Biometric Datasets Research Articles

A human vision based system for biometric images recognition

<p><span>In this paper, a universal biometric system based on human vision is proposed. From recent biological and physiological results, A human identification system that approximates the natural vision and recognition of individuals is conceived. Liquid state machine (LSM), as a recurrent spiking neural network, is highly inspired by the brain neural architecture with low training cost. However, input dimension of large scale images requires efficient processing at the cost of performance or resource overhead. This paper propose a new neural input coding for images based on frequency signals rather than pixels. Each image is filtered and fragmented then the LSM liquid (or reservoir) will receive, first, high frequency signals, then low frequency signals from each fragment. The two sets of output neurons states corresponding to each type of filter will be matched to the entire enrollment database. A weighted sum rule between the matching results will determine the right class of a biometric image. The system was tested on three different biometric datasets: face, palmprint and off-line signature, results show the reliability of the proposed approach.</span></p>

A Proposed ConvXGBoost Model for Human Activity Recognition with Multi Optimizers

The wide use of smartphones and later smartwatches equipped with a set of sensors such as location, motion, and direction blaze the trail for researchers to better recognize human activity. However, researches on using inertial or motion sensors (i.e., accelerometer, gyroscope) for human activity recognition (HAR) has intensified and reside a great confrontation to be faced. Lately, many deep learning methods have been suggested to improve the human activity classification and discrimination performance to reach an optimal accuracy. Therefore, this paper applies a Convolutional eXtreme Gradient Boosting (ConvXGBoost), which combines Convolutional Neural Network (CNN) represented by AlexNet to learn the input features automatically, followed by XGBoost decision tree used to predict the class label and thereof recognize the performed activity. Human activities are collected from sensors as time series data. Therefore, we suggested using one-dimensional AlexNet (1D AlexNet) model instead of 2D. The AlexNet model is compiled with two optimizers Adam and Stochastic Gradient Descent (SGD) which are applied consecutively. The suggested architecture was trained and evaluated on the “WISDM Smartphone and Smartwatch Activity and Biometric Dataset” that consists of raw data for eighteen activities recorded from phone and watch. The experiments revealed that using multi optimizer with a convolutional neural network improved the accuracy of recognition by 5%. Moreover, a proposed ConvXGBoost model outperformed the performance of other models works with the dataset as mentioned above with an overall accuracy of 98-99% depends on the device used.

An Efficient and Privacy-Preserving Biometric Identification Scheme Based on the FITing-Tree

Biometric identification services have been applied to almost all aspects of life. However, how to securely and efficiently identify an individual in a huge biometric dataset is still very challenging. For one thing, biometric data is very sensitive and should be kept secure during the process of biometric identification. On the other hand, searching a biometric template in a large dataset can be very time-consuming, especially when some privacy-preserving measures are adopted. To address this problem, we propose an efficient and privacy-preserving biometric identification scheme based on the FITing-tree, iDistance, and a symmetric homomorphic encryption (SHE) scheme with two cloud servers. With our proposed scheme, the privacy of the user’s identification request and service provider’s dataset is guaranteed, while the computational costs of the cloud servers in searching the biometric dataset can be kept at an acceptable level. Detailed security analysis shows that the privacy of both the biometric dataset and biometric identification request is well protected during the identification service. In addition, we implement our proposed scheme and compare it to a previously reported M-Tree based privacy-preserving identification scheme in terms of computational and communication costs. Experimental results demonstrate that our proposed scheme is indeed efficient in terms of computational and communication costs while identifying a biometric template in a large dataset.

A Study of the Influence of Data Complexity and Similarity on Soft Biometrics Classification Performance in a Transfer Learning Scenario

Transfer learning is a paradigm that consists in training and testing classifiers with datasets drawn from distinct distributions. This technique allows to solve a particular problem using a model that was trained for another purpose. In the recent years, this practice has become very popular due to the increase of public available pre-trained models that can be fine-tuned to be applied in different scenarios. However, the relationship between the datasets used for training the model and the test data is usually not addressed, specially where the fine-tuning process is done only for the fully connected layers of a Convolutional Neural Network with pre-trained weights. This work presents a study regarding the relationship between the datasets used in a transfer learning process in terms of the performance achieved by models complexities and similarities. For this purpose, we fine-tune the final layer of Convolutional Neural Networks with pre-trained weights using diverse soft biometrics datasets. An evaluation of the performances of the models, when tested with datasets that are different from the one used for training the model, is presented. Complexity and similarity metrics are also used to perform the evaluation.

