National security today requires identification of people, things and activities. Biometrics plays an important role in the identification of people, and indirectly, in the identification of things and activities. Therefore, the development of technology and systems that provide faster and more accurate biometric identification is critical to the defense of our country. In addition, the development of a broad range of biometrics is necessary to provide the range of options needed to address flexible and adaptive adversaries. This paper will discuss the importance of a number of critical areas in the development of an environment to support biometrics, including research and development, biometric education, standards, pilot projects, and privacy assurance.

Protecting portable devices is becoming more important, not only because of the value of the devices themselves, but for the value of the data in them and their capability for transactions, including m-commerce and m-banking. An unobtrusive and natural method for identifying the carrier of portable devices is presented. The method uses acceleration signals produced by sensors embedded in the portable device. When the user carries the device, the acceleration signal is compared with the stored template signal. The method consists of finding individual steps, normalizing and averaging them, aligning them with the template and computing cross-correlation, which is used as a measure of similarity. Equal Error Rate of 6.4% is achieved in tentative experiments with 36 test subjects.

As early as the 1970's it was determined that gait, or the "manner of walking" is an identifying feature of a human being. Since then, extensive research has been done in the field of computer vision to determine how accurately a subject can be identified by gait characteristics. This has necessarily led to the study of how various data collection conditions, such as terrain type, varying camera angles, or a carried briefcase, may affect the identifying features of gait. However, little or no research has been done to question whether such conditions may be inferred from gait analysis. For example, is it possible to determine characteristics of the walking surface simply by looking at statistics derived from the subject's gait? The question to be addressed is whether significant concealed weight distributed on the subject's torso can be discovered through analysis of his gait. Individual trends in subjects in response to increasing concealed weight will be explored, with the objective of finding universal trends that would have obvious security purposes.

This paper describes a novel probabilistic framework for biometric identification and data fusion. Based on intra and inter-class variation extracted from training data, posterior probabilities describing the similarity between two feature vectors may be directly calculated from the data using the logistic function and Bayes rule. Using a large publicly available database we show the two imbalanced gait modalities may be fused using this framework. All fusion methods tested provide an improvement over the best modality, with the weighted sum rule giving the best performance, hence showing that highly imbalanced classifiers may be fused in a probabilistic setting; improving not only the performance, but also generalized application capability.

In this paper, we investigate the accuracy of using a partial iris image for identification and determine which portion of the iris has the most distinguishable patterns. Moreover, we compare these results with the results of Du et. al. using the CASIA database. The experimental results show that it is challenging but feasible to use only a partial iris image for human identification.

Iris recognition imaging is attracting considerable interest as a viable alternative for personal identification and verification in many defense and security applications. However current iris recognition systems suffer from limited depth of field, which makes usage of these systems more difficult by an untrained user. Traditionally, the depth of field is increased by reducing the imaging system aperture, which adversely impacts the light capturing power and thus the system signal-to-noise ratio (SNR). In this paper we discuss a computational imaging system, referred to as Wavefront Coded(R) imaging, for increasing the depth of field without sacrificing the SNR or the resolution of the imaging system. This system employs a especially designed Wavefront Coded lens customized for iris recognition. We present experimental results that show the benefits of this technology for biometric identification.

In this paper, we describe and analyze the performance of two iris-encoding techniques. The first technique is based on Principle Component Analysis (PCA) encoding method while the second technique is a combination of Principal Component Analysis with Independent Component Analysis (ICA) following it. Both techniques are applied globally. PCA and ICA are two well known methods used to process a variety of data. Though PCA has been used as a preprocessing step that reduces dimensions for obtaining ICA components for iris, it has never been analyzed in depth as an individual encoding method. In practice PCA and ICA are known as methods that extract global and fine features, respectively. It is shown here that when PCA and ICA methods are used to encode iris images, one of the critical steps required to achieve a good performance is compensation for rotation effect. We further study the effect of varying the image resolution level on the performance of the two encoding methods. The major motivation for this study is the cases in practice where images of the same or different irises taken at different distances have to be compared. The performance of encoding techniques is analyzed using the CASIA dataset. The original images are non-ideal and thus require a sequence of preprocessing steps prior to application of encoding methods. We plot a series of Receiver Operating Characteristics (ROCs) to demonstrate various effects on the performance of the iris-based recognition system implementing PCA and ICA encoding techniques.

