DARWIN PHONES: THE EVOLUTION OF SENSING AND INFERENCE ON MOBILE PHONES PRESENTED BY: BRANDON OCHS Emiliano Miluzzo, Cory T. Cornelius, Ashwin Ramaswamy,

Slides:



Advertisements
Similar presentations
Darwin Phones: the Evolution of Sensing and Inference on Mobile Phones Emiliano Miluzzo *, Cory T. Cornelius *, Ashwin Ramaswamy *, Tanzeem Choudhury *,
Advertisements

Outline Activity recognition applications
SoNIC: Classifying Interference in Sensor Networks Frederik Hermans et al. Uppsala University, Sweden IPSN 2013 Presenter: Jeffrey.
Improvement of Audio Capture in Handheld Devices through Digital Filtering Problem Microphones in handheld devices are of low quality to reduce cost. This.
THE JIGSAW CONTINUOUS SENSING ENGINE FOR MOBILE PHONE APPLICATIONS Hong Lu,† Jun Yang,! Zhigang Liu,! Nicholas D. Lane,† Tanzeem Choudhury,† Andrew T.
D u k e S y s t e m s Sensing Meets Mobile Social Networks: The Design, Implementation and Evaluation of the CenceMe Application Emiliano Miluzzo†, Nicholas.
SoundSense: Scalable Sound Sensing for People-Centric Applications on Mobile Phones -Hong LU, Wei Pan, Nicholas D. Lane, Tanzeem Choudhury and Andrew T.
Activity, Audio, Indoor/Outdoor classification using cell phones Hong Lu, Xiao Zheng Emiliano Miluzzo, Nicholas Lane CS 185 Final Project presentation.
Sean Powers Florida Institute of Technology ECE 5525 Final: Dr. Veton Kepuska Date: 07 December 2010 Controlling your household appliances through conversation.
NEUROPHONE: BRAIN- MOBILE PHONE INTERFACE USING A WIRELESS EEG HEADSET Andrew T. Campbell, Tanzeem Choudhury, Shaohan Hu, Hong Lu, Matthew K. Mukerjee!,
SurroundSense: Mobile Phone Localization via Ambience Fingerprinting Written by Martin Azizyan, Ionut Constandache, & Romit Choudhury Presented by Craig.
Multiple Criteria for Evaluating Land Cover Classification Algorithms Summary of a paper by R.S. DeFries and Jonathan Cheung-Wai Chan April, 2000 Remote.
LYU0103 Speech Recognition Techniques for Digital Video Library Supervisor : Prof Michael R. Lyu Students: Gao Zheng Hong Lei Mo.
Top Level System Block Diagram BSS Block Diagram Abstract In today's expanding business environment, conference call technology has become an integral.
SENSING MEETS MOBILE SOCIAL NETWORKS: THE DESIGN, IMPLEMENTATION AND EVALUATION OF THE CENCEME APPLICATION Emiliano Miluzzo†, Nicholas D. Lane†, Kristóf.
A Practical Approach to Recognizing Physical Activities Jonathan Lester Tanzeem Choudhury Gaetano Borriello.
Fig. 2 – Test results Personal Memory Assistant Facial Recognition System The facial identification system is divided into the following two components:
1 HealthSense : Classification of Health-related Sensor Data through User-Assisted Machine Learning Presenter: Mi Zhang Feb. 23 rd, 2009 From Prof. Gregory.
Distributed and Efficient Classifiers for Wireless Audio-Sensor Networks Baljeet Malhotra Ioanis Nikolaidis Mario A. Nascimento University of Alberta Canada.
김덕주 (Duck Ju Kim). Problems What is the objective of content-based video analysis? Why supervised identification has limitation? Why should use integrated.
THE SECOND LIFE OF A SENSOR: INTEGRATING REAL-WORLD EXPERIENCE IN VIRTUAL WORLDS USING MOBILE PHONES Sherrin George & Reena Rajan.
WALRUS: Wireless Active Location Resolver with Ultrasound Tony Offer, Christopher Palistrant.
ALBERT PARK EEL 6788: ADVANCED TOPICS IN COMPUTER NETWORKS Energy-Accuracy Trade-off for Continuous Mobile Device Location, In Proc. of the 8th International.
A Survey of Mobile Phone Sensing Michael Ruffing CS 495.
Crowd++: Unsupervised Speaker Count with Smartphones Chenren Xu, Sugang Li, Gang Liu, Yanyong Zhang, Emiliano Miluzzo, Yih-Farn Chen, Jun Li, Bernhard.
SoundSense: Scalable Sound Sensing for People-Centric Application on Mobile Phones Hon Lu, Wei Pan, Nocholas D. lane, Tanzeem Choudhury and Andrew T. Campbell.
Sensing Meets Mobile Social Networks: The Design, Implementation and Evaluation of the CenceMe Application Emiliano Miluzzo†, Nicholas D. Lane†, Kristóf.
Ambulation : a tool for monitoring mobility over time using mobile phones Computational Science and Engineering, CSE '09. International Conference.