An Efficient Clustering-Based Segmentation Approach for Biometric Image

Background: Image analysis plays a vital role in the biometric identification system. To achieve the effective outcome of any biometric identification system, the inputted biometric image taken should be of fine quality as it greatly impacts the decision. Image segmentation is a significant aspect of image analysis that must be carried out for enhancing the quality of an image. It efficiently differentiates the foreground and background region of the inputted biometric image and facilitates further image processing simply by providing a segmented binary image which is more coherent to the system. Objective: We present an efficient clustering based image segmentation approach to obtain the quality segmented binary image that further processed to get the quality decision in biometric based identification system. Method: A centre of mass based centroid clustering approach for image segmentation has been proposed to perform binarization of an image so as the adequate and operative results can be found. Result: The performance of the proposed approach has been applied on different sets of biometric data set having different number of hand images. Resultant this approach provides the sharp and lucid images so that good enough and effective intended results can be obtained. Conclusion: The centroid based clustering approach for image segmentation outperform the existing clustering approach. In order to measure the quality of segmented binary image three statistical performance parameters are used: PSNR, SUMD and Time Elapsed (sec).

Feature based classification of voice based biometric data through Machine learning algorithm

In the era of big data and growing artificial intelligence, the requirement and necessity of biometric identification increase in a rapid manner. The digitalization and recent Pandemic crisis gives a boost to need to authorized identification which get fulfilled with biometric identification. Our paper focuses on same concept of checking the identification accuracy of machine learning algorithm REPTree on selected biometric dataset which is being deployed and evaluated on a data mining tool WEKA. Our target is to achieve more or equal to 95 percentages in order to predict the given sample data is accurately classified into our target variables values i.e. male female. The selected algorithm REPTree is a kind of decision tree classification algorithm which works on same concept as C4.5 and decision tree algorithm with speciality of generation of both kind of output i.e. discrete and continuous. The selection of algorithm gives us benefits with achievement of higher accuracy and selection of dataset also become easy with some required modification and pre-processing of data with some dimension reduction filters.

Accurate 3D Finger Knuckle Recognition Using Auto-Generated Similarity Functions

Contactless 3D finger knuckle is an emerging biometric identifier, which can provide a promising alternative for personal identification. To maximize its potential, feature representation and matching are the two critical components towards high performance. A recent pioneering work is limited by its preliminary design of the feature descriptor and the tailor-made similarity function. Although this method demonstrates decent recognition performance, there is room for improvement. This article advances the state-of-the-art method by introducing a new curvature based feature descriptor and a method to compute the similarity functions based on the statistical distribution of the encoded feature space. Our proposed feature representation utilizes an insight in 3D geometry for accurately encoding the curvature information. When computing the similarity between a pair of templates, we compute the similarity function from the probability mass distributions of the encoded feature space. Our proposed approach is scalable to templates with different sizes, and more importantly outperforms the state-of-the-art methods significantly, which is demonstrated in a publicly available database of 3D finger knuckle. In addition, we also demonstrate the generalizability of our approach by evaluating on other publicly available biometric datasets of similar patterns, i.e., 3D palmprint and finger vein.

A biometric cryptosystem scheme based on random projection and neural network

Several biometric cryptosystem techniques have been proposed to protect biometric templates and preserve users’ privacy. Although such techniques can thwart different attacks, it is difficult to achieve well non-linkability between biometric cryptosystems. In this paper, we propose a novel biometric cryptosystem scheme based on random projection (RP) and back propagation neural network (BPNN) to perform the task of biometric template protection. With the help of RP, an original biometric feature vector can be projected onto a fix-length feature vector of random subspace that is derived from a user-specific projection matrix. This process is revocable and produces unlinkable biometric templates. The proposed scheme further utilizes a BPNN model to bind a projected feature vector with a random key. Based on BPNN, a robust mapping between a projected feature vector and a random key is learned to generate an error-correction-based biometric cryptosystem. The security of the proposed scheme is analyzed and the experimental results on multiple biometric datasets show the feasibility and efficiency of the proposed scheme.

