Face Recognizer Research Articles

Passive Three Dimensional Face Recognition Using Iso-Geodesic Contours and Procrustes Analysis

We introduce a new model for personal recognition based on the 3-D geometry of the face. The model is designed for application scenarios where the acquisition conditions constrain the facial position. The 3-D structure of a facial surface is compactly represented by sets of contours (facial contours) extracted around automatically pinpointed nose tip and inner eye corners. The metric used to decide whether a point on the face belongs to a facial contour is its geodesic distance from a given landmark. Iso-geodesic contours are inherently robust to head pose variations, including in-depth rotations of the face. Since these contours are extracted from rigid parts of the face, the resulting recognition algorithms are insensitive to changes in facial expressions. The facial contours are encoded using innovative pose invariant features, including Procrustean distances defined on pose-invariant curves. The extracted features are combined in a hierarchical manner to create three parallel face recognizers. Inspired by the effectiveness of region ensembles approaches, the three recognizers constructed around the nose tip and inner corners of the eyes are fused both at the feature-level and the match score-level to create a unified face recognition algorithm with boosted performance. The performances of the proposed algorithms are evaluated and compared with other algorithms from the literature on a large public database appropriate for the assumed constrained application scenario.

High and low performers differ in the use of shape information for face recognition

Previous findings demonstrated that increasing facial distinctiveness by means of spatial caricaturing improves face learning and results in modulations of event-related-potential (ERP) components associated with the processing of typical shape information (P200) and with face learning and recognition (N250). The current study investigated performance-based differences in the effects of spatial caricaturing: a modified version of the Bielefelder famous faces test (BFFT) was applied to subdivide a non-clinical group of 28 participants into better and worse face recognizers. Overall, a learning benefit was seen for caricatured compared to veridical faces. In addition, for learned faces we found larger caricaturing effects in response times, inverse efficiency scores as well as in P200 and N250 amplitudes in worse face recognizers, indicating that these individuals profited disproportionately from exaggerated idiosyncratic face shape. During learning and for novel faces at test, better and worse recognizers showed similar caricaturing effects. We suggest that spatial caricaturing helps better and worse face recognizers accessing critical idiosyncratic shape information that supports identity processing and learning of unfamiliar faces. For familiarized faces, better face recognizers might depend less on exaggerated shape and make better use of texture information than worse recognizers. These results shed light on the transition from unfamiliar to familiar face processing and may also be relevant for developing training-programmes for people with difficulties in face recognition.

Determining and locating the closest available resources to mobile collaborators

Nowadays, organizations include a large number of physical resources (e.g., meeting rooms, classrooms, and auditoriums) and computing ones (e.g., scanners, plotters, and handheld devices) distributed among different offices and buildings. Typically, these resources have to be shared among colleagues, because it is impossible for each collaborator to own private instances of all the different resources present in the organization. In this way, resource sharing gives collaborators the opportunity of not just lending their resources to other collaborators but also benefitting from the usage of resources they do not own. However, finding shared resources in a huge organization, without a proper technological support, can be a challenge for a member of staff and obviously really hard or even impossible for an external person. The main contribution of this paper is a service-oriented architecture called Resource Availability Management Services (RAMS), which is intended to facilitate the development of groupware applications that manage the availability and suitability of human, physical and computing resources. In the case of physical resources, the RAMS architecture allows determining the available resources closest to the requesting collaborator that satisfy his/her requirements and provides information about their physical location and the shortest path to reach them. To accomplish these goals, the proposed architecture relies on the services provided by three main components: (1) a Human Face Recognizer that allows identifying and locating collaborators in an organization, (2) an Ontology-based Matchmaker, which is able to determine a set of available and accessible resources that can satisfy a collaborator’s request, and (3) a Physical Resource Locator that relies on building topologies and the walking distance method to calculate high precision relative distances between collaborators and resources.

Passive Multimodal 2-D+3-D Face Recognition Using Gabor Features and Landmark Distances

We introduce a novel multimodal framework for face recognition based on local attributes calculated from range and portrait image pairs. Gabor coefficients are computed at automatically detected landmark locations and combined with powerful anthropometric features defined in the form of geodesic and Euclidean distances between pairs of fiducial points. We make the pragmatic assumption that the 2-D and 3-D data is acquired passively (e.g., via stereo ranging) with perfect registration between the portrait data and the range data. Statistical learning approaches are evaluated independently to reduce the dimensionality of the 2-D and 3-D Gabor coefficients and the anthropometric distances. Three parallel face recognizers that result from applying the best performing statistical learning schemes are fused at the match score-level to construct a unified multimodal (2-D+3-D) face recognition system with boosted performance. Performance of the proposed algorithm is evaluated on a large public database of range and portrait image pairs and found to perform quite well.

