LYU0401 Location-Based LYU0401 Location-Based Multimedia Mobile Multimedia Mobile

Service Service Clarence FungClarence Fung

Tilen MaTilen MaSupervisor: Professor Michael LyuSupervisor: Professor Michael Lyu

Marker: Professor Alan LiewMarker: Professor Alan Liew

OutlineOutline

IntroductionIntroduction ObjectiveObjective Location-Based ServiceLocation-Based Service Current Localization MethodsCurrent Localization Methods Experimental StudyExperimental Study Wi-Fi Location SystemWi-Fi Location System Future WorkFuture Work Conclusion Conclusion

IntroductionIntroduction

In this semester, we mainly focus on tIn this semester, we mainly focus on the problem of localizationhe problem of localization

We have chosen the 1st floor of the Ho We have chosen the 1st floor of the Ho Sin-Hang Engineering Building to studSin-Hang Engineering Building to study the problem of localizationy the problem of localization

Our goal is to locate a person when he/Our goal is to locate a person when he/she is walking around on the floor she is walking around on the floor

ObjectiveObjective

To meet the need for Location-Based To meet the need for Location-Based ServiceService

To find out if Wireless LAN provide To find out if Wireless LAN provide enough information for localization in enough information for localization in 2D space2D space

Study on different localization Study on different localization algorithmsalgorithms

Develop an application in a mobile Develop an application in a mobile devicedevice

Location-Based ServiceLocation-Based Service Localization is necessary for many higher level sensLocalization is necessary for many higher level sens

or network functions such as tracking, monitoring aor network functions such as tracking, monitoring and geometric-based routing nd geometric-based routing

Three categories: Three categories:

Global location systemsGlobal location systems Wide-area location systems Wide-area location systems Indoor location systemsIndoor location systems

Systems in indoor environmentSystems in indoor environment Infrared (IR)Infrared (IR) UltrasoundUltrasound Radio signal Radio signal

Wireless LAN (WLAN)-Wireless LAN (WLAN)-Based Positioning SystemBased Positioning System

Advantages over all other systemsAdvantages over all other systems EconomicalEconomical

WLAN network usually exists already as part of tWLAN network usually exists already as part of the communications infrastructurehe communications infrastructure

Covers a large area Covers a large area Work in a large building or even across many buiWork in a large building or even across many bui

ldings. ldings. Stable systemStable system

Video- or IR-based location systems are subject tVideo- or IR-based location systems are subject to restrictions, such as line-of-sight limitationso restrictions, such as line-of-sight limitations

Current Localization Current Localization MethodsMethods

Point-based approachPoint-based approach goal is to return a single point for the mobile goal is to return a single point for the mobile

objectobject E.g. Simple Distance MatchingE.g. Simple Distance Matching

Area-based approachArea-based approach goal is to return the possible locations of the goal is to return the possible locations of the

mobile object as an area rather than a single mobile object as an area rather than a single pointpoint

E.g. Simple-Point Matching, Area-Based E.g. Simple-Point Matching, Area-Based ProbabilityProbability

Area-Based Probability Area-Based Probability (ABP)(ABP)

Advantages:Advantages: Presents the user an understanding Presents the user an understanding

of the system in a more natural and of the system in a more natural and intuitive mannerintuitive manner

High accuracyHigh accuracy More mathematical approachMore mathematical approach

Steps in using ABPSteps in using ABP

Decide the Areas

Measure Signals at Different Areas

Create a Training Set

Measure Signals at Current Position

Create a Testing Set

Find out the Probability of Being at Different Areas

Calculate Probability Density

Return the Area with Highest Probability

Applying Area-based Applying Area-based ApproachApproach

Some Terms and Some Terms and DefinitionsDefinitions

n Access Pointsn Access Points APAP11, AP, AP22, …, AP, …, APnn

Training set TTraining set T00 Offline measured signal strengths at differeOffline measured signal strengths at differe

nt locations an algorithm usesnt locations an algorithm uses Consists of a set of fingerprints (SConsists of a set of fingerprints (Sii) at m diff) at m diff

erent areas Aerent areas Aii

TT0 0 = ( A= ( Aii, S, Sii ), i = 1 … m ), i = 1 … m

Some Terms and Some Terms and DefinitionsDefinitions

Fingerprints SFingerprints Sii

Set of n signal strengths at ASet of n signal strengths at Aii, one per eac, one per each access pointh access point

SSi i = (s= (si1i1, …, s, …, sinin), where s), where sijij is the expected av is the expected average signal strength from APerage signal strength from APjj