Iris recognition has been demonstrated to be an efficient technology for doing personal identification. In this work, a method to perform iris recognition using biorthogonal wavelets is introduced. Effective use of biorthogonal wavelets using a lifting technique to encode the iris information is demonstrated. This new method minimizes built in noise of iris images using in-band thresholding in order to provide better mapping and encoding of the relevant information. Comparison of Gabor encoding, similar to the method used by Daugman and others, and biorthogonal wavelet encoding is performed. While Daugman's approach is a well-proven algorithm, the effectiveness of our algorithm is shown for the CASIA database, based on the ability to classify inter and intra class distributions, and may provide more flexibility for non-ideal images. The method was tested on the CASIA dataset of iris images with over 4,536 intra-class and 566,244 inter-class comparisons made. After calculating Hamming distances and for the selected threshold value of 0.4, FRR and FAR values were 13.6% and 0.6% using Gabor filter and 0% and 0.03% using the biorthogonal wavelets.

We show that minutiae information can reveal substantial details such as the orientation field and the class of the associated fingerprint that can potentially be used to reconstruct the original fingerprint image. The proposed technique utilizes minutiae triplet information to estimate the orientation map of the parent fingerprint. The estimated orientation map is observed to be remarkably consistent with the underlying ridge flow. We next discuss a classification technique that utilizes minutiae information alone to infer the class of the fingerprint. Preliminary results indicate that the seemingly random minutiae distribution of a fingerprint can reveal important class information. Furthermore, contrary to what has been claimed by several minutiae-based fingerprint system vendors, we demonstrate that the minutiae template of a user may be used to reconstruct fingerprint images.

The level of performance of a biometric fingerprint sensor is critically dependent on the quality of the fingerprint images. One of the most common types of optical fingerprint sensors relies on the phenomenon of total internal reflectance (TIR) to generate an image. Under ideal conditions, a TIR fingerprint sensor can produce high-contrast fingerprint images with excellent feature definition. However, images produced by the same sensor under conditions that include dry skin, dirt on the skin, and marginal contact between the finger and the sensor, are likely to be severely degraded. This paper discusses the use of multispectral sensing as a means to collect additional images with new information about the fingerprint that can significantly augment the system performance under both normal and adverse sample conditions. In the context of this paper, "multispectral sensing" is used to broadly denote a collection of images taken under different illumination conditions: different polarizations, different illumination/detection configurations, as well as different wavelength illumination. Results from three small studies using an early-stage prototype of the multispectral-TIR (MTIR) sensor are presented along with results from the corresponding TIR data. The first experiment produced data from 9 people, 4 fingers from each person and 3 measurements per finger under "normal" conditions. The second experiment provided results from a study performed to test the relative performance of TIR and MTIR images when taken under extreme dry and dirty conditions. The third experiment examined the case where the area of contact between the finger and sensor is greatly reduced.

Dynamic thresholds are a popular and effective means of binarisation, used in image processing. We show that the method is a special case of a general setting in which input data from a background window are fitted with a polynomial surface after which the data from a smaller, embedded focus window are thresholded with respect to the fitted surface. The method has been implemented and used with good results in the context of fingerprint recognition.

The purpose of this study is to quantitatively analyze the effect of fingerprint image quality on the performance of fingerprint recognition, in order to improve the identification capability of fingerprint recognition systems. In addition, this study proposes a new measure of fingerprint image quality using the Fourier spectrum. The proposed method measures the fingerprint image quality based on the global characteristics of the image. The experimental results demonstrate that the proposed algorithm shows better performance in the quality classification of fingerprint images than the existing algorithms, and this leads to a 6% improvement in the false rejection rate.

Current two-dimensional image based face recognition systems encounter difficulties with large variations in facial appearance due to the pose, illumination and expression changes. Utilizing 3D information of human faces is promising for handling the pose and lighting variations. While the 3D shape of a face does not change due to head pose (rigid) and lighting changes, it is not invariant to the non-rigid facial movement and evolution, such as expressions and aging effect. We propose a facial surface matching framework to match multiview facial scans to a 3D face model, where the (non-rigid) expression deformation is explicitly modeled for each subject, resulting in a person-specific deformation model. The thin plate spline (TPS) is applied to model the deformation based on the facial landmarks. The deformation is applied to the 3D neutral expression face model to synthesize the corresponding expression. Both the neutral and the synthesized 3D surface models are used to match a test scan. The surface registration and matching between a test scan and a 3D model are achieved by a modified Iterative Closest Point (ICP) algorithm. Preliminary experimental results demonstrate that the proposed expression modeling and recognition-by-synthesis schemes improve the 3D matching accuracy.

