Cover Image

EEG-Based Person Authentication Modelling Using Incremental Fuzzy-Rough Nearest Neighbour Technique

High level security has nurtured the arisen of Electroencephalograms (EEG) signals as a noteworthy biometrics modality for person authentication modelling. Modelling distinctive characteristics among individuals, especially in a dynamic environment involves incremental knowledge updates from time t...

Full description

Saved in:
Bibliographic Details
Main Author: Liew, Siaw Hong
Format: Thesis
Language:English
English
Published: 2016
Subjects:
R Medicine (General)
RC Internal medicine
Online Access:http://eprints.utem.edu.my/id/eprint/18351/1/EEG-Based%20Person%20Authentication%20Modelling%20Using%20Incremental%20Fuzzy-Rough%20Nearest%20Neighbour%20Technique.pdf
http://eprints.utem.edu.my/id/eprint/18351/2/EEG-Based%20Person%20Authentication%20Modelling%20Using%20Incremental%20Fuzzy-Rough%20Nearest%20Neighbour%20Technique.pdf
http://eprints.utem.edu.my/id/eprint/18351/
https://plh.utem.edu.my/cgi-bin/koha/opac-detail.pl?biblionumber=100138
Tags: Add Tag
No Tags, Be the first to tag this record!
id my.utem.eprints.18351
record_format eprints
spelling my.utem.eprints.183512021-10-10T15:30:52Z http://eprints.utem.edu.my/id/eprint/18351/ EEG-Based Person Authentication Modelling Using Incremental Fuzzy-Rough Nearest Neighbour Technique Liew, Siaw Hong R Medicine (General) RC Internal medicine High level security has nurtured the arisen of Electroencephalograms (EEG) signals as a noteworthy biometrics modality for person authentication modelling. Modelling distinctive characteristics among individuals, especially in a dynamic environment involves incremental knowledge updates from time to time. K-Nearest Neighbour (KNN) is a well-known incremental learning method which applies First-In-First-Out (FIFO) knowledge update strategy. However, it is not suitable for person authentication modelling because it cannot preserve the representative EEG signals patterns when individual characteristics changes over time. Fuzzy-Rough Nearest Neighbours (FRNN) technique is an outstanding technique to model uncertainty under an imperfect data condition. The current implementation of FRNN technique is not designed for incremental learning problem because there is no update function to incrementally reshape and reform the existing knowledge granules. Thus, this research aims to design an Incremental FRNN (IncFRNN) technique for person authentication modelling using feature extracted EEG signals from VEP electrodes. The IncFRNN algorithm updates the training set by employing a heuristic update method to maintain representative objects and eliminate rarely used objects. The IncFRNN algorithm is able to control the size of training pool using predefined window size threshold. EEG signals such as visual evoked potential (VEP) is unique but highly uncertain and difficult to process.There exists no consistant agreement on suitable feature extraction methods and VEP electrodes in the past literature. The experimental comparison in this research has suggested eight significant electrodes set located at the occipital area. Similarly, six feature extraction methods, i.e. Wavelet Packet Decomposition (WPD), mean of amplitude, coherence, crosscorrelation, hjorth parameter and mutual information were used construct the proposed person authentication model. The correlation-based feature selection (CFS) method was used to select representative WPD vector subset to eliminate redundancy before combining with other features. The electrodes, feature extraction, and feature selection analysis were tested using the benchmarking dataset from UCI repositories. The IncFRNN technique was evaluated using a collected EEG data from 37 subjects. The recorded datasets were designed in three different conditions of ambient noise influence to evaluate the performance of the proposed solution. The proposed IncFRNN technique was compared with its predecessor, the FRNN and IBk technique. Accuracy and area under ROC curve (AUC) were used to measure the authentication performance. The IncFRNN technique has achieved promising results. The results have been further validated and proven significant statistically using paired sample ttest and Wilcoxon sign-ranked test. The heuristic incremental update is able to preserve the core set of individual biometrics characteristics through representative EEG signals patterns in person authentication modelling. Future work should focus on the noise management in data acquisition and modelling process to improve the robustness of the proposed person authentication model. 2016 Thesis NonPeerReviewed text en http://eprints.utem.edu.my/id/eprint/18351/1/EEG-Based%20Person%20Authentication%20Modelling%20Using%20Incremental%20Fuzzy-Rough%20Nearest%20Neighbour%20Technique.pdf text en http://eprints.utem.edu.my/id/eprint/18351/2/EEG-Based%20Person%20Authentication%20Modelling%20Using%20Incremental%20Fuzzy-Rough%20Nearest%20Neighbour%20Technique.pdf Liew, Siaw Hong (2016) EEG-Based Person Authentication Modelling Using Incremental Fuzzy-Rough Nearest Neighbour Technique. Masters thesis, Universiti Teknikal Malaysia Melaka. https://plh.utem.edu.my/cgi-bin/koha/opac-detail.pl?biblionumber=100138
