Fusion-Based Hand Geometry Recognition Using Dempster–Shafer Theory

Abstract

This paper presents a new technique for user identi¯cation and recognition based on the fusion of hand geometric features of both hands without any pose restrictions. All the features are extracted from normalized left and right hand images. Fusion is applied at feature and also at decision level. Two probability-based algorithms are proposed for classi¯cation. The¯rst algorithm computes the maximum probability for nearest three neighbors. The second algorithm determines the maximum probability of the number of matched features with respect to a thresholding on distances. Based on these two highest probabilities initial decisions are made. The¯nal decision is considered according to the highest probability as calculated by the Dempster-Shafer theory of evidence. Depending on the various combinations of the initial decisions, three schemes are experimented with 201 subjects for identi¯cation and veri¯cation. The correct identi¯cation rate is found to be 99.5%, and the false acceptance rate (FAR) of 0.625% has been found during veri¯cation. Int. J. Patt. Recogn. Artif. Intell. 2015.29. Downloaded from www.worldscientific.com by UNIVERSITY OF OTAGO on 07/12/15. For personal use only.knowledge-based (password) or token-based (PIN) user veri¯cation systems in various required levels (low to high level) of security intelligence. 9 It is applied as one of the best reliable and legitimate human authentication systems in a constrained environment. The primary objective is to discriminate the identity of a person based on various unique biometric properties recognized by an automated system. These systems are developed by the physical (e.g. face,¯ngerprint, hand geometry, hand vein, etc.) and behavioral (e.g. gait, signature, voice, etc.) distinctiveness of an individual. 10 Di®erent human organs are employed individually (unimodal) or combined (multimodal) for this purpose. A standalone recognition decision by a unimodal system is not always reliable and robust enough to verify whether a person is genuine or imposter. So to enhance the performance, fusion 17 is very useful to authenticate an individual with multibiometrics. It basically combines several decisions taken by various expert systems. Multibiometric systems provide certain bene¯ts over the limitations of unimodal systems such as noise-e®ect, intra-class variations, non-universality, spoof-attack and°exibility. 9 Di®erent multibiometric systems are accessible and are classi¯ed according to their basic properties (e.g. multi-sensor) and level(s) of implementation (e.g. decision level). An enormous number of biometric systems with di®erent characteristics and functionalities are running successfully worldwide for several decades in the government (National ID cards), forensics (criminal investigations) and commercial applications (smart cards). 19

Personal recognition system using hand modality based on local features

“…Indeed, Kang and Wu [21] have achieved 3.69% of EER by fusing Fourier descriptors and finger area functions extracted from fingers geometry that are acquired from 638 subjects. Whereas, Bera et al [22] have achieved FAR=0.625% and FRR=0.5% using geometrical features extracted nevertheless from only 201 subjects. Since our results are achieved using a larger database containing 642 subjects, the performance obtained from the proposed system (EER=2.25%) is competitive and encouraging results are provided.…”

Two-stage human verification using HandCAPTCHA and anti-spoofed finger biometrics with feature selection

“…Based on ICA, similar work is presented in (El-Sallam et al, 2011). Also, hand biometric methods developed at various levels of fusion are studied (Bera et al, 2015). A score-level fusion with four fingers is developed using cumulative angular function based FDs, which are computed from finger contour and finger area (Kang and Wu, 2014).…”

Spoofing detection on hand images using quality assessment

“…In [32], a feature-level fusion method using hand shape, geometry, and palm print features is performed. Decision-level fusion using the hand geometric feature is presented in [7]. On the other research line, though, hand-based modalities such as fingerprint, palmprint, and finger-vein have been tested for PAD.…”