Face recognition systems generally require location of face landmarks (often eyes) as an essential preprocessing step. Eye location estimates can be assessed in absolute terms (e.g., proximity to known eye location) and also in application-specific terms (e.g., performance of a system that employs the location). This paper assesses an automatic commercial eye-finding system in absolute and application-specific terms, using two face recognition systems and a database of thousands of images. Experimental results highlight the sensitivity of both systems to accurate eye location estimation as well as quality face lighting and time lapse.

This is the first study to compare the PCA and ICP approaches to 3D face recognition, and to propose a local region approach coping with expression variation in 3D face recognition. A new algorithm for 3D face recognition is proposed for handling expression variation. It uses a surface registration-based technique for 3D face recognition. The proposed method uses a fully automatic approach to use to initialize the 3D matching. Results are presented for gallery and probe datasets of 355 subjects imaged in 3D, with significant time lapse between gallery and probe images of a given subject yielding 3,205 3D models. We find that an ICP-based method performs better than a PCA-based method. The evaluation results show that our proposed new algorithm substantially improves performance in the case of varying facial expression. We also examined subject factors in the proposed method on 3D face models by age and gender.

Spectral reflectance properties of local facial regions have been shown to be useful discriminants for face recognition. To evaluate the performance of spectral signature methods versus purely spatial methods, face recognition tests are conducted using the eigenface method for single-band images extracted from the hyperspectral images. This is the first such comparison based on the same dataset. Selected sets of bands as well as PCA transformed bands are also used for face recognition evaluation with individual band processed separately. A new spectral eigenface method which preserves both spatial and spectral features is proposed. All algorithms based on spectral and/or spatial features are evaluated under the same framework and are compared in terms of accuracy and computational efficiency.

This work addresses the problem of face modelling when the available images are taken from different viewpoints other than frontal. Examples of this situation can be found in surveillance systems, car driver images or whenever it is not possible to place a camera in front of the subjects' face. A number of solutions have been proposed in the literature to deal with these images, but in general they need special databases with explicit 3D data or multiple views to allow for its estimation. As opposed to that, we propose a method able to model faces from a single two-dimensional image captured from a generic viewpoint, as long as there are only bounded occlusions (i.e. only those produced by the nose). Taking advantage of the fact that most facial features lie approximately on the same plane, projective geometry is used to link different views. An Active Shape Model constructed with frontal view images can then be directly applied to the segmentation of pictures taken from other viewpoints.

This paper presents a novel face detection method, which integrates the discriminating feature analysis of the input image, the statistical modeling of face and nonface classes, and the Bayes classifier for multiple frontal face detection. First, feature analysis derives a discriminating feature vector by combining the input image, its 1-D Haar wavelet representation, and its amplitude projections. Second, statistical modeling estimates the conditional probability density functions, or PDFs, of the face and nonface classes, respectively. Finally, the Bayes classifier applies the estimated conditional PDFs to detect multiple frontal faces in an image. Experimental results using 853 images (containing a total of 970 faces) from diverse image sources show the feasibility of the proposed face detection method.

Human Identification based on facial images is one of the most challenging tasks in comparison to identification based on other biometric features such as fingerprints, palm prints or iris. Facial recognition is the most natural and suitable method of identification for security related applications. This paper is concerned with wavelet-based schemes for efficient face verification suitable for implementation on devices that are constrained in memory size and computational power such as PDA’s and smartcards. Beside minimal storage requirements we should apply as few as possible pre-processing procedures that are often needed to deal with variation in recoding conditions. We propose the LL-coefficients wavelet-transformed face images as the feature vectors for face verification, and compare its performance of PCA applied in the LL-subband at levels 3,4 and 5. We shall also compare the performance of various versions of our scheme, with those of well-established PCA face verification schemes on the BANCA database as well as the ORL database. In many cases, the wavelet-only feature vector scheme has the best performance while maintaining efficacy and requiring minimal pre-processing steps. The significance of these results is their efficiency and suitability for platforms of constrained computational power and storage capacity (e.g. smartcards). Moreover, working at or beyond level 3 LL-subband results in robustness against high rate compression and noise interference.