Real-Time Pose-Invariant Face Recognition Using the Efficient Second-Order Minimization and the Pose Transforming Matrix

We propose a real-time pose-invariant face recognition algorithm from a gallery of frontal images only. (i) We modified the second-order minimization method for the active appearance model (AAM). This allows the AAM to have the ability of correct convergence with little loss of frame rate. (ii) We proposed a pose transforming matrix that can eliminate warping artifacts of the warped face image from AAM fitting. This makes it possible to train a neural network as the face recognizer with one frontal face image of each person in the gallery set. (iii) We propose a simple method for pose recognition by using neural networks to select the proper pose transforming matrix. The proposed algorithm was evaluated on a set of 2000 facial images of 10 people (200 images for each person obtained at various poses), achieving a great improvement in recognition.

Development and Evaluation of two Memory Architectures for the Hardware Version of CNN Face Recognizer

The design of the proposed Convolutional Neural Network (CNN) architecture for face image recognition takes the constraints on the bandwidth of the communications between memory and processor into the account. The coarse grained parallelism which performed in the bottom layer node's calculations is reduced in consequent manner until the calculation of one simple node in the upper layer is achieved sequentially. Two methods of segmentation are used to buffer the image data required for these parallel to sequential calculations from the image RAM to multi-port RAMs. A comparison between these two methods with respect to the whole number of RAM access required to generate the system recognition code is performed. A speedup of 44 is achieved when the hardware system is implemented with the using of the 1st method of segmentation as compared to a Pentium 4, 2.4 GHz sequential computer software implementation. While a speedup of 88 is achieved when the same hardware system is implemented but with the using of the 2nd segmentation method, compared to the same mentioned sequential computer. Keywords: convolution neural networks, parallel processing, memory architecture.

Representation Plurality and Fusion for 3-D Face Recognition

In this paper, we present an extensive study of 3-D face recognition algorithms and examine the benefits of various score-, rank-, and decision-level fusion rules. We investigate face recognizers from two perspectives: the data representation techniques used and the feature extraction algorithms that match best each representation type. We also consider novel applications of various feature extraction techniques such as discrete Fourier transform, discrete cosine transform, nonnegative matrix factorization, and principal curvature directions to the shape modality. We discuss and compare various classifier combination methods such as fixed rules voting- and rank-based fusion schemes. We also present a dynamic confidence estimation algorithm to boost fusion performance. In identification experiments performed on FRGC v1.0 and FRGC v2.0 face databases, we tried to find the answers to the following questions: 1) the relative importance of the face representation technique vis-à-vis the types of features extracted; 2) the impact of the gallery size; 3) the conditions, under which subspace methods are preferable, and the compression factor; 4) the most advantageous fusion level and fusion methods; 5) the role confidence votes in improving fusion and the style of selecting experts in the fusion; and 6) the consistency of the conclusions across different databases.

View Influence Analysis and Optimization for Multiview Face Recognition

We present a novel method to recognize a multiview face (i.e., to recognize a face under different views) through optimization of multiple single-view face recognitions. Many current face descriptors show quite satisfactory results to recognize identity of people with given limited view (especially for the frontal view), but the full view of the human head has not yet been recognizable with commercially acceptable accuracy. As there are various single-view recognition techniques already developed for very high success rate, for instance, MPEG-7 advanced face recognizer, we propose a new paradigm to facilitate multiview face recognition, not through a multiview face recognizer, but through multiple single-view recognizers. To retrieve faces in any view from a registered descriptor, we need to give corresponding view information to the descriptor. As the descriptor needs to provide any requested view in space, we refer to it as 3D information that it needs to contain. Our analysis in various angled views checks the extent of each view influence and it provides a way to recognize a face through optimized integration of single view descriptors covering the view plane of horizontal rotation from -90° to 90° and vertical rotation from -30° to 30° The resulting face descriptor based on multiple representative views, which is of compact size, shows reasonable face recognition performance on any view. Hence, our face descriptor contains quite enough information of a person's face to help for recognition and eventually for search, retrieval, and browsing of photographs, videos, and 3D-facial model databases.

