صفحه 1:
CSCE 5013: Hot Topics in Mobile and Pervasive
Computing
Discussion of LOC1 and LOC2
Nilanjan Banerjee
University of Arkansas
Fayetteville, AR
nilanjan.banerjee@gmail.co
m
Acknowledgment: Romit Roychoudhuri for the slides
Hot Topic in Mobile and Pervasive Computing
صفحه 2:
۱062: ۵
صفحه 3:
Location-Based Applications (LBAs)
™ For Example:
" GeoLife shows grocery list when near Walmart
" MicroBlog queries users at a museum
" Location-based ad: Phone gets coupon at Starbucks
™ jPhone AppStore: 3000 LBAs, Android: 500 LBAs
صفحه 4:
Most emerging location based apps
do not care about the physical location
ى
صفحه 5:
Most emerging location based apps
do not care about the physical location
ى
Instead, they need the user’s logical
location 223
3
صفحه 6:
Physical Vs Logical
™ Unfortunately, most existing solutions are physical
= GPS
" GSM based
= Google Latitude
= RADAR
" Cricket
صفحه 7:
Given this rich literature,
Why not convert from
Physical to Logical Locations?
صفحه 8:
صفحه 9:
Starbucks Pizza Hut
Physical Location
Error
صفحه 10:
Starbucks Pizza Hut
Physical Location
Error
10
صفحه 11:
11
SurroundSense:
A Logical Localization Solution
صفحه 12:
12
Hypothesis
such as sound, light, color, movement, WiFi
صفحه 13:
13
Hypothesis
It is possible to localize phones by
sensing the ambience
such as sound, light, color, movement, WiFi
صفحه 14:
Multi-dimensional sensing extracts more
ambient information
Any one dimension may not be unique,
but put together, they may provide a
unique fingerprint
و
14
صفحه 15:
SurroundSense
™ Multi-dimensional fingerprint
* Based on ambient
sound/light/color/movement/WiFi
15
صفحه 16:
Should Ambiences be Unique Worldwide?
صفحه 17:
صفحه 18:
Why does it work?
The Intuition:
Economics forces nearby businesses to be fee
Not profitable to have 3 adjascent coffee shop:
with same lighting, music, color, layout, etc.
SurroundSense exploits this ambience diversit
18
صفحه 19:
SurroundSense Architecture
Fingerprin
Logical
GSM Macro | Location
Location yy
Candidate Fingerprints
19
صفحه 20:
Acoustic fingerprint
Fingerprints (amplitude distribution)
z
™ Sound: 3
(via phone 3
microphone) ع
5
2
“0.8 وم 04 02 0 02 04 06
1 Amplitude Values
@ Color: Color and light fingerprints on HSL space
1۷8 ۳ 00-5 =.
camera) 2
eet 5
5
3
20
صفحه 21:
Fingerprinting Sound
® Fingerprint generation : Signal amplitude
= Amplitude values divided in 100 equal intervals
= Sound Fingerprint = 100 normalized values
* value, = # of samples in interval x / total # of samples
™ Filter Metric: Euclidean distance
" Discard candidate fingerprint if metric > threshold ع
™ Threshold r
= Multiple 1 minute recordings at the same location
= d, = max dist ( any two recordings )
= r = max (d, of candidate locations )
21
صفحه 22:
Fingerprinting Color ‘Fe ۳2
® Floor Pictures 1
" Rich diversity across different locations
" Uniformity at the same location x ee
® Fingerprint generation: pictures in HSL space
=" K-means clustering algorithm
* Cluster’s centers + sizes
® Ranking metric
7 1 SizeOf(Cus) SizeOf (Cry)
= aa 7 2 ةك
22
صفحه 23:
Fingerprints
™ Movement: (via phone accelerometer)
Cafeteria Clothes Store Grocery
Store
۳ || | | | ۱ ll
Static ۱
۱ ۲
500 1000 1500 2000 ۰ 250 ۰ 3000 3450 3500 3550 9000 3650 3700 3760 3800 2000 2100 2200 2300 2400 2500 0
Time(s) Tine () Time (6)
23
صفحه 24:
Fingerprints
™ Movement: (via phone accelerometer)
Cafeteria Clothes Store Grocery
Store
Moving} | | , | ۱ ۱
Static “+s | ۱ | ۱ | |
۱ ۲
2:05 :20 200 جح ند دم ترجه تن هد ند نم 0 ar
Time(s) Tine () Time (6)
Queuing Seated
24
صفحه 25:
Fingerprints
@ Movement: (via phone accelerometer)
Cafeteria Clothes Store Grocery
۱۱۱
500 1000 1500 2000 2500 3000 3450/4500 3550 3600 700تؤادوة 3750 3800 2000 2100 2200 2000 200 2500 0
Time(s) Tine () Time (6)
Moving}
Static
Pause for product Short walks
browsing between product
browsing
25
صفحه 26:
Fingerprints
@ Movement: (via phone accelerometer)
Clothes Store Grocery
WL LTT 01 11
‘28008000 9460 3500 9550 9600 9850 9700 5760 3800 2flo F100 2200 2000 Ago 2500 600
Time s) Tine 6)
Walk more Quicker
stops
26
Moving
Static
3
صفحه 27:
Fingerprints
@ Movement: (via phone accelerometer)
Cafeteria Clothes Store Grocery
Store
il
| ۳۱۱۱ ۲۱
500 1000 1500 2000 ۰ 250 ۰ 3000 3450 3500 3550 9000 3650 3700 3760 3800 2000 2100 2200 2300 2400 2500 0
Time(s) Tine () Time (6)
Moving}
Static
@ WiFi: (via phone wireless card)
f(overheard WiFi APs)
27
صفحه 28:
Fingerprinting WiFi
® Fingerprint generation: fraction of time each
unique address was overheard
® Filter/Ranking Metric
" Discard candidate fingerprints which do not have similar
MAC frequencies
min( film), fo(m))
(film) + $2) ax fa(m), fa(m)) 8 درق
meM
28
صفحه 29:
Discussion
™ Time varying ambience
= Collect ambience fingerprints over different time
windows
™ What if phones are in pockets?
= Use sound/WiFi/movement
" Opportunistically take pictures
®@ Fingerprint Database
=" War-sensing
29