Reliable verification and identification can be achieved by fusing hard and soft information from multiple classifiers. Correlation filter based classifiers have shown good performance in biometric verification applications. In this paper, we develop a method of fusing soft information from multiple correlation filters. Usually, correlation filters are designed to produce a strong peak in the correlation filter output for authentics whereas no such peak should be produced for impostors. Traditionally, the peak-to-sidelobe-ratio (PSR) has been used to characterize the strength of the peak and thresholds are set on the PSR in order to determine whether the test image is an authentic or an impostor. In this paper, we propose to fuse multiple correlation output planes, by appending them for classification by a Support Vector Machine (SVM), to improve the performance over traditional PSR based classification. Multiple Unconstrained Optimal Tradeoff Synthetic Discriminant Function (UOTSDF) filters having varying degrees of discrimination and distortion tolerance are employed here to create a feature vector for classification by a SVM, and this idea is evaluated on the plastic distortion set of the NIST 24 fingerprint database. Results on this database provide an Equal Error Rate (EER) of 1.36% when we fuse correlation planes, in comparison to an average EER of 3.24% using the traditional PSR based classification from a filter, and 2.4% EER on fusion of PSR scores from the same filters using SVM, which demonstrates the advantages of fusing the correlation output planes over the fusion of just the peak-to-sidelobe-ratios (PSRs).

Multibiometric systems utilize the evidence presented by multiple biometric sources (e.g., face and fingerprint, multiple fingers of a user, multiple matchers, etc.) in order to determine or verify the identity of an individual. Information from multiple sources can be consolidated in several distinct levels, including the feature extraction level, match score level and decision level. While fusion at the match score and decision levels have been extensively studied in the literature, fusion at the feature level is a relatively understudied problem. In this paper we discuss fusion at the feature level in 3 different scenarios: (i) fusion of PCA and LDA coefficients of face; (ii) fusion of LDA coefficients corresponding to the R,G,B channels of a face image; (iii) fusion of face and hand modalities. Preliminary results are encouraging and help in highlighting the pros and cons of performing fusion at this level. The primary motivation of this work is to demonstrate the viability of such a fusion and to underscore the importance of pursuing further research in this direction.

In this work we apply a computationally efficient, closed form design of a jointly optimized filter bank of correlation filter classifiers for biometric verification with the use of multiple biometrics from individuals. Advanced correlation filters have been used successfully for biometric classification, and have shown robustness in verifying faces, palmprints and fingerprints. In this study we address the issues of performing robust biometric verification when multiple biometrics from the same person are available at the moment of authentication; we implement biometric fusion by using a filter bank of correlation filter classifiers which are jointly optimized with each biometric, instead of designing separate independent correlation filter classifiers for each biometric and then fuse the resulting match scores. We present results using fingerprint and palmprint images from a data set of 40 people, showing a considerable advantage in verification performance producing a large margin of separation between the impostor and authentic match scores. The method proposed in this paper is a robust and secure method for authenticating an individual.

Fingerprint systems are the most widespread form of biometric authentication. Used in locations such as airports and in PDA's and laptops, fingerprint readers are becoming more common in everyday use. As they become more familiar, the security weaknesses of fingerprint sensors are becoming better known. Numerous websites now exist describing in detail how to create a fake fingerprint usable for spoofing a biometric system from both a cooperative user and from latent prints. While many commercial fingerprint readers claim to have some degree of spoof detection incorporated, they are still generally susceptible to spoof attempts using various artificial fingerprint samples made from gelatin or silicone or other materials and methods commonly available on the web. This paper describes a multispectral sensor that has been developed to collect data for spoof detection. The sensor has been designed to work in conjunction with a conventional optical fingerprint reader such that all images are collected during a single placement of the finger on the sensor. The multispectral imaging device captures sub-surface information about the finger that makes it very difficult to spoof. Four attributes of the finger that are collected with the multispectral imager will be described and demonstrated in this paper: spectral qualities of live skin, chromatic texture of skin, sub-surface image of live skin, and blanching on contact. Each of these attributes is well suited to discriminating against particular kinds of spoofing samples. A series of experiments was conducted to demonstrate the capabilities of the individual attributes as well as the collective spoof detection performance.

