Scenario
21
Slide22Cluttered Indoor Scenario
22
Slide23Cluttered Indoor Scenario
23
Slide24Passive LocalizationMotivation
Indoor challenge
Proposed solution
Experimental methodologyPerformance
e
valuation
Conclusion and future work
24
Slide25Proposed Solution
High dimensional space
Measure radio signal strength (RSS) changes in multiple transmitter and receiver links.
25
Link T1 – R1
Link T2 – R2
Slide26Proposed SolutionHigh dimensional space
Cell-based localization
Flexible precision
Classification approach26
Slide27Linear Discriminant Analysis RSS measurements with user’s presence in each cell is treated as a class
k
Each
class k is Multivariate Gaussian with common covariance
Linear discriminant function:
27
Link 1 RSS (
dBm
)
Link 2 RSS (
dBm
)
k
= 1
k
= 2
k
= 3
Slide28Proposed SolutionHigh dimensional space
Cell-based
localization
Lower radio frequencySmooth the spatial variation
28
Slide29Frequency Impact
29
RSS changes smoother on 433.1 MHz than on 909.1 MHz
Slide30Frequency Impact
30
Less deep fading points!
Slide31Proposed SolutionHigh dimensional
space
Find features with fewer deep fading points
Cell-based localizationAverage the deep fading
effect
Lower radio frequency
Reduce
the deep fading
points
31
Mitigate the error caused by the
multipath effect!
Slide32Passive LocalizationMotivation
Indoor challenge
Proposed solution
Experimental methodologyPerformance
e
valuation
Conclusion and future work
32
Slide33Experimental Deployment33
Total Size:
5 × 8 m
Slide34Experimental Deployment 34
Slide35System Parameters35
Parameter
Default value
Meaning
K
32
Number of cells
P
64
Number
of pair-wise radio links
N
trn
100
Number of training
data per cell
N
tst
100
Number of testing
data per cell
Slide36System DescriptionHardware: PIP tagMicroprocessor: C8051F321
Radio chip: CC1100
Power: Lithium coin cell battery (~1 year)
Protocol: Unidirectional heartbeat (Uni-HB)
Packet size: 10 bytes
Beacon interval: 100 millisecond
36
Slide37Training MethodologyCase A: stand still at the each cell centerMeasurement only involves center of the cell
Ignore the deep fade points
Case B: random walk within each cell
Measurement includes all the spaceAverage the multi-path effects
37
Training only takes 15
mins
!
Slide38Passive LocalizationMotivation
Indoor challenge
Proposed solution
Experimental methodologyPerformance
e
valuation
Conclusion and future work
38
Slide39MetricsCell estimation accuracyThe ratio of successful cell estimations with respect to the total number of estimations.
Average error distance
Average distance between the actual location and the estimated cell’s center.
39
Slide40Localization AccuracyCell estimation accuracy:
40
Stand still
at each cell center
Random walk with in each cell
433.1 MHz
90.1
%
97.2%
909.1 MHz
82.9%
93.8%
97.2 % cell estimation accuracy with 0.36 m average error distance
Slide41Reducing Training Dataset41
100
8
Only 8 samples are good enough
Slide42Robust to Link Failure42
5 transmitter +
3 receivers =
90% cell estimation accuracy
Slide43Long-term Stability43
Slide44Multiple Subjects Localization44
Slide45Larger Deployment45
Total Size: 10 × 15 m Cell Size:
2
× 2 m13 transmitters and 9 receivers
Slide46Larger Deployment46
Cell estimation accuracy: 93.8%
Average error distance: 1.3 m
Slide47Passive LocalizationMotivation
Indoor challenge
Proposed solution
Experimental methodologyPerformance
e
valuation
Conclusion and future work
47
Slide48Conclusion and Future Work ConclusionWe propose a general probabilistic classification framework to solve the passive localization problem with:
High accuracy, low cost, robust and stable
M
ultiple subjects tracking generalization Future work
Improving multiple people tracking
Passively detect the number of people
48
Slide49Q & AThank you
49