System-on-programmable-chip implementation for on-line face recognition

In this paper, the design of a parallel architecture for on-line face recognition using weighted modular principal component analysis (WMPCA) and its system-on-programmable-chip (SoPC) implementation are discussed. The WMPCA methodology, proposed by us earlier, is based on the assumption that the rates of variation of the different regions of a face are different due to variations in expression and illumination. Given a database of sample faces for training and a query face for recognizing, the WMPCA methodology involves division of the face into horizontal regions. Each of these regions are analyzed independently by computing the eigenfeatures and comparing the same with the corresponding eigenfeatures of the faces stored in the sample database to calculate the corresponding error. The final decision of the face recognizer is based on the weighted sum of the errors computed from each of the regions. These weights are calculated based on the extent to which the various samples of the subject are spread in the eigenspace. The WMPCA methodology has a better recognition rate compared to the modular PCA approach developed by Rajkiran and Vijayan [Rajkiran, G., Vijayan, K., 2004. An improved face recognition technique based on modular PCA approach. Pattern Recognition Letters, 25(4), 429–436]. The methodology also has a wide scope for parallelism. We present an architecture that exploits this parallelism and implement the same as a system-on-programmable-chip on an ALTERA based field programmable gate array (FPGA) platform. The implementation has achieved a processing speed of about 26 frames per second at an operating frequency of 33.33 MHz.

3D shape-based face representation and feature extraction for face recognition

In this paper, we review and compare 3D face registration and recognition algorithms, which are based solely on 3D shape information and analyze methods based on the fusion of shape features. We have analyzed two different registration algorithms, which produce a dense correspondence between faces. The first algorithm non-linearly warps faces to obtain registration, while the second algorithm allows only rigid transformations. Registration is handled with the use of an average face model, which significantly fastens the registration process. As 3D facial features, we compare the use of 3D point coordinates, surface normals, curvature-based descriptors, 2D depth images, and facial profile curves. Except for surface normals, these feature descriptors are frequently used in state-of-the-art 3D face recognizers. We also perform an in-depth analysis of decision-level fusion techniques such as fixed-rules, voting schemes, rank-based combination rules, and novel serial fusion architectures. The results of the recognition and authentication experiments conducted on the 3D_RMA database indicate that: (i) in terms of face registration method, registration of faces without warping preserves more discriminatory information, (ii) in terms of 3D facial features, surface normals attain the best recognition performance, and (iii) fusion schemes such as product rules, improved consensus voting and proposed serial fusion schemes improve the classification accuracy. Experimental results on the 3D_RMA confirm these findings by obtaining %0.1 misclassification rate in recognition experiments, and %8.06 equal error rate in authentication experiments using surface normal-based features. It is also possible to improve the classification accuracy by %2.38 using fixed fusion rules when moderate-level classifiers are used.

Multihashing, human authentication featuring biometrics data and tokenized random number: A case study FVC2004

Given the recent explosion of interest in human authentication, verification based on tokenized pseudo-random numbers and the user-specific biometric feature has received much attention. These methods have significant functional advantages over solely biometrics, i.e. zero equal error rate. The main drawback of the methods proposed in the literature relies in exhibiting low performance when an “impostor” B steals the pseudo-random numbers of A and he tries to authenticate as A. In this paper, we show that a multimodal fusion, where only one biometric characteristic is combined with the pseudo-random numbers, permits to obtain a zero equal error rate when nobody steals the pseudo-random numbers, and good performance when an “impostor” B steals the pseudo-random numbers of A and he tries to authenticate as A. In this paper, we study the fusion among the score obtained by a Face Recognizer (where the face features are combined with pseudo-random numbers) and the scores of the systems submitted to FVC2004.

