The long research studies on the use of computer technology to decipher and determine the individual characteristics of handwriting have led to progress in some trends and set limits in others. At present the products for handwriting recognition and its conversion into printed text are widespread and easily accessible. There is a large number of programs which cope with the task to verify the author in a satisfactory way. It is possible to introduce and make use of enough handwriting features which are calculable and may serve as reliable criteria when the author of a particular writing is automatically confirmed or rejected.
This paper presents a framework for determining the direction of human gaze with an active multi-camera system. A fixed camera is employed in order to estimate the position of the human face and its features, like the eyes. By means of the Supervised Descent Method (SDM) for minimizing a Non-linear Least Squares (NLS) function we can compute correctly the position of the two eyes using 6 landmarks for each of them and the pose of the head. Then an active pan-tilt camera is oriented to one of the users eyes. This way a high precision gaze direction determination is accomplished.
The paper presents an approach for recognition of 3D objects using a database (DB) of precedents. Each object of interest for recognition is presented in the DB through a sufficient number of 2D projections (images), each from a different view point. If it is available a CBIR method to access the DB that is to be fast enough and noise resistant, the number of necessary view positions for each 3D object can be substantially reduced, for example, to several tens or a few hundreds of images. The authors have already applied successfully this appearance-based approach two times: i) for recognition of palm signs from a sign language alphabet and ii) for human face recognition. The recent advance in 3D scanning technologies allow to fresh up the training phase of the proposed method, i.e. the DB gathering of the necessary appearance of precedents for 3D objects, now more accurately and simply. At the same time, the true recognition remains based on images from conventional 2D cameras. This study aims to experiment the mentioned approach for the case of human ears recognition, which, according to our research is of interest to the guild on Biometrics, in the country, in Europe and worldwide.
It has been well known that there is a correlation between facial expression and person’s internal emotional state. In this paper we use an approach to distinguish between neutral and some other expression: based on the displacement of important facial points (coordinates of edges of the mouth, eyes, eyebrows, etc.). Further the feature vectors are formed by concatenating the landmarks data from Supervised Descent Method, applying PCA and use these data as an input to Support Vector Machine (SVM) classifier. The experimental results show improvement of the recognition rate in comparison to some state-of-the-art facial expression recognition techniques.
In this paper, a new method to represent human ear for biometrics purposes is introduced. Even if ear has a uniform distribution of color, human external ear characteristics are considered unique to each individual and permanent during the lifetime of an adult. For these reasons ear biometrics approaches are relying on morphological ear properties. Even if ear biometrics is a young topic a variety of approaches have been proposed to characterize the ear geometry and topology. Moreover, note that the ear morphology is the biggest human head concavity, and that its convex hull complement is mainly convex. In this connection, the matching potential for ear discrimination can be effectively exploited through an Extended Gaussian Image (EGI) representation. The original EGI representation and its correspondent concrete data-structure are here applied to ear description and discussed for human authentication and identification purposes.
The continuously increasing art market activity and international art transactions lead the market for stolen and fraudulent art to extreme levels. According to US officials, art crime is the third-highest grossing criminal enterprise worldwide. As a result, art forensics is a rising research field dealing with the identification of stolen or looted art and their collection and repatriation. Photographs of artwork provide, in several cases, the only way to locate stolen and looted items. However, it is quite common these items to be damaged as a result of excavation and illegal movement. Digital processing of photographs of damaged artwork is therefore of high importance in art forensics. This processing emphasizes on “object restoration” and although techniques from the field of image restoration can be applied it is of high importance to take into account the semantics of the artwork scene and especially the structure of objects appeared therein. In this paper, we assess the application of face image restoration techniques, applied on damaged faces appearing in Byzantine icons, in an attempt to identify the actual icons. Several biometric measurements and facial features along with a set of rules related to the design of Byzantine faces are utilized for this purpose. Preliminary investigation, applied on 25 icons, shows promising results.
Gait exhibits several advantages with respect to other biometrics features: acquisition can be performed through cheap technology, at a distance and without people collaboration. In this paper we perform gait analysis using skeletal data provided by the Microsoft Kinect sensor. We defined a rich set of physical and behavioral features aiming at identifying the more relevant parameters for gait description. Using SVM we showed that a limited set of behavioral features related to the movements of head, elbows and knees is a very effective tool for gait characterization and people recognition. In particular, our experimental results shows that it is possible to achieve 96% classification accuracy when discriminating a group of 20 people.
