International Journal of Biometrics (11 papers in press)
Human Age Classification using Appearance Features and Artificial Neural Network
by Jayant Jagtap, Manesh Kokare
Abstract: This paper presents a novel method for human age classification via face images by a computer. The proposed method classifies the human face images into four age groups: child, young, adult and senior adult by using appearance features as ageing features and Artificial Neural Network (ANN) as age classifier. The appearance features consists of both shape and textural features. Only two geometric ratios in combination with newly introduced rotation, scale and translation invariant efficient feature face angle are used as shape features. Local Binary Pattern Histogram (LBPH) of regions of interest in face image are used as textural features. The ANN is designed by using two layer feedforward back propagation neural networks. The performance of proposed age classification system is evaluated on face images from FG-NET ageing database and achieved greatly improved accuracy of 91.09% and 88.18% for male and female respectively.
Keywords: Age classification; Appearance features; Artificial neural networkrn(ANN); Local Binary Pattern Histogram (LBPH).
Undecimated Discrete Wavelet Transform for Touchless 2D Fingerprint Identification
by Salah Ahmed Saeed Othman, Tarik Boudghene Stambouli
Abstract: Several recent research efforts in biometrics have focused on developing the touchless fingerprint identification system. Most of them using imaging resulting from cameras and mobile devices. The acquired images are firstly subjected to robust preprocessing steps to localize region of interest in order to extract its features. In the literature, touchless fingerprint features are generally based on algorithms designed for minutiae analysis in touch-based images. Because of perspective distortions and deformations in the samples, minutiae-based techniques can obtain poor results. This paper investigates multi-resolution decomposition features to overcome the limitations of using traditional minutiae algorithms in term of accuracy and matching speed. These decompositions are implemented on Hong Kong polytechnic university 2D touchless fingerprint database that contains 10080 images. Experimental results illustrate successful use of Undecimated Discrete Wavelet Transform (UDWT) and Discrete Wavelet Packet Transform (DWPT) which give better performance than Discrete Wavelet Transform (DWT) and minutiae based method with less calculation cost.
Keywords: biometrics; touchless fingerprint identification; fingerprint features extraction; wavelet transform; multi-resolution decomposition; minutiae feature.
Application of geometry to RGB images for facial landmark localization A preliminary approach
by Federica Marcolin, Enrico Vezzetti, Pietro Maroso
Abstract: This study proposes a novel approach to automatically localize 11 landmarks from facial RGB images. The novelty of this method relies on the application, i.e. point-by-point mapping, of 11 Differential Geometry descriptors such as curvatures to the three individual RGB image components. Thus, three-dimensional features are applied to bidimensional facial image representations and used, via thresholding techniques, to extract the landmark positions. The method was tested on the Bosphorus database and showed global average errors lower than 5 millimetres.
The idea behind this approach is to embed this methodology in state-of-the-art 3D landmark detection methods to accomplish a full automatic landmarking by exploiting the advantages of both 2D and 3D data. Some landmarks such as pupils are arduous to be automatically extracted only via three-dimensional techniques. Thus, this method is intended as a bridging-the-gap preliminary technique that takes advantages of 2D imaging only for integrating advanced landmark localization methodologies.
Keywords: Facial Landmarks; Landmark Localization; Face Analysis; RGB Images; Differential Geometry.
Gait Recognition Based on Model-Based Methods and Deep Belief Networks
by Benouis Mohamed, Senouci Mohamed, Tlemsani Redouane, Mostefai Lotfi
Abstract: The sensitivity to illumination variations, pose, gender, age, clothing and any another source of changes, can be one of the most important challenges, in gait recognition system. In this paper, we adopt many approaches to extract signatures of human body (static model) using a model-based method, such as static body parameters, ellipse-fitting and Robust shape coding. To reduce the dimension of this features set, a principal component analysis (PCA) technique is employed. Then, a deep belief networks classifier is used to classify the gait signatures. The performance of the deep belief network (DBN) is superior to other classifiers such as k-nearest neighbor (KNN) and dynamic times warping (DTW). The comparison is performed for viewpoint changes, clothing and carrying conditions. The proposed approach has been validated on the Gait Database B.
