Upload
guest17df6
View
1.121
Download
4
Tags:
Embed Size (px)
Citation preview
1
Mobile Sensing: Leveraging Mobile Phones to Support Personal, Community, and Participatory Sensing
Nithya Ramanathan
Collaborating Faculty: Jeff Burke, Deborah Estrin, Mark Hansen, Mani Srivastava, Ruth WestUCLA Center for Embedded Networked SensingUCLA Center for Research in Engineering, Media and Performance
Staff and Graduate Students: Faisal Alquaddoomi, Betta Dawson, Jeff Goldman, Eric Howard, August Joki, Donnie Kim, Vinayak Naik, Min Mun, Nicolai Petersen, Sasank Reddy, Jason Ryder, Vids Samanta, Katie Shilton, Nathan YauUCLA Departments of Computer Science, Electrical Engineering, StatisticsCENS Urban Sensing collaborators also include: Mark Allman, Dana Cuff, Jerry Kang, Vern Paxson, Fabian Wagmister, CENS Urban Sensing funding sources include: NSF CRI, NeTS-FIND, and OIA; Cisco, Nokia, Schematic, Sun, Walt Disney Imagineering R&D
http://urban.cens.ucla.edu
Text Entry
Imagers Audio
Location (GPS)
Accelerometer
BluetoothNetwork Connectivity
What can one person do with this powerful tool?
3Reddy, Samanta, Burke, Estrin, Hansen, Srivastava
Presentation
PresentationVisualizationProcessingRaw Data
Can real-time feedback about our actions change our behavior?
Mobile SensingGrassroots data collection
Scalable
Affordable
Believable
4Reddy, Samanta, Burke, Estrin, Hansen, Srivastava
What can thousands of coordinated people do with this powerful tool?
Mobile SensingGrassroots data collection
Scalable
Affordable
Believable
5Reddy, Samanta, Burke, Estrin, Hansen, Srivastava
What can thousands of coordinated people do with this powerful tool?
Mobile SensingGrassroots data collection
Scalable Affordable Believable
6Reddy, Samanta, Burke, Estrin, Hansen, Srivastava
What can thousands of coordinated people do with this powerful tool?
Data Campaigns
7Reddy, Samanta, Burke, Estrin, Hansen, Srivastava
Phones enhance participationMake it.. Easier to collect more dataData is more credible and verifiable
Data Campaigns
8Reddy, Samanta, Burke, Estrin, Hansen, Srivastava
Phones enhance participationMake it.. Easier to collect more dataData is more credible and verifiable
No Technological Innovation
9
Image and activity data to study pollution exposure
http://www-ramanathan.ucsd.edu/ProjectSurya.html
Characterize Outdoor Activities
Infer Duration of Exposure
Collect Indoor Pollution Levels
Bluetooth temperature
sensor
Phone + GPS, accelerometer
Special soot filter
In collaboration with UCSD, SRU and TERI in India
Active Image Collection for citizen science
10
http://www.windows.ucar.edu/citizen_science/budburst
In collaboration with the ongoing citizen science initiative known as BudBurst
Reflecting on and learning from personal mobility
Cyclesense combines location data and users’ photos to give bikers daily feedback and suggestions on the quality and safety of their commutes.
In PEIR, the combination of location, time, and activity are automatically interpreted using regional air quality models to estimate participants’ exposure to particulate matter.
http://peir.cens.ucla.edu/ http://urban.cens.ucla.edu/projects/cyclesense/
12
Audio and location collection to recall family interactions
http://urban.cens.ucla.edu/projects/familydynamics/
http://www.kt.tu-cottbus.de/speech-analysis/
http://urban.cens.ucla.edu/projects/familydynamics/
In collaboration with the Semel Institute
13
While off-the-shelf components can be used to implement data campaigns, there is still room for improvement….
14
Protecting privacy while still collecting meaningful data
Example: Images of food contain compromising objects in the background
Ways to approach the problem
• Just don’t upload the sensitive data in the first place. Analysis on phone and partial sampling (FamilyDynamics)
• Give the user control over which data to release. Selective sharing (DietSense)
• Release the data, but add noise or obfuscate certain portions as needed. Location cloacking (PEIR)
• Only release aggregates of the data across some window in time or space Data aggregates (PEIR)
Design to Maximize Trust and Participation
15
Design Theme: Involve rather than burden the user by designing systems that are easy to use and understand.
Design Theme: Validate data.
16
http://urban.cens.ucla.edu/
17
PEIR Press Release
18
Passive Image Collection for Diet Recall Studies
http://urban.cens.ucla.edu/projects/dietsense/
In collaboration with the public health department at UCLA
19
Data Analysis and Using external data streams Example: Estimating Pollution without Pollution Sensors
Lifelong damage found in 13-year study of 3,600 Southland youngsters living within 500 yards of a highway. The Los Angeles Times, 1/26/07
Houston, Winer et al
Source: McConnell et al. Traffic, Susceptibility, and Childhood Asthma. Environ Health Perspect 114:766–772 (2006)
19
20
Urban Sensing: Research Challenges
Scaling and credibility.Coordinated, opportunistic sampling.
Network attestation and verification of location, time, and other context.
Encouraging sharing.Data protection and selective, resolution-controlled dissemination.
Participatory privacy
Anonymous and pseudonymous participation. Reputation, incentive, and authoring frameworks.
Finding, visualizing, and analyzing data. Data stream naming, privacy-respecting discovery, and signal search.
Server-side signal processing for data processing, browsing, and auditing.
Spatial interfaces to data and authoring.
Infrastructure for capture, review, processing. Adaptive collection protocols.
Automatic feeding of data to models.
Privacy isn’t a separate concern...
it’s embedded in the sensing and research activities...
it has variable meaning in specific circumstances and settings...
it will skew participation and data collection