In this paper a method for on-line signature verification is presented. The proposed approach consists of the following consecutive steps: feature selection and classification. Experiments are carried out on SUsig database [5] of genuine and forgery signatures of 89 users. The results obtained by applying two different types of classifiers (NN and k-nearest neighbours) are compared. For each user, several NN and kNN models are evaluated by 10-fold cross validation and LOOCV respectively. The “optimal” models are found together with their parameters: number of hidden neurons for NN, type of signature forgeries for training, input features and value of k. The influence of the signature forgery type (random and skilled) over the feature selection and verification is investigated as well.
This paper reviews the contemporary (face, gait, and fusion) computational approaches for automatic human identification at a distance. For remote identification, there may exist large intra-class variations that can affect the performance of face/gait systems substantially. First, we review the face recognition algorithms in light of factors, such as illumination, resolution, blur, occlusion, and pose. Then we introduce several popular gait feature templates, and the algorithms against factors such as shoe, carrying condition, camera view, walking surface, elapsed time, and clothing. The motivation of fusing face and gait, is that, gait is less sensitive to the factors that may affect face (e.g., low resolution, illumination, facial occlusion, etc.), while face is robust to the factors that may affect gait (walking surface, clothing, etc.). We review several most recent face and gait fusion methods with different strategies, and the significant performance gains suggest these two modality are complementary for human identification at a distance.
We present a computational framework, which combines depth and colour (texture) modalities for 3D scene reconstruction. The scene depth is captured by a low-power photon mixture device (PMD) employing the time-of-flight principle while the colour (2D) data is captured by a high-resolution RGB sensor. Such 3D capture setting is instrumental in 3D face recognition tasks and more specifically in depth-guided image segmentation, 3D face reconstruction, pose modification and normalization, which are important pre-processing steps prior to feature extraction and recognition. The two captured modalities come with different spatial resolution and need to be aligned and fused so to form what is known as view-plus-depth or RGB-Z 3D scene representation. We discuss specifically the low-power operation mode of the system, where the depth data appears very noisy and needs to be effectively denoised before fusing with colour data. We propose using a modification of the non-local means (NLM) denoising approach, which in our framework operates on complex-valued data thus providing certain robustness against low-light capture conditions and adaptivity to the scene content. Further in our approach, we implement a bilateral filter on the range point-cloud data, ensuring very good starting point for the data fusion step. The latter is based on the iterative Richardson method, which is applied for efficient non-uniform to uniform resampling of the depth data using structural information from the colour data. We demonstrate a real-time implementation of the framework based on GPU, which yields a high-quality 3D scene reconstruction suitable for face normalization and recognition.
The iris has been proposed as a highly reliable and stable biometric identifier for person authentication/recognition about two decades ago. Since then, most work in the field has been focused on segmentation and matching algorithms able to work on pictures of whole face or eye region typically captured at close distance, while preserving recognition accuracy. In this paper we present an iris matching algorithm based on spatial histograms that, while showing good recognition performance on some of the most referenced public iris dataset, is also able to perform a one-to-one comparison in a small amount of time thanks to its low computing load, thus resulting particularly suited to iris recognition applications on mobile devices.
In this paper, we present a new iris detection method based on the use of watershed segmentation. The watershed transform is used for both pupil and iris detection, in combination with image quantization, aimed at reducing the number of gray levels, and image thresholding, aimed at obtaining a tentative discrimination between foreground and background. The method has been tested on the CASIA-Iris-Interval Image database.
Soft biometrics continues to attract research interest. Traditional body and face soft biometrics have been the main research focus and have been proven, by many researchers, to be usable for identification and retrieval. Also, soft biometrics have been shown to provide several advantages over classic biometrics, such as invariance to illumination and contrast. Other than body and face, little attention has focussed on semantic descriptions of an individual, including clothing attributes. Research has yet to concern clothing characteristics as a major or complementary set of biometric traits. In this paper, we analyse the reliability and significance of clothing information for retrieval purposes. We investigate and rate the viability of semantic clothing descriptions to retrieve a subject correctly, given a verbal description of their clothing.
We investigate the possibility of using pupil size as a discriminating feature for eye-based soft biometrics. In experiments carried out in different sessions in two consecutive years, 25 subjects were asked to simply watch the center of a plus sign displayed in the middle of a blank screen. Four primary attributes were exploited, namely left and right pupil sizes and ratio and difference of left and right pupil sizes. Fifteen descriptive statistics were used for each primary attribute, plus two further measures, which produced a total of 62 features. Bayes, Neural Network, Support Vector Machine and Random Forest classifiers were employed to analyze both all the features and selected subsets. The Identification task showed higher classification accuracies (0.6194 ÷ 0.7187) with the selected features, while the Verification task exhibited almost comparable performances (~ 0.97) in the two cases for accuracy, and an increase in sensitivity and a decrease in specificity with the selected features.