New Interfaces and Approaches to Machine Learning When Classifying Gestures within Music.

Interactive music uses wearable sensors (i.e., gestural interfaces—GIs) and biometric datasets to reinvent traditional human–computer interaction and enhance music composition. In recent years, machine learning (ML) has been important for the artform. This is because ML helps process complex biometric datasets from GIs when predicting musical actions (termed performance gestures). ML allows musicians to create novel interactions with digital media. Wekinator is a popular ML software amongst artists, allowing users to train models through demonstration. It is built on the Waikato Environment for Knowledge Analysis (WEKA) framework, which is used to build supervised predictive models. Previous research has used biometric data from GIs to train specific ML models. However, previous research does not inform optimum ML model choice, within music, or compare model performance. Wekinator offers several ML models. Thus, we used Wekinator and the Myo armband GI and study three performance gestures for piano practice to solve this problem. Using these, we trained all models in Wekinator and investigated their accuracy, how gesture representation affects model accuracy and if optimisation can arise. Results show that neural networks are the strongest continuous classifiers, mapping behaviour differs amongst continuous models, optimisation can occur and gesture representation disparately affects model mapping behaviour; impacting music practice.

Abstract 14691: Heart Rate Trajectories and Clinical Correlates in Patients Recovering From Acute Myocardial Infarction: A Longitudinal Analysis of Apple Watch Heart Rate Recordings

Introduction: Vital signs can provide critical details regarding a patient’s health status for clinical decision making, and confirm feedback on treatments. Using mobile health technology, vital signs such as heart rate can be collected in larger quantities without requiring frequent clinic visits. In this study, we characterized HR trends of patients recovering from acute myocardial infarction (AMI) in the subacute period, with the goals of identifying higher risk patient subgroups. Methods: Between Nov. 2017 and Apr. 2019, type 1 AMI patients were enrolled in the Myocardial infarction, Combined device, Recovery Enhancement (MiCORE) study (NCT03760796). Patient information was collected via Research Electronic Data Capture (REDCap), and HRs were recorded using the Apple Watch and Corrie smartphone app between the time of admission to 30 days post-discharge. Associations with clinical covariates were assessed with multivariable linear mixed regression models (Figure 1). Results: We recruited 91 patients (mean age: 57 +/- 10.9, 74% male) who provided 4,447 HR recordings, with an overall mean daily HR of 74.5 bpm. Patients had a 0.168 bpm decrease in mean daily HR for each consecutive day post-discharge; the exception were patients who had a prior CABG before admission, hypertension, or type 2 diabetes, who had an associated elevation in HR across each day (Figure 1). Six of 11 AMI patients who were readmitted within 30 days exhibited at least 1 HR recording greater than 90 bpm within 3 days of readmission. Conclusions: While most AMI patients had a decrease in HR over the subacute period, we identified high-risk subgroups with increasing HRs (patients with prior CABG, HTN, and T2DM), and elevated HRs in patients three days prior to readmissions. This study highlights the potential of using longitudinal, biometric data sets in discerning higher risk patient populations post-AMI that require more targeted therapy.

Multi-Modal Biometrics based on Data Fusion

With the development of intelligent application, biometrics recognition technology has been widely concerned and applied in many fields of the real world, such as access control and payment. The traditional biometrics are usually based on single modality data of the subjects, but they are limited by the feature information capacity and the bottleneck in recognition accuracy. In this paper, a multi-modal biometric recognition framework is presented, which utilizes a multi-kernel learning algorithm to fuse heterogeneous information of different modal data. In order to extract complementary information from them, we combine the kernel matrix to form the mixed kernel matrix, and then give the final classification results. The experimental results on multiple biometric datasets show that our method can obtain higher recognition accuracy compared with the existing single mode and multi-mode fusion methods.

Morphometric analysis of all 24 hip muscles: A cadaveric study of 18 hip specimens with proposal of a new classification of muscles.