Keywords: Biometrics; gait; model-based; model free; feature extraction; principal component analysis; k-nearest neighbour; dynamic times warping; deep belief network.
Palmprint Recognition through the Fractal Dimension Estimation for Texture Analysis
by Raouia Mokni, Monji Kherallah
Abstract: Palmprint is a human physiological feature which can distinguish and identify one person from another. In the palmprint recognition biometric systems, the feature extraction is considered as the most important step. In this paper, we use the fractal approach which is both a very advanced and sophisticated method in order to extract the palmprint texture information features. This fractal approach has been widely used in recent years being considered as an active research area in image processing field. Therefore, we have implemented a new technique to extract the texture palmprint features: the Texture Analysis basing on the Fractal Dimension estimated via the Box-Counting method or TAFD-BC. Experimental results on the PolyU 2D Palmprint database prove that our proposed approach produces promising and favorable results compared to other well-known state-of-the-art techniques.
Keywords: Biometric System; Palmprint; Texture Analysis; Fractal Dimension; Box Counting; Identification.
Human Gender Classification: A Review
by Feng Lin, Wenyao Xu
Abstract: The gender recognition is essential and critical for many applications in the commercial domains such as applications of human-computer interaction and computer-aided physiological or psychological analysis, since it contains a wide range of information regarding the characteristics difference between male and female. Some have proposed various approaches for automatic gender classification using the features derived from human bodies and/or behaviors. First, this paper introduces the challenge and application of gender classification research. Then, the development and framework of gender classification are described.We compare these state-of-the-art approaches, including vision-based methods, biological information-based methods, and social network information-based methods, to provide a comprehensive review of gender classification research. Next we highlight the strength and discuss the limitation of each method. Finally, this review also discusses several promising applications for future work.
Keywords: Gender classification; vision-based feature; biometrics; bio-signals; social network information.
T-norm based classifier and FAST Features for Illumination Invariant Face Recognition
by Aruna Bhat
Abstract: Various illumination invariant face recognition approaches have been proposed in the past. However the issue that still remains unaddressed to a large extent is the applicability of such techniques in practical environment where computational costs and time are crucial factors in deciding the implementation and use of a face recognition system. To deal with these issues, the paper presents a technique to achieve high recognition performance using robust methodologies for feature extraction and classification which at the same time are also quick in delivering results. Features from Accelerated Segment Test (FAST) method is used to identify the fiducial points. Features detected in an image using FAST are known to be resilient against different kinds of variations. Also the features are detected swiftly making the FAST algorithm true to its name which encourages its usability in practical scenarios. Fiducial points are the interest points that have a well-defined position and are robustly detected. Such points identified by FAST have been found to be reasonably stable in spite of changes occurring in the image owing to variations in illumination. To boost the stability of the method, the conventionally used ID3 algorithm has been improved by using non-extensive entropy instead of Renyi entropy. Further, a classification method based on t-norms is used. This classifier is primarily based on the errors between training features and test image features which are evaluated using the triangular/t-norms. The uniqueness of the proposed method lies in its dual advantage. One is in terms of quick detection of many such features which are least susceptible to be altered by the changes in illumination conditions. Another aspect is that the robust features so detected are fed to a unique swift classifier.
Keywords: Features from Accelerated Segment Test; T-norm; ID3; Fiducial points; Frank T-norm.
Partial Silhouette-based Gait Recognition
by Soharab Hossain Shaikh, Khalid Saeed, Nabendu Chaki
Abstract: Gait analysis refers to identification of a person from the systemic study of the motions of his/her different body parts at the time of walking. It is one of the behavioural biometric approaches for human identification. Like other behavioural approaches, gait also suffers from low-repeatability leading to poor recognition accuracy. Therefore, multimodal systems are required for better recognition accuracy. With the advent of multimodal biometric systems there is a need for low-cost methods for individual biometric modalities so that the overall complexity of the system does not overshoot the real-time requirements. In view of this, a partial-silhouette-based approach for gait recognition is reported in this article. This approach is translation, rotation and scale invariant and is low-cost in terms of computational complexity. Experimental results and comparative performance analysis on benchmark dataset reveal the potential of the partial-silhouette-based approach.