Very little work has been done in determining the number of users needed to establish confidence intervals for an error rate of a biometric authentication system. The independence assumption between multiple acquisitions of an individual is too restrictive and is generally not valid. We relax this assumption and present a semi-parametric approach for estimating the within-user correlation using multivariate Gaussian copula models. We describe how to obtain confidence bands for the ROC and present the minimum requirements on the number of users needed to achieve a desired width for the ROC confidence band. Rules of thumb such as the Rule of 3 and the Rule of 30 grossly underestimate the number of users required. The underestimation becomes more severe when the correlation between any two acquisitions increases.

This paper reports on some of the challenges associated with setting up and conducting a full operational benchmark of a palm and fingerprint identification system, based on PITO's own recent experience in this field. The tests described were undertaken as part of the overall evaluation of suppliers tendering for a multi million pound contract to deliver a new national automated fingerprint service for the UK (known as IDENT1), as a successor to the existing systems, both in England and Wales, and in Scotland. The emphasis throughout was on 'operationally' representative testing and it was this that determined the design and scale of the tests, which PITO believes are the largest such tests of a national AFIS ever undertaken. The knowledge gained from performing these benchmark tests has provided PITO with extremely valuable experience in both the theoretical and practical issues surrounding the design and conduct of operational tests on large scale identification systems, and it is these issues that are discussed in this paper.

We address rejection-classification problems, which have been ignored in most prior work. For such a system, a high classification rate and a low false alarm rate are simultaneously desired. We first propose a one-class support vector representation machine (SVRM). The SVRM achieves a high test set detection rate by requiring a high training set detection rate; the SVRM reduces the false alarm rate by minimizing the upper bound of the decision region. The SVRM is then extended to a new support vector representation and discrimination machine (SVRDM) classifier to address multiple-class cases. The theoretical basis for our new SVRDM as best at rejection of non-objects (imposters in face recognition) is provided, as are new σ parameter selection methods. Test results on face recognition and verification with both pose and illumination variations are presented.

In this article, we evaluate the effectiveness of a pre-classification scheme for the fast retrieval of faces in a large image database. The studied approach is based on a partitioning of the face space through a clustering of face images. Mainly two issues are discussed. How to perform clustering with a non-trivial probabilistic measure of similarity between faces? How to assign face images to all clusters probabilistically to form a robust characterization vector? It is shown experimentally on the FERET face database that, with this simple approach, the cost of a search can be reduced by a factor 6 or 7 with no significant degradation of the performance.

Biometric identification has emerged as a reliable means of controlling access to both physical and virtual spaces. Fingerprints, face and voice biometrics are being increasingly used as alternatives to passwords, PINs and visual verification. In spite of the rapid proliferation of large-scale databases, the research has thus far been focused only on accuracy within small databases. In larger applications, response time and retrieval efficiency also become important in addition to accuracy. Unlike structured information such as text or numeric data that can be sorted, biometric data does not have any natural sorting order. Therefore indexing and binning of biometric databases represents a challenging problem. We present results using parallel combination of multiple biometrics to bin the database. Using hand geometry and signature features we show that the search space can be reduced to just 5% of the entire database.

Fingerprint matching is a key issue in research of an automatic fingerprint identification system. On the basis of triangulation in computational geometry, we develop a kind of method for fingerprint matching based on Delaunay Triangulation net in this paper. Through carrying on Delaunay Triangulation to the topological structure of fingerprint minutiae, minutiae with closer distance link to each other on the space according to the Delaunay criterion and form the Delaunay Triangulation net. Then look for some reference minutiae pairs correctly from the net. According to the reference minutiae pairs, match fingerprint on point pattern. The experimental results on FVC2000 indicate the validity of algorithm.

We present results of the largest experimental investigation of 3D ear biometrics to date. ICP-based approaches are carefully explored, and the best rank one recognition rate achieves 98.8%. Only 4 cases out of 302 are incorrectly matched. This result is encouraging in that it suggests the uniqueness of the human ear and its potential applicability as a biometric.

The human dental atlas contains a detailed description of each tooth in the mouth and their relative positions. Registering a dental radiograph to the dental atlas reveals the position and index of each tooth in the radiograph. This helps in establishing the correspondence of teeth when matching two radiographs for human identification. We propose a hidden Markov model (HMM) as an underlying representation of the dental atlas. In our model, the states representing the available teeth have discrete observations, namely the class of each tooth, and the states representing the missing teeth have continuous observations-the distance between neighboring teeth. To classify the teeth, three support vector machines (SVMs) using different feature sets are combined using the average fusion method. Experimental results show that this registration algorithm is promising.

