Real time analysis of social behavior from video and kinect recordings

This paper reports and recapitulates the progress of author's dissertation entitled on "Real Time Analysis of Social Behavior from Video and Kinect Recording" under the scrutiny of Professor Justin Dauwels.This project comprises of the prospect of extraction of human emotion using eye tracker.In this part,raw data from eye movements is used for extraction of features. The latter part of project includes the association of eye tracker and Kinect in order to enhance the accuracy of detection. Consideration of subject's eye movement as well as facial expression fine-tunes the estimation accuracy which will be covered in Chapter 5.The prime objective of this project is to initially gather reliable data for eye movements using Eye Tracker.Following which, a well designated protocol shows the subject a blend of YouTube videos to induce different emotions including happy,sad and neutral.A Large database of Kinect sensor and recording and eye tracker data will be collected in collaboration with Nielsen. Finally,analysis will be performed using pattern recognition algorithms and machine learning classifiers to detect the emotion of the subject.In this project,machine learning classifiers such as K-nearest neighbors,Naive Bayes and C4.5 Decision tree learning has been used. The prime aim is to pin point the most seemly machine learning classifier for obtaining highest accuracy for classification.Master of Science (Signal Processing