DOC PREVIEW
UCF EEL 6788 - DARWIN PHONES - THE EVOLUTION OF SENSING AND INFERENCE ON MOBILE PHONES

This preview shows page 1-2-3-19-20-39-40-41 out of 41 pages.

Save
View full document
View full document
Premium Document
Do you want full access? Go Premium and unlock all 41 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 41 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 41 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 41 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 41 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 41 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 41 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 41 pages.
Access to all documents
Download any document
Ad free experience
Premium Document
Do you want full access? Go Premium and unlock all 41 pages.
Access to all documents
Download any document
Ad free experience

Unformatted text preview:

Darwin Phones: The Evolution of Sensing and Inference on Mobile Phones Presented By: Brandon OchsWhat does Darwin do?What about battery life?Common Urban Sensing ChallengesTypes of LearningDarwin StepsClassifier EvolutionModel PoolingCollaborative InferenceDarwin Design: ComputationDarwin Design: ContextDarwin Design: Co-locationSpeaker RecognitionSpeaker ModelingClassifier Evolution: Training StepClassifier Evolution: Evolution StepSlide 17Slide 18Slide 19Privacy and TrustExperimental ResultsExperiment 1 ParametersExperiment 1 Results: Without EvolutionExperiment 2 ParametersExperiment 2 ResultsSlide 26Experiment 3 ParametersExperiment 3 ResultsSlide 29Slide 30Slide 31Experiment 4 ParametersExperiment 4 ResultsSlide 34Time and Energy MeasurementsSlide 36Possible ApplicationsFuture WorkImprovements On The PaperConclusionReferencesDARWIN PHONES: THE EVOLUTION OF SENSING AND INFERENCE ON MOBILE PHONESPRESENTED BY: BRANDON OCHSEmiliano Miluzzo, Cory T. Cornelius, Ashwin Ramaswamy, Tanzeem Choudhury, Zhigang Liu, Andrew T. Campbell, "Darwin phones: the evolution of sensing and inference on mobile phones," In Proc. of 8th ACM Conference on Mobile Systems, Applications, and Services (MobiSys), 2010, pp. 5-20.What does Darwin do?A Smartphone platform for urban sensingProof of concept model uses microphoneCommunicates with other local devices to improve inference accuracy (collaborative inference)Framework can be expanded to gatherinformation using a range of sensor dataWhat about battery life?Communicates with backend server to do the CPU-intensive machine learning algorithmsLocal devices share models rather than re-computing themSensing is enabled/disabled as the system sees fitCommon Urban Sensing ChallengesHuman burden of training classifiersAbility to perform reliably in different environments (indoor vs outdoor)The ability to scale to a large number of phones without hurting usability and battery life.Darwin overcomes all of these through classifier/model evolution, model pooling, and collaborative inferenceTypes of LearningSupervised: Given a fully-labeled training setSemi-Supervised: Given a small training set that is evolvedUnsupervised: No training set is givenDarwin StepsEvolution, Pooling, and Collaborative InferenceThese represent Darwin’s novel evolve-pool-collaborate model implemented on mobile phonesClassifier EvolutionAutomated approach to updating models over timeNeeds to account for variability in sensing conditions and settingsVariability in background noise and phone location require separate modelsModel PoolingReuses models that have already been built and evolved on other phonesExchange classification models whenever the model is available from another phoneClassifiers do not need to be retrained, which increases scalabilityCan pool models from backend serversCollaborative InferenceCombines results from multiple phonesRun inference algorithms in parallel on the same classifiersSystem is more robust to degradation in sensing qualityIncreases accuracyDarwin Design: ComputationReduces the on-the-phone computation by offloading some of the work to backend serversBackend server uses a machine learning algorithm to compute a Gaussian Mixture Model (2 hours for 15 seconds of audio)Feature vectors are computedlocallyDarwin Design: ContextContext (in/out of pocket, in/out of bag) will impact the sensing and inference capabilityClassifier evolution makes sure the classifier of an event is robust across different environmentsDarwin Design: Co-locationAccounts for a group of co-located phones running the same classification algorithm and sensing the same event but computing different inference resultsPhones pool classification models when collocated or from backend serversCompares against its own model and the co-located modelDrastically reduces classification latencyExploits diversity of different phone sensing context viewpointsSpeaker RecognitionAttempts to identify a speaker by analyzing the microphone’s audio streamSuppresses silence, low amplitude audio, and chunks that do not contain human voiceReduce false positives by pre-processing in 32ms blocksSpeaker ModelingFeature vector consisting ofMel Frequency Cepstral CoefficientsEach speaker is modeled with 20 GaussiansAn initial speaker model is built by collecting a short training sampleClassifier Evolution: Training StepShort training phase (30 seconds) used to build a model which is later evolvedFirst 15 seconds used as the training setLast 15 seconds used as baseline for evolutionClassifier Evolution: Evolution StepSemi-supervised learning strategyIf the likelihood of the incoming audio stream is much lower than any of the baselines then a new model is evolvedCollaborative InferenceLocal inference phase can be broken into three steps:Local inference operated by each individual phonePropagation of the result of the local inference to the neighboring phonesFinal inference based on the neighboring mobile phones local inference resultsEach node individually operates inference on the sensed eventResults and confidence broadcastedPrivacy and TrustRaw sensor data is not stored on or leaves the mobile phoneThe content of a conversation or raw audio data is never disclosedUsers can choose to opt out of DarwinExperimental ResultsTested using a mixture of five N97 and iPhones used by eight people over a period of two weeksAudio recorded in different locationsClassifier trained indoorsExperiment 1 ParametersThree people walk along a sidewalk of a busy road and engage in conversationThe speaker recognition application without the Darwin components runs on each of the phones carried by the peopleExperiment 1 Results: Without EvolutionExperiment 2 ParametersMeeting setting in an office environment where 8 people are involved in conversationThe phones are located at different distances from people in the meeting, some on the table and some in people’s pocketsExperiment 2 ResultsExperiment 2 ResultsExperiment 3 ParametersFive phones in a noisy restaurantThree of the five people are engaged in conversationTwo of the five phones are placed on the tablePhone 4 Is the closest phone to speaker 4 and also the closest phone to another group of people having a loud conversationExperiment 3


View Full Document

UCF EEL 6788 - DARWIN PHONES - THE EVOLUTION OF SENSING AND INFERENCE ON MOBILE PHONES

Documents in this Course
Load more
Download DARWIN PHONES - THE EVOLUTION OF SENSING AND INFERENCE ON MOBILE PHONES
Our administrator received your request to download this document. We will send you the file to your email shortly.
Loading Unlocking...
Login

Join to view DARWIN PHONES - THE EVOLUTION OF SENSING AND INFERENCE ON MOBILE PHONES and access 3M+ class-specific study document.

or
We will never post anything without your permission.
Don't have an account?
Sign Up

Join to view DARWIN PHONES - THE EVOLUTION OF SENSING AND INFERENCE ON MOBILE PHONES 2 2 and access 3M+ class-specific study document.

or

By creating an account you agree to our Privacy Policy and Terms Of Use

Already a member?