In this paper we present the work developed on off-line signature verification as a continuation of a previous work using Left-to-Right Hidden Markov Models (LR-HMM) in order to extend those models to the field of static or off-line signature processing using results provided by image connectivity analysis. The chain encoding of perimeter points for each blob obtained by this analysis is an ordered set of points in the space, clockwise around the perimeter of the blob. Two models are generated depending on the way the blobs obtained from the connectivity analysis are ordered. In the first one, blobs are ordered according to their perimeter length. In the second proposal, blobs are ordered in their natural reading order, i.e. from the top to the bottom and left to right. Finally, two LR-HMM models are trained using the (x,y) coordinates of the chain codes obtained by the two mentioned techniques and a set of geometrical local features obtained from them such as polar coordinates referred to the center of ink, local radii, segment lengths and local tangent angle. Verification results of the two techniques are compared over a biometrical database containing skilled forgeries.

Authentication systems based on biometrics are becoming popular. One of them, authentication by handwritten signature, is the most accepted due to the fact that it is easy to use. In this paper we present a comparison of methods for efficient point selection for on-line signature authentication purpose. 5 methods are presented based on different selection criteria. The efficiency of these methods is measured by the recognition rate. The comparison method used is Dynamic Time Warping.

Eye movements contain a lot of information about human being. The way the eyes are moving is very complicated and eye movement patterns has been subject of studies for over 100 years. However, surprisingly, eye movement based identification is a quite new idea presented for the first time during the Biometrics'2003 Conference in London [17]. The method has several significant advantages: compiles behavioral and physiological properties of human body, it is difficult to forge and it is affordable-with a number of ready-to-use eye registering devices (so called eye trackers). The paper introduces the methodology and presents results of the first eye movement based authorization tests.

A pilot Personal Identification Credential System (PICS) has been developed and fielded. The PICS is a wireless biometric credential that interfaces with access control systems. The PICS consists of individual handheld Personal Identification Credentials (PIC), a PICS Reader located at a facility entry control point that interfaces with the facility entry control system, and a PICS Enrollment Station. In operation, an individual approaching a facility entry point in a vehicle picks up the PIC handheld unit and places a finger on its sensor. The PIC then authenticates the user and from within the vehicle initiates two-way, secure RF communication with the PICS Reader as the vehicle approaches the gate. The PICS Reader then verifies that the individual is authorized for admittance and notifies the facility gate entry control system, which informs the sentry that the request for access was successful or unsuccessful. If the request for access is unsuccessful, the gate entry control system automatically will close the gate. This sequence of events takes place while the car is moving through a normally open entry lane. The PIC is a small, handheld device which contains the biometric sensor (fingerprint sensor), wireless RF transceiver, processor, encryption and battery. The PIC may be used while traveling in a vehicle or may be used while on foot for access to a PICS controlled man gate or secure area access portal. The PIC is small enough to be carried in a shirt pocket, or it can be left in the user's vehicle. The PIC battery will power the PIC for months and is rechargeable. Up to 10 fingers may be stored in the PIC.

This paper demonstrates the feasibility of a new method of hand-geometry recognition based on parameters derived from the contour of the hand. The contour is completely determined by the black-and-white image of the hand and can be derived from it by means of simple image-processing techniques. It can be modelled by parameters, or features, that can capture more details of the shape of the hand than what is possible with the standard geometrical features used in hand-geometry recognition. The set of features considered in this paper consists of the spatial coordinates of certain landmarks on the contour. The feature set and the recognition method used are discussed in detail. The usefulness of the proposed feature set is evaluated experimentally in a verification context. The verification performance obtained with contour-based features is compared with the verification performance of other methods described in the literature.

In this paper, a novel on-line signature verification approach is proposed for personal authentication in video surveillance systems. As we know, digit password-based authentication is the most popular manner in many network-based applications. However, if the passwords were leaked, the monitoring data are easily falsified. Biometric-based authentication using signature features is a natural and friendly approach to remedy this problem. In this study, a signature-based authentication is proposed to identify the individuals by using the template matching strategy. Some experimental results were conducted to show the effectiveness of our proposed methods.

Fingerprint recognition is implemented on Ti DSP chip Ti 5402. The algorithm is prompt and practicable. Enhance part adopts a fast orientation filter to improve inputting image quality combining fast orientation interpolation method. On the other hand, the orientation field is used in fingerprint image segment. Matching part uses point match method and combines some post-processing in order to get good match result. One fusion way is introduced in enrollment. In DSP implement phase, the system frame adopts one kind of dispatched system structure aiming at the 5402 features. Otherwise the algorithm of fingerprint recognition is modified and optimized based on DSP instructions. Experiment result indicates that system performance is good.