There are few papers investigating the morphometry of hip muscles and reporting either a single muscle or a group of synergistic muscles. With the development of hip joint simulation, hip arthroplasty implants, and gait analysis, a set of biometric data covering all the muscles that act on the hip joint is needed. Using a rigorous dissection process on 18 cadaveric hips, measurements of all 24 hip muscles and bone parameters were conducted. We measured the following: (a) total femur, femoral shaft, and neck lengths, (b) total muscle lengths, (c) intra-muscular and extra-muscular (free) tendon lengths, (d) bone angles, (e) muscle pennation, sagittal, and frontal angles, (f) muscle weight, (g) muscle volume, (h) muscle cross-sectional area, and (i) and bending moment. Data on more than 12,000 morphometric or anatomical parameters were collected. Correlation values between bone variables, muscle variables, and in-between muscle variables were computed. Based on their compliance, muscles were classified using the ratio of belly length over the sum of intra-muscular and free tendons. Values of the neck, shaft, and total femur lengths were highly correlated in relation to each other. The long muscles and the pelvitrochanteric muscles were highly correlated with femoral bone lengths. The proximal and distal intra-muscular tendon lengths were correlated to the total muscle length for all long muscles, independently of free tendon (extra-muscular) or muscle belly lengths. A very significant correlation was found between muscle weight variations among specimens. Three groups of muscles were identified based on their compliance. This is the first comprehensive anatomical morphometric study which includes all the 24 muscles acting on the hip joint. It generates a unique anatomical dataset comprising all necessary data for musculoskeletal modeling and arthroplasty implants of the hip joint. A new muscle classification was proposed based on compliance where muscles of the same group would exhibit similar compliance and functional anatomy.

Towardprecision health: applying artificial intelligence analytics to digital health biometric datasets.

The rapid development of digital health devices has enabled patients to engage in their care to an unprecedented degree and holds the possibility of significantly improving the diagnosis, treatmentand monitoring of many medical conditions. Combined with the emergence of artificial intelligence algorithms, biometric datasets produced from these digital health devices present new opportunities to create precision-based, personalized approaches for healthcare delivery. For effective implementation of such innovations to patient care, clinicians will require an understanding of the types of datasets produced from digital health technologies; the types of analytic methods including feature selection, convolution neural networking, and deep learning that can be used to analyze digital data; and how the interpretation of these findings are best translated to patient care. In this perspective, we aim to provide the groundwork for clinicians to be able to apply artificial intelligence to this transformation of healthcare.

Аналіз ефективності методів коригування промахів у системах біометричної ідентифікації на підставі електрокардіограми

Здійснено порівняння ефективності різних методів коригування промахів у біометричних системах ідентифікації. Основна ідея – виявити сегменти ЕКГ-сигналу із промахами, і замість їх вилучення з процесу ідентифікації, застосувати процедуру їх коригування. Це дасть змогу отримати більший обсяг даних і кращу статистичну базу для навчання та калібрування системи. У роботі порівнювали три різні методи усунення промахів. Перший метод базується на оцінюванні статистичного відхилення вибірок від певного номінального значення на деякий поріг. При цьому аналізується не весь сигнал одразу, а тільки його частина в межах ковзного вікна. В основі двох інших методів знаходиться ідея застосування штучних нейронних мереж, зокрема одного із їх різновидів – автоенкодерів. Відмінність між методами із використанням автоенкодерів полягає у такому: в одному випадку теж використовується ковзне вікно, що дає змогу безпосередньо задавати критерії, за якими відбувається коригування, водночас як за іншим методом виконується коригування за критеріями, які система підбирає автоматично на етапі навчання. Окрім цього, в роботі описано структуру системи біометричної ідентифікації на підставі сигналу електрокардіограми. До ключових структурних компонентів системи належать: аналоговий вимірювальний блок, АЦП та низка цифрових функціональних блоків для перетворення та аналізу сигналів. Ці блоки можуть бути імплементовані на різних обчислювальних платформах, таких як мікроконтролери, ПК, хмарні сервіси). Ці цифрові блоки виконують такі перетворення, як: низькочастотна та високочастотна фільтрація, виявлення R–піків у сигналі електрокардіограми, сегментація серцевих циклів, нормалізація за амплітудою, усунення аномалій, зменшення розмірності та класифікація. Експерименти проводили на самостійно зібраному наборі даних LBDS (Lviv Biometric Dataset). Ця база даних на момент написання статті містила понад 1400 записів для 95 різних осіб. Базова похибка ідентифікації без коригування промахів становить близько 14 %. Після застосування процедури коригування промахів похибка ідентифікації зменшилась до 2,0 % для алгоритмів на підставі автоенкодерів та до 2,9 % для алгоритмів на підставі статистичних методів. При цьому найкращі результати було досягнуто за використання LDA класифікатора у поєднанні з PCA–компресією (1,7 %), а також для KNN класифікатора без PCA–компресії (2,3 %). Проте додавання процедури коригування промахів у процес біометричної ідентифікації призводить до певного збільшення часу на опрацювання сигналу (до 20 %), що однак не критично для більшості прикладних застосувань.

