using FM radio signals Andrei Popleteev Advisors: Oscar Mayora - - PowerPoint PPT Presentation

▶
using fm radio signals
SMART_READER_LITE
LIVE PREVIEW

using FM radio signals Andrei Popleteev Advisors: Oscar Mayora - - PowerPoint PPT Presentation

Oct 09, 2023 209 likes •992 views

Indoor positioning using FM radio signals Andrei Popleteev Advisors: Oscar Mayora Venet Osmani Outline Introduction State of the art Proposed approach FM localization With local transmitters (FM L ) With broadcasting

slide-1
SLIDE 1

Indoor positioning using FM radio signals

Andrei Popleteev

Advisors: Oscar Mayora Venet Osmani

slide-2
SLIDE 2

2 21 April 2011

Outline

  • Introduction
  • State of the art
  • Proposed approach
  • FM localization

– With local transmitters (FML) – With broadcasting stations (FMB)

  • Conclusion
slide-3
SLIDE 3

3 21 April 2011

Indoor localization

  • Ambient intelligence
  • Assisted daily living
  • Activity recognition
  • Behavior analysis
  • Object tracking
slide-4
SLIDE 4

4 21 April 2011

Indoor localization

  • GPS does not work indoors.
  • Specialized systems are expensive.
  • Systems based on cellular networks:

– Good coverage – Low accuracy

  • Wi-Fi is the de-facto standard, but

– Limited coverage – High power consumption

slide-5
SLIDE 5

5 21 April 2011

Indoor localization: FM radio

  • FM radio addresses these issues, and provides:

– High coverage – Long battery life – Good accuracy

slide-6
SLIDE 6

6 21 April 2011

Source: fmscan.org

slide-7
SLIDE 7

7 21 April 2011

FM-enabled mobile devices

slide-8
SLIDE 8

8 21 April 2011

Power consumption

1200 300 50 15 200 400 600 800 1000 1200 1400 Wi-Fi (constantly active) Wi-Fi (power saving) FM receiver (with RDS) FM receiver (without RDS) Power consumption, mW

Wi-Fi data from [Anand et al. 2005] FM data from Si4703 and TDA7088 datasheets

slide-9
SLIDE 9

9 21 April 2011

State of the art

slide-10
SLIDE 10

10 21 April 2011

State of the art: FM localization

  • There are few works on FM positioning.
  • All of them consider only outdoor scenarios.
  • Achieved accuracy:

– 2005: 8 km with 50% probability (Krumm et al.) – 2009: 20 m with 67% probability (Fang et al.)

There are no results for indoors performance of FM localization.

slide-11
SLIDE 11

11 21 April 2011

State of the art: Summary

Technology Accuracy Coverage Battery life System costs Wi-Fi Medium Low Low Low Cellular Low Medium Low Low UWB High Low High High FM (outdoor) Low High High Low FM (indoor) ?

The Gap

slide-12
SLIDE 12

12 21 April 2011

Localization methods

  • Proximity-based
  • Direction-based
  • Time-based
  • Based on signal properties

– Propagation modeling – Fingerprinting Used in this work

slide-13
SLIDE 13

13 21 April 2011

Fingerprinting

Includes two phases:

  • Calibration: creation of a database matching

signal strength samples with the location.

  • Positioning: comparing the observed signal

properties to those in the database.

slide-14
SLIDE 14

14 21 April 2011

Proposed approach

slide-15
SLIDE 15

15 21 April 2011

FM radio signal sources

  • Short-range FM transmitters

– Off-the-shelf devices – No licensing required – Can transmit arbitrary sound

  • Broadcasting FM stations

– Zero cost for localization – Worldwide coverage

  • Both signal sources have been

used in this thesis

slide-16
SLIDE 16

16 21 April 2011

Experimental setup

12 m 6 m

UBiNT lab Create-Net

slide-17
SLIDE 17

17 21 April 2011

FML: positioning using local transmitters

slide-18
SLIDE 18

18 21 April 2011

FML: positioning using local transmitters

  • FML performance
  • FML vs. Wi-Fi
  • Orientation analysis
  • Accuracy degradation
slide-19
SLIDE 19

19 21 April 2011

FML positioning

  • Suitable signal features for fingerprinting:

