DSpace About DSpace Software

DSpace on Chitkara University >
Thesis - Chitkara University Himachal Pradesh >
Phd Thesis 2015 HP >

Please use this identifier to cite or link to this item: http://dspace.chitkara.edu.in/xmlui/handle/1/568

Issue Date: 31-Dec-2014
Publisher: Chitkara University Publications
Abstract: The most unstoppable and uncontrollable aspect of the mental state of humans is emotion. The emotions cannot be changed by a device but an effort can be made to predict or estimate emotion. The estimating of emotions is likely to be helpful because emotion is regarded as one of the profound factors, which influence the everyday life activities that can help us judge or choose between the available choices; and can alert us to avoid danger. The estimation of emotions can be done by using different methodologies, such as detecting facial expressions, voice estimation in stress, and measuring the changes in the physiological signals (temperature, heart rate, and GSR). This research focuses on the physiological parameters, which are measured using the sensors to decide the emotive status of a human being. The signals obtained using these measurements are authentic; though not alterable, these cannot be concealed during the measurement process, as these are generated due to the activation of the sympathetic nerves of Autonomous Nervous System (ANS).The bio-sensors have the advantage of monitoring the physiological parameters of the body (the physiological parameters are directly controlled by the autonomous nervous system and are affected by emotions). When different emotions are experienced by a human body then physiological changes are observed in terms of the skin conductance(GSR), blood volume pulse (BVP), and temperature. The proposed system design is a low power, portable, and cost effective embedded system for estimating the emotions based on collected data. It senses different emotions by using the machine learning algorithms embedded within a microcontroller (MSP430F2013).The Naïve Bayes classifier is one of the probability models that incorporates the class conditional assumptions and gives an output in the form of predicted emotion based on the data collected in the past. The algorithm’s accuracy xx improves as more the system collects more data. This predicted emotion by the Bayesian method becomes the input for the next machine learning algorithm, the Markov model. The Markov model is also implemented in microcontroller for predicting the emotion based on collected data. That means that both the algorithms are fused together to make them work in a hybrid form. For a proper fusion, a new algorithm called HYB-NAVMAR was designed to support the hybrid form. Our experimental results with the implementations of these algorithms on a TI MSP430 shows that the emotion can be predicted reasonably accurately as based on the collected data and the accuracy of prediction improves as the system collects more data.
URI: http://dspace.chitkara.edu.in/xmlui/handle/1/568
Appears in Collections:Phd Thesis 2015 HP

Files in This Item:

File Description SizeFormat
thesis_tanux.pdf3.26 MBAdobe PDFView/Open
View Statistics

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.


Valid XHTML 1.0! DSpace Software Copyright © 2002-2010  Duraspace - Feedback