sensing meets mobile social networks: the design, implementation and evaluation of the

24
Sensing Meets Mobile Social Sensing Meets Mobile Social Networks: Networks: The Design, Implementation and The Design, Implementation and Evaluation of the Evaluation of the CenceMe Application CenceMe Application Emiliano Miluzzo†, Nicholas D. Lane†, Kristóf Emiliano Miluzzo†, Nicholas D. Lane†, Kristóf Fodor†, Ronald Peterson†, Hong Lu†, Mirco Fodor†, Ronald Peterson†, Hong Lu†, Mirco Musolesi†, Shane B. Eisenman§, Xiao Zheng†, Musolesi†, Shane B. Eisenman§, Xiao Zheng†, Andrew T. Campbell† Andrew T. Campbell† Computer Science, Dartmouth College Computer Science, Dartmouth College §Electrical Engineering, Columbia University §Electrical Engineering, Columbia University Presented by Amre Shakimov CompSci 215

Upload: martena-boyer

Post on 31-Dec-2015

19 views

Category:

Documents


1 download

DESCRIPTION

Sensing Meets Mobile Social Networks: The Design, Implementation and Evaluation of the CenceMe Application. Emiliano Miluzzo†, Nicholas D. Lane†, Kristóf Fodor†, Ronald Peterson†, Hong Lu†, Mirco Musolesi†, Shane B. Eisenman§, Xiao Zheng†, Andrew T. Campbell† - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Sensing Meets Mobile Social Networks: The Design, Implementation and Evaluation of the

Sensing Meets Mobile Social Networks:Sensing Meets Mobile Social Networks:The Design, Implementation and The Design, Implementation and

Evaluation of the Evaluation of the CenceMe ApplicationCenceMe Application

Emiliano Miluzzo†, Nicholas D. Lane†, Kristóf Fodor†, Ronald Emiliano Miluzzo†, Nicholas D. Lane†, Kristóf Fodor†, Ronald Peterson†, Hong Lu†, Mirco Musolesi†, Shane B. Eisenman§, Peterson†, Hong Lu†, Mirco Musolesi†, Shane B. Eisenman§,

Xiao Zheng†, Andrew T. Campbell†Xiao Zheng†, Andrew T. Campbell†

††Computer Science, Dartmouth CollegeComputer Science, Dartmouth College

§Electrical Engineering, Columbia University§Electrical Engineering, Columbia University

Presented by Amre ShakimovCompSci 215

Page 2: Sensing Meets Mobile Social Networks: The Design, Implementation and Evaluation of the

MotivationMotivation

• Text messaging: Text messaging: •““Where R U?” Where R U?” •““What R U doing?”What R U doing?”

• Mobile phones are virtually always ON and with usMobile phones are virtually always ON and with us• Sensors in mobile phone: GPS, accelerometers, Sensors in mobile phone: GPS, accelerometers,

microphone, camera … etcmicrophone, camera … etc• Data collection through sensorsData collection through sensors

Page 3: Sensing Meets Mobile Social Networks: The Design, Implementation and Evaluation of the

Introduction of CenceMeIntroduction of CenceMe

• People-centric sensing applicationPeople-centric sensing application• Implementation on Nokia N95; Implementation on Nokia N95;

Symbian/JME VM platformSymbian/JME VM platform• Share user presence information Share user presence information

(Facebook)(Facebook)

Page 4: Sensing Meets Mobile Social Networks: The Design, Implementation and Evaluation of the

ContributionsContributions

• Design, implementation and evaluationDesign, implementation and evaluation• Lightweight classifierLightweight classifier• Trade-off: time fidelity v.s. latencyTrade-off: time fidelity v.s. latency• Complete User studyComplete User study

Page 5: Sensing Meets Mobile Social Networks: The Design, Implementation and Evaluation of the

Mobile Phone limitationsMobile Phone limitations

• OS LimitationsOS Limitations• API and Operational LimitationsAPI and Operational Limitations• Security LimitationsSecurity Limitations• Energy Management LimitationsEnergy Management Limitations

Page 6: Sensing Meets Mobile Social Networks: The Design, Implementation and Evaluation of the

Architecture Design IssuesArchitecture Design Issues

• Split-Level Classification (primitives, facts)Split-Level Classification (primitives, facts)– Customized tag Customized tag – ResiliencyResiliency– Minimize bandwidth usage/energyMinimize bandwidth usage/energy– Privacy/data integrityPrivacy/data integrity

• Power Aware Duty-Cycle (~6 hours)Power Aware Duty-Cycle (~6 hours)

Page 7: Sensing Meets Mobile Social Networks: The Design, Implementation and Evaluation of the

CenceMe ImplementationCenceMe Implementation

Operations (Phone):Operations (Phone):• Sensing Sensing • Classification to produce primitives Classification to produce primitives • Presentation of people's presence on the phone Presentation of people's presence on the phone • Upload of primitives to backend servers Upload of primitives to backend servers

Classifications (Backend Server):Classifications (Backend Server):• classifying the nature of the sound collected from the microphone classifying the nature of the sound collected from the microphone • classifying the accelerometer data to determine activity (sitting, standing, classifying the accelerometer data to determine activity (sitting, standing,

walking, running) walking, running) • scanned Bluetooth/MAC addresses in range scanned Bluetooth/MAC addresses in range • GPS readings GPS readings • random photos random photos (!!!)(!!!)