Generating Training SetGenerating Training Set

In one particular AIn one particular Aii, we read a series of , we read a series of signal strengths (ssignal strengths (sijkijk ) for a particular A ) for a particular APPj j with a constant time between samplwith a constant time between sampleses k = 1… ok = 1… oij ij ,where o,where oijij is the number of samp is the number of samp

les from APles from APj j at Aat Aii We estimate sWe estimate sij ij by averaging the series, by averaging the series,

{s{sij1ij1, s, sij2ij2…, s…, sijoijo } }

Generating Training SetGenerating Training Set

We do the same for all n APs, so we haWe do the same for all n APs, so we have the fingerprints at Ave the fingerprints at Aii, , SSi i = (s= (si1i1, …, s, …, sinin))

We do the same for all m areas, so we We do the same for all m areas, so we have the training set have the training set TT0 0 = ( A= ( Aii, S, Sii ), i = 1… m ), i = 1… m

Collecting SignalsCollecting Signals At each area chosen, we measure the signal At each area chosen, we measure the signal

strength from the access points for 1 minute strength from the access points for 1 minute Position 1 2 3 4 5 6 7 8 9 10 11 12

AP MAC address Signal Strength (dBm)

00:02:2d:28:be:9e -70 -62 -58 -67 -73 -78 -83 -86 -84 -81 -78 -55

00:02:2d:28:be:5d -67 -59 -60 -71 -76 -79 -81 -86 -81 -83 -79 -52

00:60:1d:1e:43:9b -79 -87 -85 -84 -89 -80 -76 -77 -66 -63 -77 -90

00:0f:34:f3:60:40 -63 -69 -65 -74 -76 -72 -77 -84 -76 -74 -66 -79

00:02:2d:21:39:1f -82 -78 -82 -59 -78 -73 -83 -85 -82

00:11:93:3d:6f:c0 　 -90 -85 -86 -89 -88

00:11:20:93:65:c0 　 -89 -89 -90

00:0f:34:bb:df:20 　 -89 -90 -82 -88 -88

00:0c:ce:21:1b:9d 　 -87

00:0c:85:35:33:d2 　 -88 -88

00:11:20:93:63:90 　 -89 -88

00:0c:85:35:33:d4 　 -87

00:04:76:a7:ab:a3 -90

Data ProcessingData Processing

We have chosen 7 out of 13 access We have chosen 7 out of 13 access pointspoints least contribution to localization least contribution to localization shorten computation time shorten computation time

For missing signal strengths, we For missing signal strengths, we input -92 dBm as entryinput -92 dBm as entry

Training SetTraining SetPosition 1 2 3 4 5 6 7 8 9 10 11 12

AP MAC address Signal Strength (dBm)

00:02:2d:28:be:9e -70 -62 -58 -67 -73 -78 -83 -86 -84 -81 -78 -55

00:02:2d:28:be:5d -67 -59 -60 -71 -76 -79 -81 -86 -81 -83 -79 -52

00:60:1d:1e:43:9b -79 -87 -85 -84 -89 -80 -76 -77 -66 -63 -77 -90

00:0f:34:f3:60:40 -63 -69 -65 -74 -76 -72 -77 -84 -76 -74 -66 -79

00:02:2d:21:39:1f -92 -92 -82 -78 -82 -59 -78 -73 -83 -85 -82 -92

00:11:93:3d:6f:c0 -92 -92 -92 -90 -85 -86 -89 -88 -92 -92 -92 -92

00:0f:34:bb:df:20 -92 -92 -92 -89 -90 -82 -88 -88 -92 -92 -92 -92

Getting Testing SetGetting Testing Set

The object to be localized collects a set The object to be localized collects a set of received signal strengths (RSS) wheof received signal strengths (RSS) when it is at certain locationn it is at certain location

A testing set (SA testing set (Stt) is created similar to th) is created similar to the fingerprints in the training sete fingerprints in the training set

It is a set of average signal strengths frIt is a set of average signal strengths from APs, Som APs, St t = (s= (st1t1, …, s, …, stntn))

RSSRSS

AP MAC address Signal Strength (dBm)

00:02:2d:28:be:9e -71

00:02:2d:28:be:5d -72

00:60:1d:1e:43:9b -89

00:0f:34:f3:60:40 -49

Testing SetTesting Set

AP MAC address Signal Strength (dBm)

00:02:2d:28:be:9e -71

00:02:2d:28:be:5d -72

00:60:1d:1e:43:9b -89

00:0f:34:f3:60:40 -49

00:02:2d:21:39:1f -92

00:11:93:3d:6f:c0 -92

00:0f:34:bb:df:20 -92

Applying ABPApplying ABP

Goal: return the area with a highest prGoal: return the area with a highest probability obability