institution Universiti Teknikal Malaysia Melaka
building UTEM Library
collection Institutional Repository
continent Asia
country Malaysia
content_provider Universiti Teknikal Malaysia Melaka
content_source UTEM Institutional Repository
url_provider http://eprints.utem.edu.my/
language English
English
topic R Medicine (General)
RC Internal medicine
spellingShingle R Medicine (General)
RC Internal medicine
Liew, Siaw Hong
EEG-Based Person Authentication Modelling Using Incremental Fuzzy-Rough Nearest Neighbour Technique
description High level security has nurtured the arisen of Electroencephalograms (EEG) signals as a noteworthy biometrics modality for person authentication modelling. Modelling distinctive characteristics among individuals, especially in a dynamic environment involves incremental knowledge updates from time to time. K-Nearest Neighbour (KNN) is a well-known incremental learning method which applies First-In-First-Out (FIFO) knowledge update strategy. However, it is not suitable for person authentication modelling because it cannot preserve the representative EEG signals patterns when individual characteristics changes over time. Fuzzy-Rough Nearest Neighbours (FRNN) technique is an outstanding technique to model uncertainty under an imperfect data condition. The current implementation of FRNN technique is not designed for incremental learning problem because there is no update function to incrementally reshape and reform the existing knowledge granules. Thus, this research aims to design an Incremental FRNN (IncFRNN) technique for person authentication modelling using feature extracted EEG signals from VEP electrodes. The IncFRNN algorithm updates the training set by employing a heuristic update method to maintain representative objects and eliminate rarely used objects. The IncFRNN algorithm is able to control the size of training pool using predefined window size threshold. EEG signals such as visual evoked potential (VEP) is unique but highly uncertain and difficult to process.There exists no consistant agreement on suitable feature extraction methods and VEP electrodes in the past literature. The experimental comparison in this research has suggested eight significant electrodes set located at the occipital area. Similarly, six feature extraction methods, i.e. Wavelet Packet Decomposition (WPD), mean of amplitude, coherence, crosscorrelation, hjorth parameter and mutual information were used construct the proposed person authentication model. The correlation-based feature selection (CFS) method was used to select representative WPD vector subset to eliminate redundancy before combining with other features. The electrodes, feature extraction, and feature selection analysis were tested using the benchmarking dataset from UCI repositories. The IncFRNN technique was evaluated using a collected EEG data from 37 subjects. The recorded datasets were designed in three different conditions of ambient noise influence to evaluate the performance of the proposed solution. The proposed IncFRNN technique was compared with its predecessor, the FRNN and IBk technique. Accuracy and area under ROC curve (AUC) were used to measure the authentication performance. The IncFRNN technique has achieved promising results. The results have been further validated and proven significant statistically using paired sample ttest and Wilcoxon sign-ranked test. The heuristic incremental update is able to preserve the core set of individual biometrics characteristics through representative EEG signals patterns in person authentication modelling. Future work should focus on the noise management in data acquisition and modelling process to improve the robustness of the proposed person authentication model.
format Thesis
author Liew, Siaw Hong
author_facet Liew, Siaw Hong
author_sort Liew, Siaw Hong
title EEG-Based Person Authentication Modelling Using Incremental Fuzzy-Rough Nearest Neighbour Technique
title_short EEG-Based Person Authentication Modelling Using Incremental Fuzzy-Rough Nearest Neighbour Technique
title_full EEG-Based Person Authentication Modelling Using Incremental Fuzzy-Rough Nearest Neighbour Technique
title_fullStr EEG-Based Person Authentication Modelling Using Incremental Fuzzy-Rough Nearest Neighbour Technique
title_full_unstemmed EEG-Based Person Authentication Modelling Using Incremental Fuzzy-Rough Nearest Neighbour Technique
title_sort eeg-based person authentication modelling using incremental fuzzy-rough nearest neighbour technique
publishDate 2016
url http://eprints.utem.edu.my/id/eprint/18351/1/EEG-Based%20Person%20Authentication%20Modelling%20Using%20Incremental%20Fuzzy-Rough%20Nearest%20Neighbour%20Technique.pdf
http://eprints.utem.edu.my/id/eprint/18351/2/EEG-Based%20Person%20Authentication%20Modelling%20Using%20Incremental%20Fuzzy-Rough%20Nearest%20Neighbour%20Technique.pdf
http://eprints.utem.edu.my/id/eprint/18351/
https://plh.utem.edu.my/cgi-bin/koha/opac-detail.pl?biblionumber=100138
_version_ 1715193891611738112
score 13.211869

Similar Items