Biometric Access Control with High Dimensional Facial Features

Citation

Ooi, Shih Yin and Pang, Ying Han and Khor, Ean Yee (2016) Biometric Access Control with High Dimensional Facial Features. Information Security and Privacy, 9723. pp. 437-445. ISSN 0302-9743

[img] Text
121.pdf
Restricted to Repository staff only

Download (1MB)

Abstract

Access control is vital to prevent adversary from stealing resources from data centres. The security of traditional authentication means, such as password and Personal Identification Number (PIN), are imperfect for access control. In this paper, a reliable facial biometric access control with promising authentication performance is proposed. In our study, facial feature representation is computed based on ICA modelling, descriptor binarization, bitwise operation on the bit maps and effective compression via whitening PCA. The proposed technique is namely Binarized Independent Component Pattern (BICP). BICP training module integrates ICA methodology to construct ICA filter bank from natural image patches. Each face image is convoluted with the filters for the corresponding ICA responses. The ICA responses are further processed via feature binarization, and XOR bitwise operation before convert to code map. Next, block-wise histogramming is applied on each code map. By concatenating the regional histograms, it produces a set of high dimensional BICP descriptor, which will be further scaled and compressed. Empirical results show the remarkable performance of BICP on facial expression, illumination, time span and facial makeup effects.

Item Type: Article
Uncontrolled Keywords: Access control, Face biometric, ICA filters, XOR operation, Binary pattern
Subjects: T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK7800-8360 Electronics
Divisions: Faculty of Information Science and Technology (FIST)
Depositing User: Ms Rosnani Abd Wahab
Date Deposited: 09 Feb 2018 16:34
Last Modified: 09 Feb 2018 16:34
URII: http://shdl.mmu.edu.my/id/eprint/6681

Downloads

Downloads per month over past year

View ItemEdit (login required)