– Received signal strength (RSS) – Audio signal-to-noise ratio (SNR) – Stereo channel separation (SCS)

slide-20
SLIDE 20

20 21 April 2011

Signal properties vs. distance

0.0 0.2 0.4 0.6 0.8 1.0 1 2 3 4 5 6 7 8 Normalized value Distance from transmitter, meters

SNR SCS RSSI

Receiver: Brando USB FM radio

slide-21
SLIDE 21

21 21 April 2011

FML positioning performance

0.0 0.2 0.4 0.6 0.8 1.0 1 2 3 4 5 6 7 8 Confidence Error distance, meters SNR SCS RSSI Baseline

Receiver: Brando USB FM radio; grid: 1 m.

slide-22
SLIDE 22

22 21 April 2011

FML positioning accuracy (RSSI)

2.65m @ 95% 0.97m @ 50%

1 2 3 4 5 6 7 8 9 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Error, meters Confidence

Gaussian Processes

3.88m @ 95% 0.93m @ 50%

1 2 3 4 5 6 7 8 9 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Error, meters Confidence

kNN

Receiver: HTC Artemis; grid: 0.5 m.

slide-23
SLIDE 23

23 21 April 2011

FML: positioning using local transmitters

  • FML positioning
  • FML vs. Wi-Fi
  • Orientation analysis
  • Accuracy degradation
slide-24
SLIDE 24

24 21 April 2011

FML versus Wi-Fi

FM Wi-Fi

1 2 3 4 5 6 7 8 9 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Error, meters Confidence

Gaussian Processes

1 2 3 4 5 6 7 8 9 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Error, meters Confidence

kNN

Receiver: HTC Artemis; grid: 1 m. FM RSSI granularity reduced to ensure a fair comparison.

slide-25
SLIDE 25

25 21 April 2011

FML: positioning using local transmitters

  • FML positioning
  • FML vs. Wi-Fi
  • Orientation analysis
  • Accuracy degradation
slide-26
SLIDE 26

26 21 April 2011

Effect of orientation

  • Human body influences the signal distribution

by reflecting and attenuating radio waves.

  • This might impact the localization accuracy.

– It does for Wi-Fi. – Does it for FM?

slide-27
SLIDE 27

27 21 April 2011

Effect of orientation

  • Four datasets collected,
  • ne for each direction.
  • “All FM” – accuracy when

all four datasets are utilized.

  • Other graphs - accuracy

within each dataset.

Confidence Error distance, meters

 User direction has no significant effect on FM localization accuracy.

slide-28
SLIDE 28

28 21 April 2011

Recognition of orientation

  • Is it possible to

detect the orientation using FM RSS fingerprints?

10 20 30 40 50 60 Exact Adjacent Opposite Probability, % Recognized orientation

 No, the result is random.

slide-29
SLIDE 29

29 21 April 2011

FML: positioning using local transmitters

  • FML positioning
  • FML vs. Wi-Fi
  • Orientation analysis
  • Accuracy degradation
slide-30
SLIDE 30

30 21 April 2011

What if…

slide-31
SLIDE 31

31 21 April 2011

Signal strength distribution

RSSI coordinate

Before:

RSSI coordinate

Now:

slide-32
SLIDE 32

32 21 April 2011

Accuracy degradation

  • Signal fingerprints change with time due to:

– Furniture layout – Air temperature and humidity – Hardware temperature

  • These fluctuations affect the accuracy.
  • The solution: periodic recalibration

– Requires personnel or additional hardware – Is tedious and expensive

slide-33
SLIDE 33

33 21 April 2011

  • Recalibration performed automatically

when the device position is known:

– In a cradle – On a nightstand – Connected to a wall charger

  • No additional hardware required
  • Transparent for the user

Spontaneous recalibration

slide-34
SLIDE 34

34 21 April 2011

1 2 3 4 5 6 7 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Error, meters Confidence

Effect of recalibration

Original Degraded Recalibrated

slide-35
SLIDE 35

35 21 April 2011

FMB: positioning using broadcasting FM stations

slide-36
SLIDE 36

36 21 April 2011

FMB: positioning using broadcasting FM stations

  • FMB performance
  • FMB vs. Wi-Fi and GSM
  • Signal stability and people’s presence
  • Power consumption
slide-37
SLIDE 37

37 21 April 2011

FMB experiments

  • Performed in the same 12x6 m testbed

(with slightly changed layout).