Guest editorial: Special issue on computer vision beyond the visible spectrum

It was three years ago that we had our first special issue inComputer Vision Beyond the Visibel Spectrum (CVBVS) inthe Machine Vision and Applications journal [1]. At thattime, we expected that computer vision research wouldgrow significantly in the nonvisible spectrum, particularly inthe biometrics area. This second special issue in the Imageand Vision Computing journal comes to verify ourexpectations. New CVBVS methods and systems in facedetection, tracking, and recognition enable identification ata distance under any lighting conditions. As a result, thisnew technology heralds the migration of human identifi-cation screening from access control to surveillanceapplications. These adavances promise to revolutionizethe security industry. Also, methods and systems for defenseapplications, which is the traditional CVBVS domain, grewmore sophisticated and effective. Target detection andrecognition in Synthetic Aperture Radar (SAR) imagery isat the very core of these advancements.Allthepapersincludedinthisspecialissuewerepresentedin preliminary form at the IEEE Workshop in ComputerVision Beyond the Visible Spectrum, held in Kauai, Hawaii,on December 14, 2001. The selected papers were a smallsubset of the total number of papers presented at theWorkshop and underwent two rounds of further rigorousreview and updates to ensure a high quality outcome.In the first paper Face Detection in the Near-IRSpectrum, J. Dowdall et al. introduce a new face detectionmethod based on the fusion of two near-infrared bands. Theresulting phenomenology is such that it allows reliable facedetection with simple algorithmic means like integralprojection. The authors realized a prototype system basedon their method and performed comparative experimentaltests with a state-of-the-art visible spectrum face detector.The near-infrared detector exhibited superior performanceat various ambient illumination conditions and in thepresence of extreme face rotation. This face detectionsystem in combination with a face recognizer is employedon a trial basis for gate control in a practical application.In the second paper Tracking Human Faces in InfraredVideo, C.K. Eveland et al. describe a novel face detectionand tracking system in thermal infrared. In contrast to thenear-infrared face detector proposed by J. Dowdall et al. inthe first paper, this system is totally passive. The underlyingmethod has two components: first, a face segmentationscheme that is based on a thermal emission model from thehuman skin; and second, a condensation algorithm that usesthe segmentation model for tracking the face over time. Themethod is characterized by mathematical elegance and issupported by promising experimental results.The third paper is on Statistical Hypothesis Pruning forRecognizing Faces from Infrared Images by A. Srivastava.This paper complements the face detection and trackingmethods presented in the first two papers. Recognition isaccomplished by hypothesis pruning using spectraldecomposition of the observed images and their lower-order statistics. This is another elegant method that issupported by comparative experimental results. The authorsused the data sets released by the first two papers, whichreinforces the complimentary nature of the approachestowards an infrared identification system.The fourth paper Genetic Algorithm Based FeatureSelection for Target Detection in SAR Images, by B. Bhanuand Y. Lin introduces a Genetic Algorithm (GA) approachfor target feature selection in SAR imagery. The GAalgorithm is driven by a new fitness function based on

An integrated automatic face detection and recognition system

This paper proposes an integrated system for unconstrained face recognition in complex scenes. The scale and orientation tolerant system comprises a face detector followed by a recognizer. Given a color input image of a person, the face detector encloses the face from the complex scene within a circular boundary, and locates the position of the nose. A radial grid mapping centered on the nose is then performed to extract a feature vector within the boundary. The feature vector is input to a radial basis function neural network classifier for face identification. The proposed face detector achieved an average detection rate of 95.8% while the face recognizer achieved an average recognition rate of 97.5% on a database of 21 persons with variations in scale, orientation, natural illumination and background. The two modules were combined to form an automatic face recognition system that was evaluated in the context of a security system using a video database of 21 users and 10 intruders, acquired in an unconstrained environment. A recognition rate of 93.5% with 0% false acceptance rate was achieved.

FACE RECOGNITION USING SUPPORT VECTOR MACHINES WITH LOCAL CORRELATION KERNELS

This paper presents a real-time face recognition system. For the system to be real time, no external time-consuming feature extraction method is used, rather the gray-level values of the raw pixels that make up the face pattern are fed directly to the recognizer. In order to absorb the resulting high dimensionality of the input space, support vector machines (SVMs), which are known to work well even in high-dimensional space, are used as the face recognizer. Furthermore, a modified form of polynomial kernel (local correlation kernel) is utilized to take account of prior knowledge about facial structures and is used as the alternative feature extractor. Since SVMs were originally developed for two-class classification, their basic scheme is extended for multiface recognition by adopting one-per-class decomposition. In order to make a final classification from several one-per-class SVM outputs, a neural network (NN) is used as the arbitrator. Experiments with ORL database show a recognition rate of 97.9% and speed of 0.22 seconds per face with 40 classes.

Argus: the digital doorman

When you've visited someone in a large apartment or office complex, chances are that a security guard in the lobby granted you access. Perhaps over time, the guard has learned to associate you with the person you plan to visit and immediately notifies that person over the building intercom when you arrive. Argus, named after the vigilant watchman from Greek mythology, is an automated version of such a security guard: a system for automatic visitor identification. We successfully implemented and tested Argus at Just Research. To detect visitors, Argus's digital camera photographs the building entrance at regular intervals, and a motion detection algorithm identifies potential scenes containing visitors. Using a neural-network-based face detector, Argus extracts faces from these images. A memory-based face recognition system examines these faces and attempts to find visually similar matches in its stored database of visitors. An interface agent notifies system users whenever visitors arrive. Users can also provide feedback to Argus in the event of misclassified visitors. Because the face recognizer can learn online, Argus immediately incorporates these corrections into its face recognition data set.