Fusion biometric modal contributes in two aspects. It can not only improve the biometric recognition accuracy, but also gives a comparatively safe strategy, since it is difficult for intruders to achieve multi-biometric information simultaneously, especially the iris information. The contourlet transform is a new two-dimensional extension of the wavelet transform using multiscale and directional filter banks. The contourlet expansion is composed of basis images oriented at various directions in multiple scales, with flexible aspect ratios. In this paper, by using Contourlet transform, we extract the features of retina and iris, and fuse them at feature level and utilize Hamming distance for matching purpose to provide a higher accuracy than unimodal system. The experimental results show that our biometric system based on the integration of retina and iris traits achieve an EER=0.0413%.
The goal of this paper is to review automatic systems for forensic speaker recognition (FSR) based on scientifically approved methods for calculation and interpretation of biometric evidence. The objective of this paper is not to promote one speaker recognition method against another, but is to make available to the biometric research community data-driven methodology combining automatic speaker recognition techniques and a rigorous forensic experimental background. Forensic speaker recognition is the process of determining if a specific individual (suspected speaker) is the source of a questioned speech recording (trace). This paper aims at reviewing forensic automatic speaker recognition (FASR) methods that provide a coherent way of quantifying and presenting recorded speech as biometric evidence, as well as the assessment of its strength (likelihood ratio) in the Bayesian interpretation framework compatible with interpretations in other forensic disciplines. Forensic speaker recognition has proven an effective tool in the fight against crime, yet there is a constant need for more research due to the difficulties involved because of the within-speaker (within-source) variability, between-speakers (between-sources) variability, and differences in recording sessions conditions.
Biometrics has historically found its natural mate in Forensics. The first applications found in the literature and over cited so many times, are related to biometric measurements for the identification of multiple offenders from some of their biometric and anthropometric characteristics (tenprint cards) and individualization of offender from traces found on crime-scenes (e.g. fingermarks, earmarks, bitemarks, DNA). From sir Francis Galton, to the introduction of AFIS systems in the scientific laboratories of police departments, Biometrics and Forensics have been "dating" with alternate results and outcomes. As a matter of facts there are many technologies developed under the "Biometrics umbrella" which may be optimised to better impact several Forensic scenarios and criminal investigations. At the same time, there is an almost endless list of open problems and processes in Forensics which may benefit from the introduction of tailored Biometric technologies. Joining the two disciplines, on a proper scientific ground, may only result in the success for both fields, as well as a tangible benefit for the society. A number of Forensic processes may involve Biometric-related technologies, among them: Evidence evaluation, Forensic investigation, Forensic Intelligence, Surveillance, Forensic ID management and Verification.The COST Action IC1106 funded by the European Commission, is trying to better understand how Biometric and Forensics synergies can be exploited within a pan-European scientific alliance which extends its scope to partners from USA, China and Australia.Several results have been already accomplished pursuing research in this direction. Notably the studies in 2D and 3D face recognition have been gradually applied to the forensic investigation process. In this paper a few solutions will be presented to match 3D face shapes along with some experimental results.
Tracking of human beings represents a hot research topic in the field of video analysis. It is attracting an increasing attention among researchers thanks to its possible application in many challenging tasks. Among these, action recognition, human/human and human/computer interaction require body-part tracking. Most of the existing techniques in literature are model-based approaches, so despite their effectiveness, they are often unfit for the specific requirements of a body-part tracker. In this case it is very hard if not impossible to define a formal model of the target. This paper proposes a multi-anchor tracking system, which works on 8 bits color images and exploits the mutual information to track human body parts (head, hands, …) without performing any foreground/background segmentation. The proposed method has been designed as a component of a more general system aimed at human interaction analysis. It has been tested on a wide set of color video sequences and the very promising results show its high potential.
A new approach to soft biometrics aims to use human labelling as part of the process. This is consistent with analysis of surveillance video where people might be imaged at too low resolution or quality for conventional biometrics to be deployed. In this manner, people use anatomical descriptions of subjects to achieve recognition, rather than the usual measurements of personal characteristics used in biometrics. As such the labels need careful consideration in their construction, and should demonstrate correlation consistent with known human physiology. We describe our original process for generating these labels and analyse relationships between them. This gives insight into the perspicacity of using a human labelling system for biometric purposes.
Multifractality Sw , auto- and crosscorrelations h_2 of mediolateral and anteroposterior postural sways of healthy young and elderly subjects is studied by the MDFA, WTMM and MDCA methods. MDFA and MDCA reveal a random walk like time series with h_2∼ 1.7 . Vibrating soles or aging decreases h_2 of the elderly persons. Sw of the mediolateral sways is lesser than the anteroposterior sways. The random permutation of the series vanishes the multifractality which is related with the long-range power-low correlations.