Approach: compute the likelihood of tApproach: compute the likelihood of the testing set (She testing set (Stt) that matches the fing) that matches the fingerprint for each area (Serprint for each area (Sii) )

Applying ABPApplying ABP

Assumptions:Assumptions: Signal received from different APs are Signal received from different APs are

independent independent

For each APFor each APjj, j = 1…n, the sequence of , j = 1…n, the sequence of RSS sRSS sijkijk, k = 1… o, k = 1… oijij, at each A, at each Ai i in Tin Too is mo is modeled as a Gaussian distributiondeled as a Gaussian distribution

Applying Bayes’ ruleApplying Bayes’ rule We compute the probability of being aWe compute the probability of being a

t different areas At different areas Aii, on given the testin, on given the testing set Sg set St t

P(AP(Ai i |S|Stt) = P(S) = P(St t |A|Aii)* P(A)* P(Aii)/ P(S)/ P(Stt) (1)) (1) P(SP(Stt) is a constant) is a constant Assume the object is equally likely to be at any Assume the object is equally likely to be at any

location. P(Alocation. P(Aii) is a constant) is a constant

P(AP(Ai i |S|Stt) = c*P(S) = c*P(St t |A|Aii)) (2) (2)

Area Based ProbabilityArea Based Probability We compute P(SWe compute P(St t |A|Aii) for every area A) for every area Aii ,i=1…m, using ,i=1…m, using

the Gaussian assumptionthe Gaussian assumption

Finding Probability DensityFinding Probability Density the object must be at one of the 12 areas the object must be at one of the 12 areas ΣP(Ai | St) =1 for all i ΣP(Ai | St) =1 for all i

Max{P(AMax{P(Ai i |S|Stt) } = Max{c*P(S) } = Max{c*P(St t |A|Aii) } ) } = Max{P(S= Max{P(St t |A|Aii) }) }

Return the area AReturn the area Aii with top probability with top probability

Gaussian Distribution Gaussian Distribution In our application, In our application,

we can take μ as the we can take μ as the expected average siexpected average signal strengths for thgnal strengths for the access point to be e access point to be calculated calculated

we take σ as 8.5we take σ as 8.5

Integral of Normal Integral of Normal FunctionFunction

Find probability Find probability by integrationby integration

Take interval as Take interval as 1 1

Error function erf(x)Error function erf(x) Express Integral of Express Integral of

Normal Function in Normal Function in terms of erfterms of erf

Approximate value Approximate value of erf by a seriesof erf by a series

Choose iteration of Choose iteration of 5050

Experimental StudyExperimental Study Area 5 is near Area 5 is near

the North-West the North-West stairway on the stairway on the 1st floor1st floor

deep purple line deep purple line is on the top of is on the top of other linesother lines

Localization Localization system returns system returns the correct the correct resultresult

ABP Localization Graph at Area 5

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

1 5 9 13 17 21 25 29 33 37 41Sample No

Prob

ability

Probability at A1

Probability at A2

Probability at A3

Probability at A4

Probability at A5

Probability at A6

Probability at A7

Probability at A8

Probability at A9

Probability at A10

Probability at A11

Probability at A12

Accuracy of Localization Accuracy of Localization SystemSystem

Default Default sample size sample size of testing of testing set = 4set = 4

80 testing 80 testing sets for each sets for each of the 12 of the 12 locations locations

8580

75 74

8575 71

87.5

52.5

70

87.5

6975.94

0

10

20

30

40

50

60

70

80

90

Percentage

1 2 3 4 5 6 7 8 9 10 11 12 Overall

Area

Accuracy at Different Locations

Accuracy of Localization Accuracy of Localization SystemSystem

1 5 9 13 17 2125

76 87.6 93 96 98 99 99.5

0102030405060708090100

Percentage

Sample No

System Overall Accuracy

Other Factors affecting Other Factors affecting Accuracy Accuracy

Property of signalsProperty of signals The strength of signals fluctuatesThe strength of signals fluctuates

Hardware failureHardware failure access points fails to give out signals or access points fails to give out signals or

give out signals at unusual strengthgive out signals at unusual strength Change in environment Change in environment

addition access points on the floor addition access points on the floor opening the doorsopening the doors

Orientation in collecting signal Orientation in collecting signal

Wi-Fi Location System Wi-Fi Location System (WLS) (WLS)