Can self-reported encoding strategy and recognition skill be diagnostic of performance in eyewitness identifications?

The relationship between 3 witness factors and identification accuracy, as well as calibration and diagnosticity of confidence, was investigated. A total of 384 participants in an eyewitness experiment rated their facial recognition skill, general memory skill, and self-reported encoding strategy on a questionnaire presented after the photo-confrontation. Participants who rated themselves to be good face recognizers showed a slightly higher overall accuracy with a more diagnostic confidence-accuracy relation. Participants who reported that they relied on a holistic encoding strategy were associated with more accurate identifications and a stronger confidence-accuracy relation than those who reported an analytic encoding strategy. Degree of self-reported general memory skill was not diagnostic of identification performance.

Face recognition/detection by probabilistic decision-based neural network

This paper proposes a face recognition system, based on probabilistic decision-based neural networks (PDBNN). With technological advance on microelectronic and vision system, high performance automatic techniques on biometric recognition are now becoming economically feasible. Among all the biometric identification methods, face recognition has attracted much attention in recent years because it has potential to be most nonintrusive and user-friendly. The PDBNN face recognition system consists of three modules: First, a face detector finds the location of a human face in an image. Then an eye localizer determines the positions of both eyes in order to generate meaningful feature vectors. The facial region proposed contains eyebrows, eyes, and nose, but excluding mouth (eye-glasses will be allowed). Lastly, the third module is a face recognizer. The PDBNN can be effectively applied to all the three modules. It adopts a hierarchical network structures with nonlinear basis functions and a competitive credit-assignment scheme. The paper demonstrates a successful application of PDBNN to face recognition applications on two public (FERET and ORL) and one in-house (SCR) databases. Regarding the performance, experimental results on three different databases such as recognition accuracies as well as false rejection and false acceptance rates are elaborated. As to the processing speed, the whole recognition process (including PDBNN processing for eye localization, feature extraction, and classification) consumes approximately one second on Sparc10, without using hardware accelerator or co-processor.

Neural information processing in real-world face-recognition applications

One of the challenges of neural information processing is to achieve, at least partially, a similar performance to humans on systems for automated visual face recognition. Commercial applications of new technology don't care about the underlying architecture or paradigms. The architecture simply needs to work all the time and be easy to use. That was the essence of ZN's product design plan when we began developing our ZN-Face access control system. ZN-Face relies on von der Malsburg's graph matching, which is robust enough to deal with the low quality pictures encountered outside the laboratory when developing automated image acquisition from real world scenes. (In this way, of course, the underlying neural system's robustness is essential, because otherwise we could not have fulfilled the works-all-the-time requirement.) At ZN, we developed the complete hardware and software setup for the biometric access control device. We optimized and adapted the algorithms to the specific verification task-that is, "is the person in question identical to the cardholder?"and tested it in the hardware setup, leading to a 99.5% performance verification rate. As with most face recognizers, ZN-Face requires the cooperation of users, who must orient their heads toward the camera during picture acquisition (/spl plusmn/15/spl deg/). Today's algorithms can only partially solve the challenge of generalizing from, for example, a half profile view to a frontal view.

Determination of metallic lead in lead mixtures

Nowadays, organizations include a large number of physical resources (e.g., meeting rooms, classrooms, and auditoriums) and computing ones (e.g., scanners, plotters, and handheld devices) distributed among different offices and buildings. Typically, these resources have to be shared among colleagues, because it is impossible for each collaborator to own private instances of all the different resources present in the organization. In this way, resource sharing gives collaborators the opportunity of not just lending their resources to other collaborators but also benefitting from the usage of resources they do not own. However, finding shared resources in a huge organization, without a proper technological support, can be a challenge for a member of staff and obviously really hard or even impossible for an external person. The main contribution of this paper is a service-oriented architecture called Resource Availability Management Services (RAMS), which is intended to facilitate the development of groupware applications that manage the availability and suitability of human, physical and computing resources. In the case of physical resources, the RAMS architecture allows determining the available resources closest to the requesting collaborator that satisfy his/her requirements and provides information about their physical location and the shortest path to reach them. To accomplish these goals, the proposed architecture relies on the services provided by three main components: (1) a Human Face Recognizer that allows identifying and locating collaborators in an organization, (2) an Ontology-based Matchmaker, which is able to determine a set of available and accessible resources that can satisfy a collaborator’s request, and (3) a Physical Resource Locator that relies on building topologies and the walking distance method to calculate high precision relative distances between collaborators and resources.