Page 8: Sensing Meets Mobile Social Networks: The Design, Implementation and Evaluation of the

Phone SoftwarePhone Software

Page 9: Sensing Meets Mobile Social Networks: The Design, Implementation and Evaluation of the

ClickStatusClickStatus

Page 10: Sensing Meets Mobile Social Networks: The Design, Implementation and Evaluation of the

Backend SoftwareBackend Software

Page 11: Sensing Meets Mobile Social Networks: The Design, Implementation and Evaluation of the

Phone classifiers (1/2)Phone classifiers (1/2)

• AudioAudio– Feature extractionFeature extraction– ClassificationClassification

Page 12: Sensing Meets Mobile Social Networks: The Design, Implementation and Evaluation of the
Page 13: Sensing Meets Mobile Social Networks: The Design, Implementation and Evaluation of the

Phone classifiers (2/2)Phone classifiers (2/2)

• ActivityActivity

Page 14: Sensing Meets Mobile Social Networks: The Design, Implementation and Evaluation of the

Backend ClassifierBackend Classifier

• ConversationConversation• Social ContextSocial Context

– Neighborhood conditionsNeighborhood conditions– Social StatusSocial Status

• Mobility Mode DetectorMobility Mode Detector• LocationLocation• ““Am I Hot”Am I Hot”

– Nerdy, party animal, cultured, healthy, greenyNerdy, party animal, cultured, healthy, greeny

Page 15: Sensing Meets Mobile Social Networks: The Design, Implementation and Evaluation of the

System PerformanceSystem Performance

• Classifier accuracyClassifier accuracy• Impact of mobile phone placement on bodyImpact of mobile phone placement on body

– 8 users8 users– Annotations as ground truth for comparison with Annotations as ground truth for comparison with

classifier outputsclassifier outputs

• Environmental conditionsEnvironmental conditions• Sensing duty cyclesSensing duty cycles

Page 16: Sensing Meets Mobile Social Networks: The Design, Implementation and Evaluation of the

General ResultGeneral Result

Page 17: Sensing Meets Mobile Social Networks: The Design, Implementation and Evaluation of the

Phone placement on bodyPhone placement on body• Pocket, lanyard, clipped to beltPocket, lanyard, clipped to belt• Insignificant impact conversation v.s. Non-conversationInsignificant impact conversation v.s. Non-conversation

Page 18: Sensing Meets Mobile Social Networks: The Design, Implementation and Evaluation of the

Environmental impactEnvironmental impact• Independent of activity classificationIndependent of activity classification• More important: transition between locationsMore important: transition between locations

Page 19: Sensing Meets Mobile Social Networks: The Design, Implementation and Evaluation of the

Duty Cycle (1/2)Duty Cycle (1/2)• Problem detecting short term eventProblem detecting short term event• Experiment: 8 people. Reprogram different duty cycles.Experiment: 8 people. Reprogram different duty cycles.

Page 20: Sensing Meets Mobile Social Networks: The Design, Implementation and Evaluation of the

Power BenchmarksPower Benchmarks• Measuring battery voltage, current, temperatureMeasuring battery voltage, current, temperature• Battery lifetime: 6.22+/- 0.59 hoursBattery lifetime: 6.22+/- 0.59 hours

Page 21: Sensing Meets Mobile Social Networks: The Design, Implementation and Evaluation of the

Memory and CPU BenchmarksMemory and CPU Benchmarks

Page 22: Sensing Meets Mobile Social Networks: The Design, Implementation and Evaluation of the

User StudyUser Study• Survey user experienceSurvey user experience• Feedback:Feedback:

– Positive from all usersPositive from all users– Willing to share detail status and presence information Willing to share detail status and presence information

on Facebookon Facebook– Privacy not an issuePrivacy not an issue– Stimulate curiosity among usersStimulate curiosity among users– Self-learning on activity patterns and social statusSelf-learning on activity patterns and social status

Page 23: Sensing Meets Mobile Social Networks: The Design, Implementation and Evaluation of the

Rooms for improvementRooms for improvement

• Battery life to up to 48 hoursBattery life to up to 48 hours• Finer grained privacy policy settings.Finer grained privacy policy settings.• Shorter classification timeShorter classification time

Page 24: Sensing Meets Mobile Social Networks: The Design, Implementation and Evaluation of the

ConclusionConclusion

• A complete design, implementation and evaluationA complete design, implementation and evaluation• First application to retrieve and publish sensing presenceFirst application to retrieve and publish sensing presence• A complete user study and feedback for future A complete user study and feedback for future

improvementimprovement