  • 76 active FM stations detected.
  • 3 local FM transmitters for comparison.
  • KNN classifier, leave-one-out evaluation.
slide-38
SLIDE 38

38 21 April 2011

FMB localization performance

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1 2 3 4 5 6 7 8 9

Confidence Error distance, meters

FMb (all beacons) FML (3 beacons) Baseline

0.91m @ 50% 4.71m @ 95%

slide-39
SLIDE 39

39 21 April 2011

FM station selection

  • More stations in fingerprint result in:

– More accurate localization, but – Higher computational load – Longer scanning times

  • Do all the stations contribute equally?
  • Is there a trade-off between the number of

stations and localization performance?

slide-40
SLIDE 40

40 21 April 2011

Station selection methods

  • Naïve approach: select stations with

– strongest signals; – weakest signals.

  • Alternative approach: select the stations which

vary the most across the test points.

slide-41
SLIDE 41

41 21 April 2011

Station selection methods

1 2 3 4 5 6 1 7 13 19 25 31 37 43 49 55 61 67 73

Median error, meters Fingerprint width (number of stations)

Strongest Weakest Highest diversity

slide-42
SLIDE 42

42 21 April 2011

FMB with 10% of stations

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1 2 3 4 5 6 7 8 9

Confidence Error distance, meters

FMb (all beacons) FMb (7 beacons) Baseline

slide-43
SLIDE 43

43 21 April 2011

FMB: positioning using broadcasting FM stations

  • FMB performance
  • FMB vs. Wi-Fi and GSM
  • Signal stability and people’s presence
  • Power consumption
slide-44
SLIDE 44

44 21 April 2011

FMB versus Wi-Fi

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1 2 3 4 5 6 7 8 9

Confidence Error distance, meters

FMb (all beacons) Wi-Fi (all beacons) Baseline

76 FM and 17 distinct Wi-Fi beacons

slide-45
SLIDE 45

45 21 April 2011

FMB versus GSM

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1 2 3 4 5 6 7 8 9

Confidence Error distance, meters

FMb (all beacons) FMb (7 beacons) GSM (all beacons) Baseline

15 distinct GSM beacons GSM RSSI acquired with HTC Artemis

slide-46
SLIDE 46

46 21 April 2011

FMB localization: Summary

Localization accuracy for different technologies (in meters) measured in the same conditions.

Confidence FMB Wi-Fi GSM FMB (7 stations) 50% 0.9 1.6 3.1 1.3 67% 1.3 1.9 4.2 2.1 90% 3.4 3.5 6.2 4.0 95% 4.7 4.0 9.1 4.9

slide-47
SLIDE 47

47 21 April 2011

FMB: positioning using broadcasting FM stations

  • FMB performance
  • FMB vs. Wi-Fi and GSM
  • Signal stability and people’s presence
  • Power consumption
slide-48
SLIDE 48

48 21 April 2011

Signal stability analysis

  • Human bodies interact with radio waves.
  • Thus, people are an unpredictable factor that

influences signal distribution and thus localization performance.

  • FM radio waves are longer than Wi-Fi waves –

this leads to differences in signal propagation.

slide-49
SLIDE 49

49 21 April 2011

Signal stability: Experiment 1

  • Environment: a student mensa

– Lunch time (crowded) – Evening (empty)

  • 50 minutes duration; 84 fingerprints

– 26 FM stations – 5 Wi-Fi access points

  • RSS samples normalized according to device’s

minimum and maximum values.

slide-50
SLIDE 50

50 21 April 2011

Signal stability: Experiment 1

Empty Crowded

0.00 0.01 0.02 0.03 0.04 0.0 0.1 0.2 0.3 0.4

RSSI s.d. Mean RSSI

0.00 0.01 0.02 0.03 0.04 0.0 0.1 0.2 0.3 0.4

RSSI s.d. Mean RSSI FM Wi-Fi

slide-51
SLIDE 51

51 21 April 2011

Signal stability: Experiment 2

  • An office environment:

– Empty, daytime – Empty, nighttime – Populated

  • 6 hours duration; 592 fingerprints

– 23 FM stations – 13 Wi-Fi access points

  • RSS normalized as in the previous experiment.
slide-52
SLIDE 52

52 21 April 2011

Signal stability: Experiment 2

Wi-Fi FM

0.00 0.02 0.04 0.06 0.08 0.10 0.1 0.2 0.3 0.4 0.5 0.6

s.d. Mean

Empty, day Empty, night Populated, day 0.000 0.005 0.010 0.015 0.020 0.05 0.1 0.15 0.2 0.25

s.d. Mean

slide-53
SLIDE 53

53 21 April 2011

FMB: positioning using broadcasting FM stations

  • FMB performance
  • FMB vs. Wi-Fi and GSM
  • Signal stability and people’s presence
  • Power consumption
slide-54
SLIDE 54

54 21 April 2011

Power consumption results

  • Battery life: the time

for a completely charged device to switch off.

  • Unused modules

were turned off.

10 20 30 40 50 10 20

Battery life, hours Interval between scans, seconds

FM (3 channels) FM (45 channels) Wi-Fi Baseline

26 7 13 32

Measured with Samsung Omnia 2 smartphone.

 FM provides 2.6 to 5.5 times longer battery life than Wi-Fi.

42 28

slide-55
SLIDE 55

55 21 April 2011

Conclusion

slide-56
SLIDE 56

56 21 April 2011

Summary

  • FM radio:

– provides a good localization accuracy; – can be used in sensitive environments; – provides longer battery life than Wi-Fi; – is more robust to people’s presence; – is readily available.

slide-57
SLIDE 57

57 21 April 2011

Contributions (1/2)

  • Demonstration of feasibility of indoor

localization using FM radio signals from:

– Short-range FM transmitters

  • Accuracy comparable to Wi-Fi.

– Broadcasting FM stations

  • Accuracy superior than GSM;
  • Accuracy superior than Wi-Fi

(for confidence levels up to 90%).

slide-58
SLIDE 58

58 21 April 2011

Contributions (2/2)

  • Quantitative evaluation of influence of human

presence on FM and Wi-Fi RSS characteristics.

  • A method to counter accuracy degradation of

fingerprinting-based systems.

slide-59
SLIDE 59

59 21 April 2011

Publications

  • A.Papliatseyeu, V.Osmani and O.Mayora. Indoor Positioning Using FM Radio.

International Journal of Handheld Computing Research, 3(2010). PP. 19–31.

  • A.Matic, A.Popleteev, V.Osmani, and O.Mayora-Ibarra. FM Radio for Indoor

Localisation with Spontaneous Recalibration. Journal of Pervasive and Mobile Computing, 6(2010). PP. 642–656.

  • A.Papliatseyeu, A.Matic, V.Osmani, and O.Mayora-Ibarra. Indoor Positioning Using
  • ff-the-shelf FM Radio Devices. Abs. volume IPIN–2010. PP. 41–42.
  • A.Matic, A.Papliatseyeu, V.Osmani, and O.Mayora-Ibarra. Tuning to Your Position:

FM-radio based Indoor Localization with Spontaneous Recalibration. Proc. PerCom–2010. PP. 153–161.

  • A.Papliatseyeu, N.Kotilainen, O.Mayora-Ibarra, and V.Osmani. FINDR: Low-cost

indoor positioning using FM radio. Proc. MobilWare–2009. PP. 15–26.

slide-60
SLIDE 60

Thank you

slide-61
SLIDE 61

61 21 April 2011

slide-62
SLIDE 62

62 21 April 2011

FM stereo signal encoding

L+R L-R

RDS

Pilot

Noise

15 19 23 38 53 57

Frequency, kHz

slide-63
SLIDE 63

63 21 April 2011

Signal strength representation

FM Wi-Fi Unified (dB) 40..50 “Excellent”

  • 50

30..39 “Very good”

  • 60

20..29 “Good”

  • 70

10..19 “Low”

  • 80

1..9 “Very low”

  • 90

“No signal”

slide-64
SLIDE 64

64 21 April 2011

FM with Wi-Fi

ss1 ss2 ss3 ss1 ss2 ss3 ss1 ss2 ss3 ss4 ss5 ss6

FM fingerprint Wi-Fi fingerprint Combined wide fingerprint

slide-65
SLIDE 65

65 21 April 2011

FML combined with Wi-Fi

1 2 3 4 5 6 7 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Error, meters Confidence