Development Tool for Location-Based SyDevelopment Tool for Location-Based Systemstem

Simplify development stepsSimplify development steps Increase the efficiency and productivity Increase the efficiency and productivity It divides into 3 componentsIt divides into 3 components

Wi-Fi Signal Scanner (WSS)Wi-Fi Signal Scanner (WSS) Wi-Fi Data Processor (WDP)Wi-Fi Data Processor (WDP) Wi-Fi Location Detector (WLD)Wi-Fi Location Detector (WLD)

Wireless LAN Wireless LAN TerminologyTerminology

Media Access Control address (MAC Media Access Control address (MAC Address)Address) 48 bits long48 bits long unique hardware address unique hardware address e.g. 00:50:FC:2A:A9:C9e.g. 00:50:FC:2A:A9:C9

Service set identifier (SSID)Service set identifier (SSID) 32 character32 character Wireless LAN identifierWireless LAN identifier

Receive Signal Strength Indicator (RSSI) Receive Signal Strength Indicator (RSSI) signal strength signal strength unit is in dBm unit is in dBm

OverviewOverview

Platform:Platform: Window CEWindow CE Window XP, 2000Window XP, 2000

Technology:Technology: IEEE 802.11bIEEE 802.11b

ToolsTools Embedded Visual C++ 4.0Embedded Visual C++ 4.0 Visual Studio .NET 2003Visual Studio .NET 2003

Tradition Development Tradition Development Procedure (TDP)Procedure (TDP)

The followings in the Tradition The followings in the Tradition Development ProcedureDevelopment Procedure

Studying the technology

Software Design

Algorithm design

Final System

1-2 week

1-2 week

2-3 week

Wi-Fi Location System Wi-Fi Location System Development Procedure Development Procedure

(WLP) (WLP) Collecting Data

Processing Data

Deploying and Test System

Several hours

1 day

Final System

Several days

Using Wi-Fi Signal Scanner

Using Wi-Fi Data Processor

Using Wi-Fi Location Detector

Comparison between TDP Comparison between TDP and WLPand WLP

Using WLP, we can develop Using WLP, we can develop Location-Based System in a short Location-Based System in a short time.time.

This work can be done by non-This work can be done by non-professionalsprofessionals

It simplifies Development StepsIt simplifies Development Steps

Wi-Fi Signal Scanner Wi-Fi Signal Scanner

To collect the To collect the signal strength signal strength received from received from access pointsaccess points

Collected DataCollected Data

Strength Signal

Number of Received

Signal

Total of Received

Signal

Mean of Received

Signal

Wi-Fi Data Processor Wi-Fi Data Processor

To process collected dataTo process collected data

Position Region

Access Point

Region

Setting and Information

Region

Wi-Fi Data ProcessorWi-Fi Data Processor

Two main steps in WDPTwo main steps in WDP Filter out useless dataFilter out useless data Set parameters at each positionSet parameters at each position

DataData NameName Point at Map PicturePoint at Map Picture

Wi-Fi Location Detector Wi-Fi Location Detector

Three functions in WLDThree functions in WLD To detect the location in the target place To detect the location in the target place To show the detected position name and To show the detected position name and

corresponding position at the Map Picture corresponding position at the Map Picture To show calculated probabilityTo show calculated probability

Three modes in WLD Three modes in WLD Data ModeData Mode Map ModeMap Mode Probability Mode Probability Mode

Data ModeData Mode

To show the sample dataTo show the sample data

Map ModeMap Mode

Position

Name

Probability ModeProbability Mode

To show calculated probability at each To show calculated probability at each positionposition

ConclusionConclusion

We are success in applying Area-Based We are success in applying Area-Based ProbabilityProbability

We have done experiments on accuracy We have done experiments on accuracy of algorithmof algorithm

We have implemented Location-Based We have implemented Location-Based Development Tool—Wi-Fi Location Development Tool—Wi-Fi Location SystemSystem

Based on our knowledge and developed Based on our knowledge and developed tools in localization, we are able to tools in localization, we are able to further develop a location-based service further develop a location-based service

Future WorkFuture Work

Ho Sin-Hang Engineering Building Ho Sin-Hang Engineering Building Tour Guide ServiceTour Guide Service

Multimedia Application with video Multimedia Application with video streaming streaming

Improvement in Localization Improvement in Localization Algorithm Algorithm

Increase the Accuracy in Localization Increase the Accuracy in Localization Research on 3D localization Research on 3D localization

algorithm in an building algorithm in an building

Q&AQ&A

DEMODEMO

THE ENDTHE END