The palmprint recognition has become a focus in biological recognition and image processing fields. In this process, the features extraction (with particular attention to palmprint principal line extraction) is especially important. Although a lot of work has been reported, the representation of palmprint is still an open issue. In this paper we propose a simple, efficient, and accurate palmprint principal lines extraction method. Our approach consists of six simple steps: normalization, median filtering, average filters along four prefixed directions, grayscale bottom-hat filtering, combination of bottom-hat filtering, binarization and post processing. The contribution of our work is a new method for palmprint principal lines detection and a new dataset of hand labeled principal lines images (that we use as ground truth in the experiments). Preliminary experimental results showed good performance in terms of accuracy with respect to three methods of the state of the art.
Biomedical and health applications are representing nowadays a very attractive area for electronics devices. The state of the art shows a multitude of solutions based on in body sensors devices to measure and survey human physiological parameters enabling a distance medical monitoring systems and higher out-of-hospital care services for patients. These emergent systems have a lot of challenges related to their power consumption, size and complexity. In this article, we focus on one of the technological locks of the electronic gastric pill, which is the antenna. A miniaturized patch antenna a key element of the receiving system is presented. The antenna size and performances are very important to consider as it have to be integrated on a jacket in order to collect the transmitted information from an electronic gastric pill to trace cartography of the collected data. Different miniaturization techniques of patch antennas are combined with the use high dielectric constant substrate in order to reach a tradeoff between the antenna performances and the size. The designed antenna resonates in the desired ISM frequency band (430 MHz) and was simulated and optimized with the presence of a model of the human body for a more realistic use case. The antenna size reduction achieves 81.56% with good performances in terms of gain (-29.4dB).
With the growing use of biometric authentication systems in the past years, spoof fingerprint detection has become increasingly important. In this work, we implement and evaluate two different feature extraction techniques for software-based fingerprint liveness detection: Convolutional Networks with random weights and Local Binary Patterns. Both techniques were used in conjunction with a Support Vector Machine (SVM) classifier. Dataset Augmentation was used to increase classifier's performance and a variety of preprocessing operations were tested, such as frequency filtering, contrast equalization, and region of interest filtering. The experiments were made on the datasets used in The Liveness Detection Competition of years 2009, 2011 and 2013, which comprise almost 50,000 real and fake fingerprints' images. Our best method achieves an overall rate of 95.2% of correctly classified samples - an improvement of 35% in test error when compared with the best previously published results.
Compressive Sensing (CS) is a newly introduced signal processing technique that enables to recover sparse signals from fewer samples than the Shannon sampling theorem would typically require. It is based on the assumption that, for a sparse signal, a small collection of linear measurements contains enough information to allow its reconstruction. Combining the acquisition and compression stages, CS is a very promising technique to develop ultra low power wireless bio-signal monitoring systems. In this paper we present a Compressive Sensing framework for ECG signals based on a universal Gaussian over-complete dictionary that permits to successfully increase the reconstruction quality performance. The purpose of the proposed dictionary is to improve ECG signal sparsity in order to achieve a higher compression ratio. Numerical experiments demonstrate that our method achieves improved performance with respect to state-of-the-art CS schemes.
There is an urgent need for a biometric that can be used for reliable identification of very young children (0 - 4 years of age), to fight child trafficking and improve vaccination uptake in the developing world. It remains unclear if the most common adult biometrics, such as face, fingerprint and iris, are stable during the first few years of life.
Finger vein recognition is a recent biometric application, which relies on the use of human finger vein patterns beneath the skin's surface. While several methods have been proposed in the literature, its applicability to uncontrolled scenarios has not yet been shown. To this purpose this paper first introduces the VERA database, a new challenging publicly available database of finger vein images. This corpus consists of 440 index finger images from 110 subjects collected with an open device in an uncontrolled way. Second, an evaluation of state-of-the-art finger vein recognition systems is performed, both on the controlled UTFVP database and on the new VERA database. This is achieved using a new open source and extensible finger vein recognition framework, which allows fair and reproducible benchmarks. Experimental results show that challenging recording conditions such as misalignments of the fingers lead to an absolute degradation in equal error rate of 2.75% up to 24.10% on VERA when compared to the best performances on UTFVP.
