EMS: EXPRESSION BASED MOOD SHARING FOR SOCIAL NETWORKS
Md Munirul HaqueMohammad AdibuzzamanDepartment of Mathematics, Statistics, and Computer ScienceMarquette University
OUTLINE Motivation State of the art Classification of models FACS Drawbacks Comparison Open issues System Overview System Architecture Implementation
MOTIVATION Facebook has 500 million active users twitter has 190 million visitors per month Number of smart phone users has crossed 45
millions Many mood applications in FB
My Mood SpongeBob Mood The Mood Weather Report Name and Mood Analyzer Manual setting
STATE OF THE ART Active Appearance Model (AAM) Computer Expression Recognition Tool (CERT) Eigenface, Eigeneye, Eigenlips Artificial Neural Network (ANN) Relevance Vector Machine (RVM)
CLASSIFICATION
3D
Automatic Facial Expression Detection
Principal Component Analysis (PCA)
DimensionFocusClassification Technique
Artificial Neural Network (ANN)
Linear Discriminant Analysis (LDA) 2DPatientsAdultsNeonates
FACS
FACS
DRAWBACKS Reliability on clear frontal image Out-of-plane head rotation Right feature selection Fail to use temporal and dynamic information Considerable amount of manual interaction Noise, illumination, glass, facial hair, skin
color issues Computational cost Mobility Intensity Reliability.
COMPARISON
EIGENFACE, EIGENEYE, EIGENLIPS
Eigenfaces for the training image set
CHARACTERISTICS Real Time Mood to Social Media Location Aware Sharing Mood Aware Sharing
Mobility Resources of Behavioral Research Context Aware Event Manager
OPEN ISSUES Deception of Expression (suppression,
amplification, simulation) Difference in Cultural, Racial, and Sexual
Perception Intensity Dynamic Features
SYSTEM OVERVIEW
SYSTEM ARCHITECTURE
AXIS2
Apache Tomcat Container
Application Server
SOAP/Web Service Engine
Expression Detection Script
Server
WAMP
PHP Web Server
Browser/MobileHTTP Call
Client
FIG: Expression Detection Architecture
MATLAB
JAVA Library runs on MCR
MATLAB Builder JA
TRAINING DATABASE
WEB CLIENT
FUTURE WORK Build a Facebook application which will
capture the user image using device camera(webcam or mobile camera).
Feed that image to the MATLAB Script and get Expression detected.
Do a survey on the user response of the Facebook application
Increase accuracy Images- not present in the database Confusion matrix
CONCLUSION Most computational research requires the
extensive ability of MATLAB for different computations like image processing, forecasting and other areas.
Using web service to run a MATLAB script will help do research on Computational sciences research.
Q/A Any question?? Comments Suggestion