Saphiraweb: An open-source cloud platform for E-health analysis

Here we present a cloud platform for e-health analysis. The goal of this research is to offers a standardized way to secure registry and share biometric impedance data from many volunteers of distinct locations. We perform the corporal analysis of a dataset from elderly volunteers who perform regular gym exercises. Our goal is to follow the evolution of health quality according to the percentage of body fat, hydration, and corporal lean mass. These factors can potentially indicate how physical exercises can contribute to improving the health quality of this population. Our result is two-fold: (a) a qualitative overview of the development of an open-source system for e-health analysis has many issues to host a useful large biometric dataset and (b) our preliminary analysis indicates that regular physical exercises contribute reducing corporal body fat, and improve the quality of health.

Innovations in research and clinical care using patient-generated health data.

Patient-generated health data (PGHD), or health-related data gathered from patients to help address a health concern, are used increasingly in oncology to make regulatory decisions and evaluate quality of care. PGHD include self-reported health and treatment histories, patient-reported outcomes (PROs), and biometric sensor data. Advances in wireless technology, smartphones, and the Internet of Things have facilitated new ways to collect PGHD during clinic visits and in daily life. The goal of the current review was to provide an overview of the current clinical, regulatory, technological, and analytic landscape as it relates to PGHD in oncology research and care. The review begins with a rationale for PGHD as described by the US Food and Drug Administration, the Institute of Medicine, and other regulatory and scientific organizations. The evidence base for clinic-based and remote symptom monitoring using PGHD is described, with an emphasis on PROs. An overview is presented of current approaches to digital phenotyping or device-based, real-time assessment of biometric, behavioral, self-report, and performance data. Analytic opportunities regarding PGHD are envisioned in the context of big data and artificial intelligence in medicine. Finally, challenges and solutions for the integration of PGHD into clinical care are presented. The challenges include electronic medical record integration of PROs and biometric data, analysis of large and complex biometric data sets, and potential clinic workflow redesign. In addition, there is currently more limited evidence for the use of biometric data relative to PROs. Despite these challenges, the potential benefits of PGHD make them increasingly likely to be integrated into oncology research and clinical care.

Optical PTFT Asymmetric Cryptosystem-Based Secure and Efficient Cancelable Biometric Recognition System

Recently, biometric systems are extensively and commonly utilized for authentication and verification applications. The security issue and the dependence on a specific biometric for the biometric verification process are the main challenges confronted in biometric systems. The security issue comes due to the exploitation of the original biometrics in stored servers. Therefore, if any attacks have been introduced to the stored biometrics, they will be missed indefinitely. Consequently, the stored original biometrics must be secured through maintaining and storing these templates away from exploitation in their servers. So, there is a need for designing a cancelable biometric recognition system (CBRS) that is a promising protection trend in biometric verification and authentication fields. The CBRS is based on the conversion of biometric data or its features to a different arrangement. In this article, a novel CBRS based on the suggested optical PTFT (Phase Truncated Fourier Transform) asymmetric encryption algorithm is introduced. In the proposed algorithm, two different distributions of phases in the output and Fourier planes are maintained as deciphering keys, and thus, the encryption keys will not be utilized for the decryption process. This leads to the advantage that the two ciphering keys may be utilized as public secret keys to encrypt distinct biometric images. Consequently, the suggested PTFT cryptosystem is an asymmetric encryption/decryption technique compared to the preceding related optical encryption techniques that are symmetric techniques such as Optical Scanning Holography (OSH) and Double Random Phase Encoding (DRPE). The suggested PTFT asymmetric encryption algorithm also has a wonderful practical performance in security applications. One of the main contributions of the proposed optical PTFT asymmetric encryption algorithm is that it removes the linearity features of the optical OSH and DRPE symmetric encryption algorithms through its great features of the phase truncation nonlinear operation. Subsequently, this produces an encrypted biometric template with two public keys, and the authenticated user can retrieve the original biometric template utilizing two private keys with achieving a high security and cancelability performance for the stored biometrics. To confirm the efficacy of the suggested optical encryption algorithm for developing a secure CBRS, various biometric datasets of face, ear, palmprint, fingerprint, and iris images are examined and analyzed. Extensive comparative analyses are performed amongst the suggested algorithm and the optical OSH and DRPE encryption algorithms. The experimental outcomes achieved for performance quality assessment assure that the suggested CBRS is reliable, robust, and realistic. It has great security and cancelability proficiency that expose excellent cancelable biometric recognition performance even in the existence of noise. Moreover, the performed experiments declare that the suggested CBRS guarantee an average FRR (False Reject Rate) of 0.0012, EER (Equal Error Rate) of 0.0019, and FAR (False Accept Rate) of 0.0030, and an average AROC (Areas under the Receiver Operating Characteristic) of 0.9996.