Biometric measurements are now often routinely adopted as a robust means of determining individual identity. Such an approach is clearly beneficial in a variety of scenarios, including those relating to medical environments. In the medical context, however, the use of biometric data can potentially offer other valuable opportunities for harnessing the power of biometrics which have a more direct bearing on healthcare monitoring and treatment delivery. In this paper we focus on the prediction of "soft" biometric data and, in particular, we describe an approach which aims to predict "higher level" characteristics about an individual, such as those which may broadly be described as emotional or mental state. We show how such a capability can be utilised in healthcare scenarios, and specifically, by presenting some initial analysis of results from newly acquired data in a keystroke-based data collection task, we identify the most crucial issues which must be addressed if our basic predictive technique is to be developed for practical viability.
In this paper we propose the Electronic Multimedia Health Fascicle (EMHF), a truly new software system for the very large number of available electronic health records. It allows the physician to see at a glance the patient's clinical biometric measurements and biologic parameters, so as to be able to link any alarming physical status to his recent medical history. Web based, accessible from any mobile device, and easy to use by both physicians and patients, the system facilitates patient-medical interaction. Using the system can also promote better adherence to medical guidelines by the physicians and to medical prescriptions and advice by the patients.
Regarding the palms recognition system studies, despite achieving a high success rate, hygiene problems in systems with contact and problems arising from changes in the alignment of the hand pose in non-contact ones have been encountered. To resolve these problems, 3D palmprint recognition systems have been developed, however these systems have not had the opportunity to spread due to expensive technologies used as well as low screening rates. In this study, to calculate the position and orientation of a palm in 3D space, a stereo camera system is proposed, with the help of these values by adjusting the geometric image of the palm is converted to a 2D environment. From these images, patterns used for recognition are extracted by using Active Appearance Models. Thus a non-contact system has been achieved, but also patterns that can be used by powerful 2D recognition systems have also been produced.
This paper presents the performance analysis of Alize/Lia_Ral algorithms in forensic speaker verification applications. In particular, in this work we evaluate the performance impact of speech signal degradation considering the background noise level, speech rate variation, audio signal length used for testing, GSM radio channel, etc. The Alize/Lia_Ral platform has demonstrated a strong dependence on the ambient noise and a slight dependence on Lombard effect, bandwidth reduction, length of the audio signal, and changes in speech rate.
Many research studies demonstrated that recognition based on ear biometrics offers an accuracy which is comparable to face trait, especially in controlled settings. Our proposal is to exploit it to avoid the problem of newborn swap, which is possible and actually happens, most of all in crowded maternity wards of big hospitals. We tested the viability of this solution using a dataset of ear images of newborns, and the obtained results testify that it is possible to decrease the probability of an error using this technique.
In this paper, we present a non-invasive, low-cost, wire-free video processing-based approach to neonatal apnoea detection. Our method consists in evaluating the presence or absence of apnoea events through an innovative analysis of a motion signal extracted from a video live capturing or recording of a patient. In particular, we pre-process the video by a recently proposed selective magnification algorithm, which has the purpose of emphasizing respiratory movements. Subsequently, by relying on a motion detection method based on the difference of consecutive frames, we extract a signal representative of the “quantity” of movement. Then, since breathing is characterized by periodic movements of specific body parts (e.g., the chest), using the Maximum Likelihood (ML) criterion we detect the presence or absence of a periodic component in the motion signal, so that the presence/absence of the respiratory movements and, therefore, of apnoea episodes can be inferred. Our method is tested on a newborn with recurrent apnoea events, affected by Congenital Central Hypoventilation Syndrome (CCHS). With the proposed method, we can identify 90-100% of the apnoea events detected by polysomnography, depending on the acceptable detection delay. The results, although preliminary, are thus very promising and show that apnoea events can be identified with non-invasive, low-cost, wire-free devices.
This paper introduces and evaluates our point-triplet spin-image descriptor, a novel descriptor that requires three vertices to be computed. This descriptor is able to encode surface information, within a spherical neighbourhood with radius r defined from a triplet's baricenter, into a surface signature. We believe that this new descriptor could be useful within a number of graph based retrieval applications; however, here we evaluate its performance within 3D face processing in the first instance. In doing so, this descriptor is embedded into a system designed to simultaneously localise the nose-tip and the two inner-eye corners of a human face. First, candidate triplets are gathered using the structured graph matching approach “relaxation by elimination” with a basic graph of three vertices and three arcs. Next, these candidate landmark-triplets are evaluated as in a binary decision problem. Hence, a point-triplet spin-image feature for each candidate landmark-triplet is computed and evaluated according to its Mahalanobis distance. This investigation includes two state of the art datasets, the Face Recognition Grand Challenge (FRGC) and CurtinFaces, as well as a performance comparison between this point-triplet spin-image and another point-triplet descriptor, named weighted-interpolated depth map which give us promising results and encourages our face processing research.