In this paper a novel approach for face authentication is proposed, based on the Hidden Markov Model (HMM) tool. While this technique has been largely and successfully employed in face recognition systems, its use in the authentication context has poorly been investigated. The method proposed in this paper extracts from the image a sequence of partially overlapped images, from which different kinds of simple and quickly computable features are extracted. The face template is obtained by modelling the sequence with a continuous Gaussian Hidden Markov Model. Given an unknown subject, the authentication phase is carried out by thresholding the likelihood of the given face with respect to the HMM template. The proposed approach has been thoroughly tested on the ORL database, also applying different parameters' configurations. A comparison with two other state-of-the-art approaches is also reported. The results obtained are really promising, showing the wide applicability of the Hidden Markov Models methodology.

Authentication of individuals is rapidly becoming an important issue. The authors previously proposed a Pen-input online signature verification algorithm. The algorithm considers a writer’s signature as a trajectory of pen position, pen pressure, pen azimuth, and pen altitude that evolve over time, so that it is dynamic and biometric. Many algorithms have been proposed and reported to achieve accuracy for on-line signature verification, but setting the threshold value for these algorithms is a problem. In this paper, we introduce a user-generic model generated by AdaBoost, which resolves this problem. When user- specific models (one model for each user) are used for signature verification problems, we need to generate the models using only genuine signatures. Forged signatures are not available because imposters do not give forged signatures for training in advance. However, we can make use of another's forged signature in addition to the genuine signatures for learning by introducing a user generic model. And Adaboost is a well-known classification algorithm, making final decisions depending on the sign of the output value. Therefore, it is not necessary to set the threshold value. A preliminary experiment is performed on a database consisting of data from 50 individuals. This set consists of western-alphabet-based signatures provide by a European research group. In this experiment, our algorithm gives an FRR of 1.88% and an FAR of 1.60%. Since no fine-tuning was done, this preliminary result looks very promising.

We present a MIFE (Most Information Feature Extraction) approach, which extract as abundant as possible information for the face classification task. In the MIFE approach, a facial image is separated into sub-regions and each sub-region makes individual’s contribution for performing face recognition. Specifically, each sub-region is subjected to a sub-region based adaptive gamma (SadaGamma) correction or sub-region based histogram equalization (SHE) in order to account for different illuminations and expressions. Experiment results show that the proposed SadaGamma/SHE correction approach provides an efficient delighting solution for face recognition. MIFE and SadaGamma/SHE correction together achieves lower error ratio in face recognition under different illumination and expression.

This paper presents an experimental study of the implementation of a face recognition system in embedded systems. To investigate the feasibility and practicality of real time face recognition on such systems, a door access control system based on face recognition is built. Due to the limited computation power of embedded device, a semi-automatic scheme for face detection and eye location is proposed to solve these computationally hard problems. It is found that to achieve real time performance, optimization of the core face recognition module is needed. As a result, extensive profiling is done to pinpoint the execution hotspots in the system and optimization are carried out. After careful precision analysis, all slow floating point calculations are replaced with their fixed-point versions. Experimental results show that real time performance can be achieved without significant loss in recognition accuracy.

This paper describes an analytical determination of the chemical composition and variation of the chemicals in body odor among individuals through headspace solid phase micro-extraction gas chromatography/mass spectrometry (SPME-GC/MS) as a possible biometric measurement. Biometrics is the technique of measuring a physical characteristic or personal trait of an individual and comparing that characteristic to a database for the purpose of identification. It is known that the human body produces odor, and that this odor is distinguishable through the use of canines. The chemical composition of an individual’s body odor is considered a physical characteristic. This paper discusses and compares the odor profiles produced from the hands as well as the armpit regions of different individuals. Chromatographic distinction among the individuals studied is shown through a combination of the relative ratios of common compounds and the presence of differing compounds.