Cloud-assisted secure biometric identification with sub-linear search efficiency

Cloud computing has been one of the critical solutions to reduce heavy storage and computation burden of biometric identification. To protect the privacy of biometric data against untrusted cloud servers, outsourced biometric databases are usually encrypted by users. Performing biometric identification over encrypted data without revealing privacy to cloud servers attracts more and more attention. Several secure biometric identification solutions have been proposed to solve this challenging problem. However, these schemes still suffer from various limitations, such as low search efficiency and heavy computation burden on users. In this paper, we propose a novel cloud-assisted biometric identification scheme based on the asymmetric scalar-product preserving encryption (ASPE) and spatial data structures such as the R-tree index, which simultaneously achieves sub-linear search efficiency and low computation burden on users. Specifically, we construct an R-tree index on the biometric dataset and encrypt the index with ASPE. Then we modify the original search algorithm in the R-tree index and design a secure search algorithm based on ASPE to find the nearest neighbor result over the encrypted R-tree index. Through theoretical analysis and extensive experiments, we demonstrate the effectiveness and efficiency of our proposed approach.

It is All in the System's Parameters: Privacy and Security Issues in Transforming Biometric Raw Data into Binary Strings

Biometrics traits such as faces, fingerprints, and irises, are becoming prevalent in computer security applications: from authentication systems to identification systems. Given the sensitive nature of biometrics, a great deal of effort is put into protecting the biometric data after it is acquired — from secure sketch and fuzzy extractors to the use of secure multiparty computations (in protocols such as SCiFI or GSHADE). While these solutions make sure that the extracted values (e.g., binary strings or vectors) that correspond to the biometrics are kept privately and securely, their practical implementations are not optimal with respect to privacy guarantees in the process of extracting the information from the raw biometric data. This paper analyses current solutions for protected systems and discusses the existing and potential problems in the security and privacy of their feature extraction and the binarization processes. As an illustrative example, we show a PoC of an attack on a feature extraction solution from facial images, used in several protected systems, and show that it reveals information which is very close to the training image of the user. As we argue in this paper, other solutions provide privacy for the system's users but make use of external set of biometric data which is often quite large, thus facing privacy and ownership issues associated with the external set of people. The take home message of this paper is: Many of the existing “privacy preserving” solutions neglect the privacy and security aspects of the feature extraction and binarization processes. Hence, we urge future research to close this gap in the security and privacy of biometric systems.

Eyebrows and eyeglasses as soft biometrics using deep learning

Occlusion due to eyeglasses is one of the main challenges affecting the face and general ocular recognition, including eyebrow matching. In this study, the authors propose a convolutional neural network (CNN)-based method for (a) eyeglasses detection and segmentation to mitigate its impact on personal recognition in mobile devices and (b) use the shape of the glasses as a soft token of identity (something that one has). They evaluated the efficacy of the proposed eyeglasses segmentation on eyebrow matching and eyeglasses-based user authentication. To this front, various texture and deep features were evaluated. Using the publicly available large-scale visible ocular biometric dataset, they show that the proposed methods provide (a) eyeglasses detection and segmentation accuracies of 100 and 97% using CNNs, (b) a 2.51% reduction in eyebrow matching error by removing eyeglass occlusions and (c) eyeglasses matching with a 96.6% accuracy.