Gaussian Processes

1 2 3 4 5 6 7 8 9 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Error, meters Confidence

kNN

FM Wi-Fi Combined

Receiver: HTC Artemis; grid: 1 m.

slide-66
SLIDE 66

66 21 April 2011

FM with Wi-Fi

  • Combined coverage
  • Longer battery life
  • Improved accuracy (by up to 22%)
slide-67
SLIDE 67

67 21 April 2011

Spontaneous recalibration

RSSI coordinate

Before:

RSSI coordinate

Now:

Reference positions

slide-68
SLIDE 68

68 21 April 2011

Spontaneous recalibration

RSSI coordinate

change change

RSSI coordinate Reference point neighbours are also updated Weight Distance from the ref. point 1

Propagation model

slide-69
SLIDE 69

69 21 April 2011

FMB accuracy vs. number of stations

slide-70
SLIDE 70

70 21 April 2011

1 2 3 4 5 6

Strongest Weakest Highest diversity

1 2 3 4 5 6 7

Error distance, m

2 4 6 8 10 12 1 7 13 19 25 31 37 43 49 55 61 67 73

Fingerprint width (number of stations)

50% 67% 95%

slide-71
SLIDE 71

71 21 April 2011

Station selection methods

  • Highest diversity approach shows best results.
  • Stronger stations perform similarly to weaker

stations.

  • In previous works, stronger FM stations

provided better median accuracy.

  • The contradiction is due to the difference

between indoor and outdoor signal changes.

slide-72
SLIDE 72

72 21 April 2011

Outdoors vs. indoors

Signal strength Distance

Outdoor RSSI difference Indoor RSSI difference Outdoor RSSI difference Indoor RSSI difference

Strong signal Weak signal

Indoor scale Indoor scale Outdoor scale Outdoor scale

slide-73
SLIDE 73

73 21 April 2011

Indoor obstacles

slide-74
SLIDE 74

74 21 April 2011

Localization performance

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1 2 3 4 5 6 7 8 9

Confidence Error distance, meters

FMb (all beacons) FML (3 beacons) Wi-Fi (all beacons) GSM (all beacons) Baseline

slide-75
SLIDE 75

75 21 April 2011

Application scenario

1 2 3 4 5 6 coffee uLab uLab uLab unknown unknown uLab uLab uLab uLab uLab uLab uLab coffee uLab unknown unknown uLab uLab uLab unknown unknown coffee uLab uLab coffee unknown coffee uLab uLab unknown unknown uLab uOffice uLab uLab unknown unknown uLab uLab uLab unknown uLab unknown uLab uLab unknown unknown uOffice unknown unknown unknown uOffice uLab coffee unknown unknown unknown coffee coffee coffee unknown unknown uLab uLab uLab unknown unknown coffee unknown unknown unknown uOffice uLab unknown 0.1 0.2 0.3 0.4 0.5 0.6 0.7 09:30 10:30 11:30 12:30 13:30 14:30 15:30 16:30 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 11:30 12:30 13:30 14:30 15:30 16:30 17:30 10:30 11:30 12:30 13:30 14:30 15:30 16:30

Questionnaire response Location Activity level Time

Activity level (body) Activity level (device) Subjective activity Subjective tensity

slide-76
SLIDE 76

76 21 April 2011

Application scenario

coffee uLab uLab uLab unknown unknown uLab uLab uLab uLab uLab uLab uLab coffee uLab unknown unknown uLab uLab uLab unknown unknown coffee uLab uLab coffee unknown coffee uLab uLab unknown unknown uLab uOffice uLab uLab unknown unknown uLab uLab uLab unknown uLab unknown uLab uLab unknown unknown uOffice unknown unknown unknown uOffice uLab coffee unknown unknown unknown coffee coffee coffee unknown unknown uLab uLab uLab unknown unknown coffee unknown unknown unknown uOffice uLab unknown 5 10 15 20 25 30 35 40 09:30 10:30 11:30 12:30 13:30 14:30 15:30 16:30 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 11:30 12:30 13:30 14:30 15:30 16:30 17:30 10:30 11:30 12:30 13:30 14:30 15:30 16:30

Location Distance in signal space Time