“SoundSense: Scalable Sound Sensing for People-Centric Applications on Mobile Phones” Authors: Hong Lu, Wei Pan, Nicholas D. Lane, Tanzeem Choudhury and.
Design, Implementation and Evaluation of CenceMe Application COSC7388 – Advanced Distributed Computing Presentation By Sushil Joshi.
SoundSense by Andrius Andrijauskas. Introduction  Today’s mobile phones come with various embedded sensors such as GPS, WiFi, compass, etc.  Arguably,
9 th Conference on Telecommunications – Conftele 2013 Castelo Branco, Portugal, May 8-10, 2013 Sara Candeias 1 Dirce Celorico 1 Jorge Proença 1 Arlindo.
VBS Documentation and Implementation The full standard initiative is located at Quick description Standard manual.
1 ENTROPY-BASED CONCEPT SHIFT DETECTION PETER VORBURGER, ABRAHAM BERNSTEIN IEEE ICDM 2006 Speaker: Li HueiJyun Advisor: Koh JiaLing Date:2007/11/6 1.
Energy Efficient Location Sensing Brent Horine March 30, 2011.
SocialWeaver: Collaborative Inference of Human Conversation Networks Using Smartphones Chengwen Luo and Mun Choon Chan School of Computing National University.
Nicholas D. Lane, Hong Lu, Shane B. Eisenman, and Andrew T. Campbell Presenter: Pete Clements Cooperative Techniques Supporting Sensor- based People-centric.
The Second Life of a Sensor: Integrating Real-World Experience in Virtual Worlds using Mobile Phones Mirco Musolesi, Emiliano Miluzzo, Nicholas D. Lane,
A Baseline System for Speaker Recognition C. Mokbel, H. Greige, R. Zantout, H. Abi Akl A. Ghaoui, J. Chalhoub, R. Bayeh University Of Balamand - ELISA.
Look who’s talking? Project 3.1 Yannick Thimister Han van Venrooij Bob Verlinden Project DKE Maastricht University.
Singer similarity / identification Francois Thibault MUMT 614B McGill University.
1 City With a Memory CSE 535: Mobile Computing Andreea Danielescu Andrew McCord Brandon Mechtley Shawn Nikkila.
Training Conditional Random Fields using Virtual Evidence Boosting Lin Liao, Tanzeem Choudhury †, Dieter Fox, and Henry Kautz University of Washington.
1.Research Motivation 2.Existing Techniques 3.Proposed Technique 4.Limitations 5.Conclusion.
Voice Activity Detection based on OptimallyWeighted Combination of Multiple Features Yusuke Kida and Tatsuya Kawahara School of Informatics, Kyoto University,
Providing User Context for Mobile and Social Networking Applications A. C. Santos et al., Pervasive and Mobile Computing, vol. 6, no. 1, pp , 2010.
Counting How Many Words You Read
Indoor Positioning System
ROVER TECHNOLOGY PRESENTED BY Gaurav Dhuppar Final Year I.T. GUIDED BY Ms. Kavita Bhatt Lecturer I.T.
Internet of Things. IoT Novel paradigm – Rapidly gaining ground in the wireless scenario Basic idea – Pervasive presence around us a variety of things.
Predicting Voice Elicited Emotions
Sensing Meets Mobile Social Networks: The Design, Implementation and Evaluation of the CenceMe Application Emiliano Miluzzo†, Nicholas D. Lane†, Kristóf.
Pocket, Bag, Hand, etc. - Automatically Detecting Phone Context through Discovery Emiliano Miluzzoy, Michela Papandreax, Nicholas D. Laney, Hong Luy, Andrew.
Efficient Opportunistic Sensing using Mobile Collaborative Platform MOSDEN.
Big Data Quality Challenges for the Internet of Things (IoT) Vassilis Christophides INRIA Paris (MUSE team)
Voice Activity Detection Based on Sequential Gaussian Mixture Model Zhan Shen, Jianguo Wei, Wenhuan Lu, Jianwu Dang Tianjin Key Laboratory of Cognitive.
Huber Flores Social-aware Hybrid Mobile Offloading A contribution for edge and fog computing? Huber Flores
Traffic State Detection Using Acoustics
Information Technology Deanship
Bag-of-Visual-Words Based Feature Extraction
Week 01 Comp 7780 – Class Overview.
Vijay Srinivasan Thomas Phan
Introduction to Pattern Recognition
Anindya Maiti, Murtuza Jadliwala, Jibo He Igor Bilogrevic
iSRD Spam Review Detection with Imbalanced Data Distributions
Title of poster... M. Author1, D. Author2, M. Author3
AUDIO SURVEILLANCE SYSTEMS: SUSPICIOUS SOUND RECOGNITION
Xin Qi, Matthew Keally, Gang Zhou, Yantao Li, Zhen Ren
John H.L. Hansen & Taufiq Al Babba Hasan
Sensor Networks – Motes, Smart Spaces, and Beyond
Presentation transcript:

DARWIN PHONES: THE EVOLUTION OF SENSING AND INFERENCE ON MOBILE PHONES PRESENTED BY: BRANDON OCHS Emiliano 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

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 gather information using a range of sensor data

What 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 fit

Common 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 inference

Types of Learning  Supervised: Given a fully-labeled training set  Semi-Supervised: Given a small training set that is evolved  Unsupervised: No training set is given

Darwin Steps  Evolution, Pooling, and Collaborative Inference  These represent Darwin’s novel evolve-pool-collaborate model implemented on mobile phones

Classifier 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 models

Model 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 servers

Collaborative 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 accuracy

Darwin 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 computed locally

Darwin 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 environments

Darwin 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 viewpoints

Speaker 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 blocks

Speaker Modeling  Feature vector consisting of Mel Frequency Cepstral Coefficients  Each speaker is modeled with 20 Gaussians  An initial speaker model is built by collecting a short training sample

Classifier 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 evolution

Classifier 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 evolved

Collaborative 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 broadcasted

Privacy 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 Darwin

Experimental 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 indoors

Experiment 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 people

Experiment 1 Results: Without Evolution

Experiment 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 pockets

Experiment 2 Results

Experiment 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 conversation

Experiment 3 Results

Experiment 4 Parameters  Five people walk along a sidewalk and three of them are talking  The greatest improvement is observed by speaker 1, whose phone is clipped to their belt

Experiment 4 Results

Time and Energy Measurements  Baselines for power use determined  Measurements performed using the Nokia Energy Profiler tool  No data gathered for the iPhone  Smart duty cycling required later to save battery life

Time and Energy Measurements

Possible Applications  Virtual square application  Social application for a group of friends  Place discovery application  Use collaborative inference to determine location  Friend Tagging application  Exploit face recognition to tag friends on pictures

Future Work  Duty cycling for improved battery life  Simplified classification techniques

Improvements On The Paper  Studies don’t show conclusive evidence; there should be separate control models for each of the scenarios

Conclusion  The Darwin system combines classifier evolution, model pooling, and collaborative inference  Results indicate that the performance boost offered by Darwin off sets problems with sensing context  The Darwin system provides a scalable framework that can be used for other urban sensing applications

References  [1] Emiliano 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  [2] H. Ezzaidi and J. Rouat. Pitch and MFCC Dependent GMM Models for Speaker Identification systems. In Electrical and Computer Engineering, Canadian Conference on, volume 1, 2004  [3] H. Ezzaidi and J. Rouat. Pitch and MFCC Dependent GMM Models for Speaker Identification systems. In Electrical and Computer Engineering, Canadian Conference on, volume 1, 2004.