Here, we present a computational photography-based biometric scanner to capture deep wrist vascular arcades using multi-zone NIR LED banks and multi-exposure HDR per each illumination zone, achieving exceptional pseudo-color imaging of the wrist volume, in addition to other anatomical features such as high-resolution skin textures. Using NIR illumination on dorsal, lateral, and ventral positions, deeper vascular structures are encoded in red and green, and superficial textures and veins in blue. As an internal biometric, this modality inherently provides stronger security and privacy. Implemented on an NVIDIA Jetson platform, we further leverage secure computation using ARM TrustZone isolation, encrypted data storage, and secure boot functionalities. Experimental results validate the effectiveness and computational efficiency of the proposed system, demonstrating its potential for secure, real-time, and privacy-preserving biometric applications.

Generative AI has revolutionized modern machine learning by providing unprecedented realism, diversity, and efficiency in data generation. This technology holds immense potential for biometrics, including for securing sensitive and personally identifiable information. Given the irrevocability of biometric samples and mounting privacy concerns, biometric template security and secure matching are among the most sought-after features of modern biometric systems. This paper proposes a novel obfuscation method using Generative AI to enhance biometric template security. Our approach utilizes synthetic facial images generated by a Generative Adversarial Network (GAN) as "random chaff points" within a secure vault system. Our method creates n sub-templates from the original template, each obfuscated with m GAN chaff points. During verification, s closest vectors to the biometric query are retrieved from each vault and combined to generate hash values, which are then compared with the stored hash value. Thus, our method safeguards user identities during the training and deployment phases by employing the GAN-generated synthetic images. Our protocol was tested using the AT&T, GT, and LFW face datasets, achieving ROC areas under the curve of 0.99, 0.99, and 0.90, respectively. Our results demonstrate that the proposed method can maintain high accuracy and reasonable computational complexity comparable to those unprotected template methods while significantly enhancing security and privacy, underscoring the potential of Generative AI in developing proactive defensive strategies for biometric systems.

Given the popularity and importance of biometric personal identification, new modalities continue to be explored by the research community, including those based on the infrared scans of the target tissues. This paper presents a study of functional near-infrared spectroscopy (fNIRS) as a potential biometric modality. We assess the subject-specificity of the fNIRS’s functional and structural signatures using a multichannel fNIRS dataset collected over the forehead during various motor tasks and rest states. We detail our preprocessing, feature extraction (with Uniform Manifold Approximation and Projection), and feature selection (with permutation feature importance). Using cosine similarity matching between enrollment and validation features from an average of four mental actions, we obtained an area under the ROC curve (AUC) of 0.86. We also compare the efficacy of various classifiers, such as tree-based, neural network-based, and gradient-boosting-based models. An ensemble classifier improved the verification accuracy and AUC over the test data set to 96% and 0.99, respectively.

Many computer vision methods have been proposed for the affective assessment using facial expressions from full-face images. In many use cases, however, only the ocular region may be available due to the application of masks, clothing items, or privacy issues. In this paper, we show the utility of a robust and yet light deep learning model for ocular affect assessment using cross-dataset evaluation, where we train the model on one dataset and perform testing on another dataset with different demographics. We compare a MobileNet-V2 deep learning model, using transfer learning, with a more traditional method using histogram of oriented gradients (HOG) features with support vector machine (SVM) classifier. Experiments were conducted on the FACES dataset for training with six facial expressions and tested on the more diverse Chicago faces dataset(CFD) to show how evaluated models generalize not only in cross-dataset evaluation but also in the presence of new ethnicities not present during training. The experimental results show that the deep learning model can provide an average accuracy of 76.77% overall facial expressions when compared to HOG and SVM's 62.47% for this challenging cross-dataset emotion assessment using only eye regions.

In this paper, we present an energy efficient motion sensing platform for wireless body area networks that supports real-time, persistent gesture recognition through motion tracking sensors. The platform consists of multiple heterogeneous wearable sensors. The custom-built body sensors include inertial sensors, a low-power microcontroller and a 2.4 GHz radio transceiver. Signal classifiers such as Fishers linear discriminant classifiers, static neural networks, and focused time delay neural networks (fTDNN) are employed to classify the signals obtained from the wearable sensors. It was found that at a sampling rate of 10 Hz and just 4 bits/sample, the fTDNN classifier achieves 88% classification rate. Our results show that reducing the sampling and quantization rates could be used in energy constrained sensor networks for gesture recognition.

