Comparison of PCA and 2DPCA Accuracy with K-Nearest Neighbor Classification in Face Image Recognition

Face recognition is a special pattern recognition for faces that compare input image with data in database. The image has a variety and has large dimensions, so that dimension reduction is needed, one of them is Principal Component Analysis (PCA) method. Dimensional transformation on image causes ve...

Full description

Saved in:
Bibliographic Details
Main Authors: Sri Sutarti, Anggyi Trisnawan Putra, Endang Sugiharti
Format: Article
Language:English
Published: Universitas Negeri Semarang 2019-05-01
Series:Scientific Journal of Informatics
Subjects:
Online Access:https://journal.unnes.ac.id/nju/index.php/sji/article/view/18553
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850281794986835968
author Sri Sutarti
Anggyi Trisnawan Putra
Endang Sugiharti
author_facet Sri Sutarti
Anggyi Trisnawan Putra
Endang Sugiharti
author_sort Sri Sutarti
collection DOAJ
description Face recognition is a special pattern recognition for faces that compare input image with data in database. The image has a variety and has large dimensions, so that dimension reduction is needed, one of them is Principal Component Analysis (PCA) method. Dimensional transformation on image causes vector space dimension of image become large. At present, a feature extraction technique called Two-Dimensional Principal Component Analysis (2DPCA) is proposed to overcome weakness of PCA. Classification process in 2DPCA using K-Nearest Neighbor (KNN) method by counting euclidean distance. In PCA method, face matrix is changed into one-dimensional matrix to get covariance matrix. While in 2DPCA, covariance matrix is directly obtained from face image matrix. In this research, we conducted 4 trials with different amount of training data and testing data, where data is taken from AT&T database. In 4 time testing, accuracy of 2DPCA+KNN method is higher than PCA+KNN method. Highest accuracy of 2DPCA+KNN method was obtained in 4th test with 96.88%. while the highest accuracy of PCA+KNN method was obtained in 4th test with 89.38%. More images used as training data compared to testing data, then the accuracy value tends to be greater.
format Article
id doaj-art-193a849f786241b2a3b44cf1849adb5f
institution OA Journals
issn 2407-7658
language English
publishDate 2019-05-01
publisher Universitas Negeri Semarang
record_format Article
series Scientific Journal of Informatics
spelling doaj-art-193a849f786241b2a3b44cf1849adb5f2025-08-20T01:48:09ZengUniversitas Negeri SemarangScientific Journal of Informatics2407-76582019-05-0161647210.15294/sji.v6i1.185538766Comparison of PCA and 2DPCA Accuracy with K-Nearest Neighbor Classification in Face Image RecognitionSri Sutarti0Anggyi Trisnawan Putra1Endang Sugiharti2Universitas Negeri SemarangUniversitas Negeri SemarangUniversitas Negeri SemarangFace recognition is a special pattern recognition for faces that compare input image with data in database. The image has a variety and has large dimensions, so that dimension reduction is needed, one of them is Principal Component Analysis (PCA) method. Dimensional transformation on image causes vector space dimension of image become large. At present, a feature extraction technique called Two-Dimensional Principal Component Analysis (2DPCA) is proposed to overcome weakness of PCA. Classification process in 2DPCA using K-Nearest Neighbor (KNN) method by counting euclidean distance. In PCA method, face matrix is changed into one-dimensional matrix to get covariance matrix. While in 2DPCA, covariance matrix is directly obtained from face image matrix. In this research, we conducted 4 trials with different amount of training data and testing data, where data is taken from AT&T database. In 4 time testing, accuracy of 2DPCA+KNN method is higher than PCA+KNN method. Highest accuracy of 2DPCA+KNN method was obtained in 4th test with 96.88%. while the highest accuracy of PCA+KNN method was obtained in 4th test with 89.38%. More images used as training data compared to testing data, then the accuracy value tends to be greater.https://journal.unnes.ac.id/nju/index.php/sji/article/view/18553face recognition, feature extraction, pca, 2dpca, k-nearest neighbor.
spellingShingle Sri Sutarti
Anggyi Trisnawan Putra
Endang Sugiharti
Comparison of PCA and 2DPCA Accuracy with K-Nearest Neighbor Classification in Face Image Recognition
Scientific Journal of Informatics
face recognition, feature extraction, pca, 2dpca, k-nearest neighbor.
title Comparison of PCA and 2DPCA Accuracy with K-Nearest Neighbor Classification in Face Image Recognition
title_full Comparison of PCA and 2DPCA Accuracy with K-Nearest Neighbor Classification in Face Image Recognition
title_fullStr Comparison of PCA and 2DPCA Accuracy with K-Nearest Neighbor Classification in Face Image Recognition
title_full_unstemmed Comparison of PCA and 2DPCA Accuracy with K-Nearest Neighbor Classification in Face Image Recognition
title_short Comparison of PCA and 2DPCA Accuracy with K-Nearest Neighbor Classification in Face Image Recognition
title_sort comparison of pca and 2dpca accuracy with k nearest neighbor classification in face image recognition
topic face recognition, feature extraction, pca, 2dpca, k-nearest neighbor.
url https://journal.unnes.ac.id/nju/index.php/sji/article/view/18553
work_keys_str_mv AT srisutarti comparisonofpcaand2dpcaaccuracywithknearestneighborclassificationinfaceimagerecognition
AT anggyitrisnawanputra comparisonofpcaand2dpcaaccuracywithknearestneighborclassificationinfaceimagerecognition
AT endangsugiharti comparisonofpcaand2dpcaaccuracywithknearestneighborclassificationinfaceimagerecognition