Breadcrumb

 
 

User Preference Extraction from Bio-Signals: An Experimental Study

Title:

User Preference Extraction from Bio-Signals: An Experimental Study

Aurup, Golam Mohammad Moshiuddin (2011) User Preference Extraction from Bio-Signals: An Experimental Study. Masters thesis, Concordia University.

[img]
Preview
PDF - Accepted Version
3691Kb

Abstract

Abstract
User Preference Extraction from Bio-Signals: An Experimental Study

Golam Mohammad Moshiuddin Aurup

The purpose of this study is to extract user preferences about a product from emotional responses. Literature on psychology reveals that human preferences are related to their emotions. In addition, literature on emotion recognition reveals that emotion can be extracted from physiological signals like the heart rate, skin conductance, brain signal etc. In this study, these two streams were brought together and a new approach was proposed to extract preference of users through the analysis of emotional responses from physiological signals. Brain signal (electroencephalography or EEG) was chosen in this regard for its relevance in emotion recognition literature. For the experimental study, Thought Technology’s biofeedback system was used in order to capture and process the users’ EEG signals while users are exposed to various images where each image represents a possible feature of a product. Experiments were performed in two phases. In the first phase, proposed method, hardware setting, and the hypothesis relating user preference and EEG values were tested on an image library. In the second phase, images relating to a product (automobile in our case) design or features were used to get preferences between competing products.

In the first phase, International Affective Picture System (IAPS) library images were used to develop experiments. This library provides a large database for emotion affecting images and is widely accepted in the emotion detection literature. Three participants and three two-image sets were used in this study. The relationship between preference and extracted values were established through graph plot and trend line analysis and effects of repetition of experiments or images were identified. Results supported that analysis of EEG signals can distinguish pleasant and unpleasant feeling about images. A maximum of 80% accuracy was obtained in establishing relationship between preference and signal values. Left frontal side of the brain provided with the best results and was utilized in the rest of the study. Possibility to use different frequency bands of EEG signal was also studied in this phase.

In the second phase of experiments, 8 image-sets relating to automobile design and features were used for a group of 11 participants. 60% of the participants responded with 70% or more accuracy. It was found that the cognitive preference of a participant was stronger than the aesthetic preference whenever there was a conflict between the two. Accuracy rate showed by participants varied with the quality of the tests; i.e. with the factors like image resolution, clarity, composition, subject, and background of images; and with the capability of the participant to identify the images properly.

Literature on brain activity reports that, for some people, the left side of the brain is more active than the right one. The opposite is true for others. The hypothesis relating preference and extracted values was corrected in this regard. The corrected hypothesis was termed the reverse hypothesis. At the beginning of phase 2 experiments, 4 experiments were developed with IAPS images to identify if the participant followed the preliminary hypothesis or the reversed one. The results showed that most participants performed better in the experiments with product images than the experiments with standard IAPS images.

Divisions:Concordia University > Faculty of Engineering and Computer Science > Mechanical and Industrial Engineering
Item Type:Thesis (Masters)
Authors:Aurup, Golam Mohammad Moshiuddin
Institution:Concordia University
Degree Name:M.A. Sc.
Program:Industrial Engineering
Date:15 February 2011
Thesis Supervisor(s):Akgunduz, Ali
ID Code:7078
Deposited By:GOLAM MOHAMMAD AURUP
Deposited On:08 Jun 2011 16:03
Last Modified:08 Jun 2011 16:03
All items in Spectrum are protected by copyright, with all rights reserved. The use of items is governed by Spectrum's terms of access.

Repository Staff Only: item control page

Document Downloads

More statistics for this item...

Concordia University - Footer