Deep belief network-based hybrid model for multimodal biometric system for futuristic security applications

Abstract

Biometrics is the technology to identify humans uniquely based on face, iris, and fingerprints, etc. Biometric authentication allows the person recognition automatically on the basis of behavioral or physiological characteristics. Biometrics are broadly employed in several commercial as well as the official identification systems for automatic access control. This paper introduces the model for multimodal biometric recognition based on score level fusion method. The overall procedure of the proposed method involves five steps, such as pre-processing, feature extraction, recognition score using Multi- support vector neural network (Multi-SVNN) for all traits, score level fusion, and recognition using deep belief neural network (DBN). The first step is to input the training images into pre-processing steps. Thus, the pre-processing of three traits, like iris, ear, and finger vein is done. Then, the feature extraction is done for each modality to extract the features. After that, the texture features are extracted from pre-processed images of the ear, iris, and finger vein, and the BiComp features are acquired from individual images using a BiComp mask. Then, the recognition score is computed based on the Multi-SVNN classifier to provide the score individually for all three traits, and the three scores are provided to the DBN. The DBN is trained using the chicken earthworm optimization algorithm (CEWA). The CEWA is the integration of the chicken swarm optimization (CSO), and earthworm optimization algorithm (EWA) for the optimal authentication of the person. The analysis proves that the developed method acquired a maximal accuracy of 95.36%, maximal sensitivity of 95.85%, and specificity of 98.79%, respectively.

Introduction

In recent years, biometrics plays an important role in security systems. Biometrics is categorized into image-based systems and signal-based systems. Biometrics is a tool for differentiating subjects in a reliable manner using behavioural or physical traits [2]. Signal-based systems contain the identification of Electrocardiography (ECG) as well as the identification of the speaker. Image-based systems consist of gestures, hand-written signature, voice, hand geometry, gait, iris recognition, and face [1,35]. The biometric system is a promising and constantly evolving technology used in the automatic system to identify the person's efficiently and uniquely without the need to remember or carry anything, like Ids and passwords [3]. Several studies proved that iris trait contains a number of merits than other biometric system based on the features like face [36] and fingerprint [15], this makes the iris system to be commonly accepted in many applications to accurate and high-reliability biometric systems [3,37]. The biometric system is broadly categorized into two kinds, like multimodal, and unimodal biometric systems. The unimodal biometric framework establishes the identity of the person based on the single information source, such as left iris, right iris, and face [3]. In the multimodal biometric system, when the system works under-identification mode, the outcome of the classifier is viewed by a list of ranks obtained from candidates, which represents the possible matches [8]. Designing and implementing the multimodal biometric system requires a number of factors, which influence the overall performance of the system [3].

From the various biometric modalities studied nowadays, the iris, finger vein, and ear recognition are measured as the major reliable biometric modalities with less error rate [4]. Since 1987, Iris recognition is the popular biometric recognition system by Aran as well as Leonard [5,1]. In the last two years, iris recognition [7] has the fastest-growing field of research [6,7]. The purpose of iris recognition is to capture and analyze the images for identification purposes. Iris localization is the first important step for finding the outer and inner boundaries of the region of iris [8,9]. Iris recognition has been applied in several biometric applications, like border crossing control, intelligent unlocking, security [34] and crime screening, and border crossing control, and so on [10]. Another one possible biometric source is the ear. In machine vision, ear biometrics plays a very important role in forensic science. Ears have several benefits over more introduced biometrics; as Bertillon said that they have a stable and rich structure that is conserved from birth to old age [11]. In the last few years, finger vein images have gained more attention for personal authentication based on convenience and security as it uses the features within the human body [12].

Various approaches are used for the localization of iris, like Distance Regularized Level Set Evolution (DRLSE), integrodifferential operator, Circular Hough Transform (CHT), and Active Contour (AC). Tsai [13] uses a fuzzy matching strategy to identify the feature of iris points. Here, the similarity score table is utilized to compare the feature points in matching algorithms. Several types of research introduced neural networks for iris recognition [14]. The deep learning approach gained tremendous success in the computer vision area and accomplishes the existing performance in image classification [12]. The features of fingerprints and finger-veins are extracted by Gabor filter, and the hamming distance is employed for matching using binarization [14]. Some of the broadly utilized pattern representation techniques are Linear discriminant analysis (LDA) -kernel-based [18], and principal component analysis (PCA) -based [17] ear recognitions. In the earlier technique, PCA is applied for ear recognition in which the recognition performance is obtained under less condition and afterward approaches employs a combination of kernel, and LDA to solve the disadvantages posed by PCA [16]. The challenges faced by biometric authentication are concerning the leakage of data, which is obtained from the output of fake identification [3]. The widths and directions of finger veins differ, hence it is complex to find the best directions and frequencies of Gabor filter. If the iris image is obtained from non-ideal conditions then, segmentation and localization become very challenging [27]. The degradation of the image quality is high associated with the visible illumination of the eye image determined in dynamic environments [4]. Hence, there is a need for a biometric system, which overcome these challenges.

In this research paper, the multimodal biometric system is developed using score level fusion. The score-level fusion method is designed based on CEWA. The CEWA is the combination of CSO and EWA. Initially, the pre-processing is performed from the input images, and the feature extraction is performed individually for ear, finger vein, and iris images. After that, the extracted features from three modalities are given to the BiComp mask to obtain the BiComp features, and then the recognition score is estimated using the Multi-SVNN classifier based on individual features obtained from the ear, iris, and finger-vein. Then, the score level fusion is performed from three modalities, and the three scores are provided to the DBN for the optimal authentication of the person.

The main contribution of the research paper:

• Proposed a new hybrid algorithm, named CEWA, which is used to train the DBN classifier for the effective authentication of the person. The proposed CEWA is the combination of CSO and EWA

The paper is structured as follows: Section 2 deliberates the literature review of the existing methods of the multimodal biometric system, and the proposed DBN-based hybrid model is deliberated in Section 3. Section 4 describes the results and discussion of the proposed method, and finally, concludes the paper in Section 5.