Article Preview
TopIntroduction
Biometrics consists of the mathematical analysis of the biological characteristics of a person and it has for objective to determine its identity in an irrefutable way (Patil, 2016).
Contrary to what we know or what we possess, the biometrics is based on what we are and so allows avoiding the duplication, the theft, the forgetting or the loss. Unimodal biometric systems suffer from a variety of problems such as noise in sensed data, non-universality, inter-class similarities, and spoof attacks (H. Benaliouche, 2014). Multimodal biometric systems are a relatively new approach to overcome these limitations by assuring important improvements over the unimodal biometric system, in terms of higher accuracy and resistance to spoofing.
Multimodal biometric systems have been subjected to many research efforts and several approaches have been proposed to build these systems by efficiently combining biometric data from several sensors. In (Wang, 2008), the author proposed a robust multimodal biometric authentication scheme that integrates iris, face and palmprint data based on multiple parallel support vector machines (SVMs) score level fusion.
Kisku et al. (Kisku, 2011) presented multimodal biometric approaches which integrate face and palmprint at the feature extraction level. The system has attained 98.75% recognition rate with 0% false acceptance rate (FAR).
In (Maryam, 2012), Maryam et al. proposed the fusion of face and iris to obtain a robust recognition system.
Global (subspace linear discriminant analysis) and local (local binary pattern) feature extractors are used to extract face and iris features independently.
Face and iris scores are normalized using hyperbolic tangent (tanh) normalization, and then the fusion of these two modalities is performed using weighted sum rule.
Aly et al. (Aly, 2013), proposed a new approach for adaptive combination of Iris, Palmprint, Finger_Knuckle biometrics to ensure multiple levels of security. The score level fusion rule is adapted using Particle Swarm Optimization (PSO).
Maryam et al. (Maryam, 2014) proposed face-iris multimodal biometric system using score level and feature level fusion. Principal Component Analysis (PCA), subspace Linear Discriminant Analysis (LDA), subpattern-based PCA, modular PCA and Local Binary Patterns (LBP) are global and local feature extraction methods applied on face and iris images. Feature selection is performed using PSO at feature level fusion step to reduce the dimension of feature vectors. The matching score level fusion scheme using weighted sum rule and tanh normalization.
Cherifi et al. (Cherifi, 2015) proposed the use of a hybrid algorithm GA-PSO to find the optimum weights associated with the modalities being fused at the score level.
Singla et al. (Singla, 2016) proposed a multimodal biometric fusion system that fuses results from both Wavelets transform and PCA. This system combines three biometric traits, including face, fingerprint and iris.
Patil et al. (Patil, 2016) presented a fusion of face and palmprint at the sensor level, feature extraction level, matching the score level and decision level. For feature extraction LDA (Linear Discriminate Analysis) is used for face and palmprint biometric.
Kumar et al. (Kumar, 2017) combines fingerprint and iris at the matching score level for automatic identification of an individual. Minutiae matching and Edge detection techniques are used in this purpose.