Recent reported advances in smartphone based ocular biometric recognition in visible spectrum demonstrated the efficacy of deep-learning schemes. In this paper, we evaluate convolutional neural networks (CNNs) pretrained for large scale object recognition, namely VGG-16, VGG-19, InceptionNet and ResNet, and fine-tuned for ocular recognition using RGB images captured by smartphones. Fine-tuning pretrained CNN models is advantageous in case of insufficient training data, and the partial training is faster compared to custom CNN trained from scratch. Experiments on VISOB dataset yielded TPR of up to 100% at FPR of 10-4using VGG-16 model fine-tuned for ocular recognition.

Automated age classification has drawn significant interest in numerous applications such as marketing, forensics, human-computer interaction, and age simulation. A number of studies have demonstrated that age can be automatically deduced from face images. However, few studies have explored the possibility of computational estimation of age information from other modalities such as fingerprint or ocular region. The main challenge in age classification is that age progression is person-specific which depends on many factors such as genetics, health conditions, life style, and stress level. In this paper, we investigate age classification from ocular images acquired using smart-phones. Age information, though not unique to the individual, can be combined along with ocular recognition system to improve authentication accuracy or invariance to the ageing effect. To this end, we propose a convolutional neural network (CNN) architecture for the task. We evaluate our proposed CNN model on the ocular crops of the recent large-scale Adience benchmark for gender and age classification captured using smart-phones. The obtained results establish a baseline for deep learning approaches for age classification from ocular images captured by smart-phones.

Biometrie authentication is being exceedingly utilized into mobile devices as an alternative to passwords. However, for security and privacy reasons, it is important to protect the biometric template. In this paper, we introduce a method to obfuscate and match certain biometric templates comprised of multiple local descriptors derived around spatial interest points. Obfuscation starts by insertion of chaff (fake) interest points along with their respective synthesized descriptors that are statistically similar to original descriptors. Fusion of local matches along with global outlier rejection is used to mitigate the impact of the resulting obfuscation on biometric matching accuracy. The efficacy of the proposed method is demonstrated through ocular and face recognition experiments.

With the unprecedented mobile technology revolution, a number of ocular biometric based personal recognition schemes have been proposed for mobile use cases. The aim of this competition is to evaluate and compare the performance of mobile ocular biometric recognition schemes in visible light on a large scale database (VISOB Dataset ICIP2016 Challenge Version) using standard evaluation methods. Four different teams from universities across the world participated in this competition, submitting five algorithms altogether. The submitted algorithms applied different texture analysis in a learning or a non-learning based framework for ocular recognition. The best results were obtained by a team from Norwegian Biometrics Laboratory (NTNU, Norway), achieving an Equal Error Rate of 0.06% over a quarantined test set.

Vascular Similarly Measurement (VSM) is an important tool in many biomedical applications. However, designing a robust computational VSM remains a challenge. We investigate different wavelet families and their orders to find their efficacy as feature extractors for computational VSM. Using a 50-subject dataset of RGB ocular surface vasculature images, we show that a compact feature vector composed of wavelet packet energies derived from Db1 wavelets, in conjunction with Fisher linear discriminant analysis and judged by the ensuing ROCs, is best suited for this task. Coif1 and Rbio2.4 were found to be the next best two wavelets for this purpose. Repetition of the same experiments using neural networks confirmed the optimality of the above suite of features for VSM.

Using a dSLR camera with macro LED light, 11 samples containing light and moderately cracked concrete surfaces were imaged with perpendicular and angled illumination. Textural features from gray level co-occurrence matrix statistics were derived, from which 3-6 salient features were selected. Cross validation accuracies were as high as 94% using neural network classifiers, indicating the feasibility of rapid, automatic concrete cracking assessment using COTS digital imaging.

We used Fisher linear discriminant analysis (LDA), static neural networks (NN), and focused time delay neural networks (TDNN) for gesture recognition. Gestures were collected in form of acceleration signals along three axes from six participants. A sports watch containing a 3-axis accelerometer, was worn by the users, who performed four gestures. Each gesture was performed for ten seconds, at the speed of one gesture per second. User-dependent and user-independent k-fold cross validations were carried out to measure the classifier performance. Using first and second order statistical descriptors of acceleration signals from validation datasets, LDA and NN classifiers were able to recognize the gestures at an average rate of 86% and 97% (user-dependent) and 89% and 85% (user-independent), respectively. TDNNs proved to be the best, achieving near perfect classification rates both for user-dependent and user-independent scenarios, while operating directly on the acceleration signals alleviating the need for explicit feature extraction.