Recent studies on ECG signals proved that they can be employed as biometric traits able to obtain sufficient accuracy in a wide set of applicative scenarios. Most of the systems in the literature, however, are based on templates consisting in vectors of integer or floating point numbers. While any numerical representation is inherently binary, in here we consider as binary templates only those codings in which similarity or distance metrics can be directly applied to the for performing identity comparisons. With respect to templates composed by integer or floating point values, the use of binary templates presents important advantages, such as smaller memory space, and faster and simpler matching functions. Binary templates could therefore be adopted in a wider range of applications with respect to traditional ECG templates, like wearable devices and body area networks. Moreover, binary templates are suitable for most of the biometric template protection methods in the literature. This paper presents a novel approach for computing and processing binary ECG templates (HeartCode). Experimental results proved that the proposed approach is effective and obtains performance comparable to more mature biometric methods for ECG recognition, obtaining Equal Error Rate (EER) of 8.58% on a significantly large database of 8400 samples extracted from Holter acquisitions performed in uncontrolled conditions.
Smartphones and tablet computers are being actively studied for the performance of biometric recognition in visible spectrum. Owing to robust performance of iris recognition, many works have investigated the performance in visible spectrum. Increasing popularity of iris recognition in the visible spectrum has further resulted in using smartphones for the same. The extraction of robust features for visible spectrum iris recognition is vital to meet the expected accuracy of recognition. In this work, we explore K - means clustering based feature extraction to obtain robust features. K-means clustering is a fast alternative training method that is not computationally expensive and can easily be extended to large scale systems. The robust features extracted serves best for the unconstrained iris recognition on smartphones in visible spectrum. The proposed feature extraction technique has been extensively evaluated on publicly available smartphone iris database from BIPLab. The best Equal Error Rate of 0.31% is achieved using the proposed technique on images captured using iPhone in indoor scenario.
Recent studies in biometrics focus on one dimensional physiological signals commonly acquired in medical applications, like electrocardiogram (ECG), electroencephalograms (EEG), phonocardiogram (PCG), and photoplethysmogram (PPG). In this context, an important application is in continuous authentication scenarios since physiological signals are frequently captured for long time periods in order to monitor the health status of the patients.
The electrocardiogram (ECG) is an electrophysiologic signal that conveys signature biometric data about individuals. While this signal varies across time and with mental state, and is susceptible to noise artifact, this paper presents an approach for classifying individuals that is robust to these signal properties and noise sources. To show maximal information extraction, we take the discrete wavelet transform of a single lead ECG for feature extraction and identify individuals using a multiclass support vector machine. This approach utilizes temporal and spectral information to characterize ECG waveform morphology and timing with low computational cost. To demonstrate the effectiveness of this approach, records from the MIT-BIH normal sinus rhythm database are used to test subject identification, yielding high accuracy and positive predictive value.
A biometric identification system determines the identity of a given biometric input data, among a set of gallery identities stored in the database. The existing identification techniques typically base the decision on the match scores representing the similarity between the query and the template of each gallery. The strategies proposed so far for combining different biometric identification systems usually use only the rank one identity unimodal output in order to compute the integrated evidence of the i th person. In this paper, we propose a novel framework for fusion at hybrid rank-score level based on graph models where an additional information concerning the non-matched scores is exploited. Experiments were carried out on two multimodal databases and they demonstrate the effectiveness of the proposed approach for identification performance improvement.
Identification of humans via ECG is being increasingly studied because it can have several advantages over the traditional biometric identification techniques. However, difficulties arise because of the heartrate variability. In this study we analysed the influence of QT interval correction on the performance of an identification system based on temporal and amplitude features of ECG. In particular we tested MLP, Naive Bayes and 3-NN classifiers on the Fantasia database. Results indicate that QT correction can significantly improve the overall system performance.
In this paper, we investigate the use of a local discriminative feature space for fingerprint liveness detection. In particular, we rely on the Weber Local Descriptor (WLD), which is a powerful and robust descriptor recently proposed for texture classification. Inspired by Weber's law, it consists of two components, differential excitation and orientation, evaluated for each pixel of the image. Joint histograms of these components are then processed to build the discriminative features used to train a linear kernel SVM classifier. Experimental results with different databases and different sensors show WLD to perform favorably compared to the state-of-the-art methods in fingerprint liveness detection. In addition, by combining WLD with LPQ (Local Phase Quantization) results further improve significantly.