In the modern world, the ever-growing need to ensure a system's security has spurred the growth of the newly emerging technology of biometric identification. The present paper introduces a novel set of facial biometrics based on quantified facial asymmetry measures in the frequency domain. In particular, we show that these biometrics work well for face images showing expression variations and have the potential to do so in presence of illumination variations as well. A comparison of the recognition rates with those obtained from spatial domain asymmetry measures based on raw intensity values suggests that the frequency domain representation is more robust to intra-personal distortions and is a novel approach for performing biometric identification. In addition, some feature analysis based on statistical methods comparing the asymmetry measures across different individuals and across different expressions is presented.

Human face recognition is currently a very active research area with focus on ways to perform robust biometric identification. Many face recognition algorithms have been proposed, among the different approaches, frequency domain methods, like advanced correlation filters have been shown to exhibit better tolerance to illumination variations than traditional methods. In this paper, we propose a new frequency domain face recognition method which combines the Gabor transforms and a quaternion correlation filter for face recognition when the illumination conditions are changed. The Gabor transform provides optimally localized spatial and frequency domain representation of the original face images, and the quaternion correlation filters can jointly process multi-channel subbands for more robust face recognition. The numerical experiments show that the proposed method outperforms the previously compared advanced correlation filter methods.

There has been a strong trend lately in face processing research away from geometric models towards appearance models. Appearance-based methods employ dimensionality reduction to represent faces more compactly in a low-dimensional subspace which is found by optimizing certain criteria. The most popular appearance-based method is the method of eigenfaces that uses Principal Component Analysis (PCA) to represent faces in a low-dimensional subspace spanned by the eigenvectors of the covariance matrix of the data corresponding to the largest eigenvalues (i.e., directions of maximum variance). Recently, Random Projection (RP) has emerged as a powerful method for dimensionality reduction. It represents a computationally simple and efficient method that preserves the structure of the data without introducing significant distortion. Despite its simplicity, RP has promising theoretical properties that make it an attractive tool for dimensionality reduction. Our focus in this paper is on investigating the feasibility of RP for face recognition. In this context, we have performed a large number of experiments using three popular face databases and comparisons using PCA. Our experimental results illustrate that although RP represents faces in a random, low-dimensional subspace, its overall performance is comparable to that of PCA while having lower computational requirements and being data independent.

This paper describes an embedded multi-user login system based on fingerprint recognition. The system, built using the Sitsang development board and embedded Linux, implements all fingerprint acquisition, preprocessing, minutia extraction, match, identification, user registration, and template encryption on the board. By careful analysis of the accuracy requirement as well as the arithmetic precision to be used, we optimized the algorithms so that the whole system can work in real-time in the embedded environment based on Intel(R) PXA255 processor. The fingerprint verification, which is the core part of the system, is fully tested on a fingerprint database consists of 1149 fingerprint images. The result shows that we can achieve an accuracy of more than 95%. Field testing of 20 registered users has further proved the reliability of our system. The core part of our system, then embedded fingerprint authentication, can also be applied in many different embedded applications concerning security problems.

In automatic fingerprint identification system, incomplete or rigid template may lead to false rejection and false matching. So, how to improve quality of the template, which is called template improvement, is important to automatic fingerprint identify system. In this paper, we propose a template improve algorithm. Based on the case-based method of machine learning and probability theory, we improve the template by deleting pseudo minutia, restoring lost genuine minutia and updating the information of minutia such as positions and directions. And special fingerprint image database is built for this work. Experimental results on this database indicate that our method is effective and quality of fingerprint template is improved evidently. Accordingly, performance of fingerprint matching is also improved stably along with the increase of using time.

In an automatic fingerprint identification system, image enhancement step and match component can benefit from the estimation of fingerprint image quality. In this paper, a criterion for fingerprint image quality is established based on the gradient of fingerprint image. We test the quality measuring method using the proposed criterion on the fingerprint image database, DB3 in FVC2002. From the experimental results , it can be shown that the method is available for estimating the fingerprint image quality.

A facial pose estimation system is presented that functions with illumination variations present. Pose estimation is a useful first stage in a face recognition system. A separate minimum noise and correlation energy (MINACE) filter is synthesized for each pose. To select the MINACE parameter c for the filter for a pose, a training set of illumination differences of several faces at that pose, and a validation set of other poses (illumination differences of several faces at a few other poses) is used in the automated filter-synthesis step. However, the filter for each pose is a combination of faces at only that pose. The pose estimation system is evaluated using images from the CMU Pose, Illumination and Expression (PIE) database. The classification performance (PC) scores are presented for several pose estimation tests. The pose estimate will be used for a subsequent image transformation of a test face to a reference pose for face identification.