Keywords: Gait analysis; appearance-based approach; human identification; partial-silhouette-based approach; gait biometrics.
A Clustering based Indexing Approach for Biometric Databases using Decision-Level Fusion
by Ilaiah Kavati, Munaga VNK Prasad, Chakravarthy Bhagvati
Abstract: In this paper, we proposes a clustering-based indexing mechanism for biometric databases. The proposed technique relies mainly on a small set of preselected images called representative images. First, the database is partitioned into set of clusters and one image from each cluster is selected for the representative image set. Then, for each image in the database, an index code is computed by comparing it against the representative images. Further, an efficient storage structure (i.e., index space) is developed and the biometric images are arranged in it like traditional database records so that a quick search is possible. During identification, a list of candidates which are very similar to the query are retrieved from the index space. Further, to make full use of the clustering, we also retrieve the candidate identities from the selected clusters which are similar to query. Finally, the candidate identities from the index space and cluster space are fused using decision-level fusion. Experimental results on different databases shows a significant performance improvement in terms of response time and identification accuracy compared to the existing indexing methods.
Keywords: Clustering; Indexing; Representative images; Match scores; Decision-level fusion; Palmprints; Hand veins.
Extraction and selection of Binarized Statistical Image Features for fingerprint recognition
by Ahlem Adjimi, Abdenour Hacine-Gharbi, Philippe Ravier, Messaoud Mostefai
Abstract: Due to their simplicity and efficiency, image-based descriptors are currently very used in the task of fingerprint recognition. Among these descriptors, the histogram based descriptors are of widespread use. In this work, we use a novel histogram based descriptor called Binarized Statistical Image Features (BSIF). The gray value of each pixel is passed through a data-learned linear filter which output is converted into a binary code string. BSIF is a texture descriptor similar to the Local Binary Patterns (LBP) and Local Phase Quantization (LPQ), but the difference is the way the filters are learned. The BSIF descriptor depends on two main parameters which are the filter window size and the number of bits in the binary code string. In this work, we have evaluated the BSIF descriptor on the standard FVC2002 DB1 and DB4 databases, using different filter sizes with different bit lengths. We have extracted the BSIF histograms from sub-images around the core point of the fingerprint image and concatenated them to construct the final features vector. The experiments have shown that an increasing number of extracted sub-images lead to an increasing recognition rate. But increasing the number of sub-images also leads to higher dimension histogram estimations which decreased performance of the system regarding computing time and memory capacity. To avoid this problem we have used a feature selection method called Interaction capping (ICAP) which selects the relevant bins of the BSIF histogram based on the mutual information measure. This procedure permits to reduce the dimensionality of the BSIF histogram. The obtained results clearly showed that using the BSIF descriptor as a novel fingerprint features extraction method had a good effect on the recognition rates which outperformed the rates obtained with LBP or LPQ descriptors except for the smallest code string (length of 5). The results also showed that using feature selection method could reduce the dimensionality leading to a less computational complexity of the system.
Keywords: Biometrics; fingerprints; fingerprint identification;features extraction; BSIF; LBP; LPQ; texture descriptors; histogram; recognition rate; feature selection; mutual information; ICAP.
A fingerprint matching algorithm using bit-plane extraction method with phase-only correlation
by Florence Francis-Lothai, David B.L. Bong
Abstract: This paper introduces a new method in fingerprint feature extraction based on bit-plane. A bit-plane image requires smaller storage space than a grayscale image. Region of Interest (ROI) of a fingerprint image is extracted by using a modified blob analysis method, then the core point of the ROI is detected for dimension reduction process. Before bit-plane is extracted, the fingerprint image is enhanced by using Fourier transform. Bit-plane 7 of the enhanced image is used as the input for fingerprint matching with Phase-Only Correlation (POC) function. Experiment results showed that the storage space requirement for a fingerprint database could be reduced up to 87% per image for a bit-plane image compared with a grayscale image. The proposed fingerprint matching algorithm achieves 81.16% of recognition rate on FVC2002-Db1a database and 89.78% on FingerDOS database.
Keywords: bit-plane; biometric; feature extraction; fingerprint matching algorithm; Phase-Only correlation.