abs-introduction
TRANSCRIPT
Indoor Localization without
Infrastructure using the
Acoustic Background
Spectrum
Stephen P. Tarzia, Peter A.
Dinda, Robert P. Dick, Gokham
MemikMobiSys 2011
Jose Alvarez
Sezaki Lab.
Background & Objectives
Motivations
◦ Indoor Localization
◦ WiFi not always available and is not a
definite solution
◦ Improved accuracy
Background & Objectives
Related works
◦ Radio
Sensors as beacons
WiFi
◦ Other than radio
Accelerometer and Compass
SoundSense and SurroundSense
Background & Objectives
Distinctive elements of this method
◦ Transient noise v/s Background noise
=> Listen to background noises
◦ Look at frequency domain
Points of Interest
69% accuracy for 33 rooms
Implementation using smartphones
(Batphone in AppleStore)
Effectively combined sound technique
with WiFi
Proposed Method
Database side
◦ Training by gathering labeled samples (4
per room)
Client side
◦ Euclidian distance measure for selecting
candidates
Proposed Method
Results
Quality of the
recording device did
not impact accuracy
For 2 rooms
discrimination
accuracy goes up to
92%
Results
Chatter state
performance
drops
dramatically
Choosing
different
frequency band
helps in
conversation
state
Results
During the experiments the device
(Ipod touch) was with the screen on at
all times, with the brightness setting at
the dimmest => Battery life was 7.9
hours
11.5% CPU usage
16.7 MB of main memory
Battery life extended for a reasonable
time.
Why this method?
Indoor localization is not solved
completely yet
Using room’s acoustic fingerprint
seems like a creative method for room
identification
Hardware on smartphones is more
than enough for a feasible
implementation (Batphone)