Psychophysiological measurements of startle eyeblink can provide information about the state of an individual regarding sensory, attentional, cognitive, and affective processing, and thus reveal valences of interest for affective computing. However, eyeblink is usually measured using intrusive contact electromyographic (EMG) electrodes, accompanied by a laborious manual process of feature extraction. We introduce a new noninvasive automatic system using high-speed video recording of startle blinks in conjunction with data-driven feature selection and support vector machine (SVM) ensembles to classify startle eyeblinks. Using a prestimulus (prepulse) to produce robust modulation of acoustically elicited startle eyeblinks, we tracked the blinks using 250 frames per second video, and extracted different features from eyelid displacement and velocity signals. The SVMs were able to determine whether a trial had contained startle or prepulse+startle stimuli with an accuracy of up to 73 percent (five-fold cross validation). By fusing the decisions made on different feature sets, an ensemble of seven SVMs increased this rate to almost 79 percent. Since startle eyeblinks are robustly modulated by not only sensory events (such as the prepulse used in this study) but also affective and cognitive states, eyelid tracking using high-speed video, in conjunction with the introduced classification method, is an effective and user-friendly alternative to EMG for classification of startle blinks to infer users' affective-cognitive states.

In this paper, we show the feasibility of using high-speed video for measurement of startle eyeblinks as a new augmentative modality for biometric security, as blinks can reveal emotional states of interest in security screenings using nonintrusive measurements. Using neural network as classifiers, this initial study shows that upper eyelid tracking at 250 frames per second can categorize startle blinks with accuracies comparable to those of the well-established but intrusive EMG-based measures of muscles in charge of eyelid closure.

Given their multiresolution temporal and spectral locality, wavelets are powerful candidates for decomposition, feature extraction, and classification of non-stationary electroencephalographic (EEG) signals for brain-computer interface (BCI) applications. Wavelet feature extraction methods offer several options through the choice of wavelet families and decomposition architectures. The classification results of EEG signals generated from imagined motor, cognitive, and affective tasks are presented using support vector machine (SVM) classifiers, indicating that these methods are suitable for imagined motor, cognitive and affective classification. Classifier performances of better than 80% for six imagined motor tasks, and for two affective tasks were achieved. Three cognitive tasks were successfully classified with 70% accuracy. The methods can be used with a variety of EEG signal reference methods and electrode placement locations. Wavelet features performed satisfactorily in the presence of noise when the classifiers were presented with contaminated training data.

In an earlier work we had explored the possibility of utilizing the vascular pattern of the sclera, episclera, and conjunctiva as a biometric indicator. These blood vessels, which can be observed on the white part of the human eye, demonstrate rich and seemingly unique details in visible light, and can be easily imaged using commercially available digital cameras. In this work we discuss a new method to represent and match the textural intricacies of this vascular structure using wavelet-derived features in conjunction with neural network classifiers. Our experimental results, based on the evidence of 50 subjects, indicate the potential of the proposed scheme to characterize the individuality of the ocular surface vascular patterns and further confirm our assertion that these patterns are indeed unique across individuals.

Among the more challenging problems in the design of temporal neural networks are the incorporation of short and long-term memories and the choice of network topology. Delayed copies of network signals can form short-term memory (STM), whereas feedback loops can constitute long-term memories (LTM). This paper introduces a new general evolutionary temporal neural network framework (GETnet) for the automated design of neural networks with distributed STM and LTM. GETnet is a step towards the realization of general intelligent systems that can be applied to a broad range of problems. GETnet utilizes nonlinear moving average and autoregressive nodes and sub-circuits that are trained by enhanced gradient descent and evolutionary search in architecture, synaptic delay, and synaptic weight spaces. The ability to evolve arbitrary time-delay connections enables GETnet to find novel answers to classification and system identification tasks. A new temporal minimum description length policy ensures creation of fast and compact networks with improved generalization capabilities. Simulations using Mackey-Glass time series are presented to demonstrate the above stated capabilities of GETnet.

Fingerprint scanners may be susceptible to spoofing using artificial materials, or in the worst case, dismembered fingers. An anti-spoofing method based on liveness detection has been developed for use in fingerprint scanners. This method quantifies a specific temporal perspiration pattern present in fingerprints acquired from live claimants. The enhanced perspiration detection algorithm presented here improves our previous work by including other fingerprint scanner technologies; using a larger, more diverse data set; and a shorter time window. Several classification methods were tested in order to separate live and spoof fingerprint images. The dataset included fingerprint images from 33 live subjects, 33 spoofs created with dental material and Play-Doh, and fourteen cadaver fingers. Each method had a different performance with respect to each scanner and time window. However, all the classifiers achieved approximately 90% classification rate for all scanners, using the reduced time window and the more comprehensive training and test sets.

A method for temporal pattern recognition for continuous time signals is addressed. It is shown how a simple form of back-propagation can be used in conjunction with a temporal error signal to adapt both the weights and path delays of a continuous time delay feed forward multi-layer neural network with hard-limited output. An instance of such a network is simulated and some of the results are discussed. During the initial tests the network showed robust capabilities for detection of temporal patterns, including fast recognition of onsets of new waveforms in presence of moderately heavy noise and phase and frequency distortions.