Unconstrained Conditions Research Articles

Objective Detection of Cigarette Smoking from Physiological Sensor Signals.

Cigarette smoking has severe health impacts on those who smoke and the people around them. Several wearable sensing modalities have recently been investigated to collect objective data on daily smoking, including detection of smoking episodes from breathing patterns, hand to mouth behavior, and characteristic hand gestures or cigarette lighting events. In order to provide new insight into ongoing research on the objective collection of smoking-related events, this paper proposes a novel method to identify smoking events from the associated changes in heart rate parameters specific to smoking. The proposed method also accounts for the breathing rate and body motion of the person who is smoking to better distinguish these changes from intense physical activities. In this research, a human study was first performed on 20 daily cigarette smokers to record heart rate, breathing rate, and body acceleration collected from a wearable chest sensor consisting of an ECG and bioimpedance measurement sensor and a 3D inertial sensor. Each participant spent ~2 hours in a laboratory environment (mimicking daily activities that included smoking 4 cigarettes) and ~24 hours under unconstrained free-living conditions. A support vector machine-based classifier was developed to automatically detect smoking episodes from the captured sensor signals using fifteen features selected by a forward-feature selection method. In a leave one subject out cross-validation, the proposed approach detected smoking events (187 out of total 232) with the sensitivity and F-score accuracy of 0.87 and 0.79, respectively, in the laboratory setting (known activities) and 0.77 and 0.61, respectively, under free-living conditions. These results validate the proof-of-concept that, although further research is necessary for performance improvement, characteristic changes in heart rate parameters could be a useful indicator of cigarette smoking even under free-living conditions.

Distortion of pipe-flow development by boundary layer growth and unconstrained inlet conditions

This study extends the analysis of the canonical developing pipe-flow problem to realistic inlet conditions affecting emerging jets. A comparison of simulations to existing theory reveals adverse phenomena caused by the inlet: the velocity profile inversion and flow separation (vena contracta) at a sharp inlet. Beginning with the simple uniform inflow, the inversion is shown to persists at significantly higher Re (Re = 2000) than previously reported. It is found to be caused by the theory’s neglected radial velocity, resulting from the boundary layer’s displacement effect. Rescaling of the inlet axial coordinate is shown to collapse all centerline velocity curves above Re = 100, thus elucidating the known weak dependence of entrance-length on Re. The sharp-inlet separation bubble is found not to occur below Re ≅ 320 although this inlet profile overrides the boundary layer’s effect. Furthermore, the bubble’s downstream length increases rapidly with Re, whereas its upstream length grows gradually and proportionally to its thickness—here identified as its characteristic-scale. Beyond the bubble, the profile relaxes to a monotonic form—captured beyond x/(Re·R) = 0.005, if theory is modified using the bubble’s characteristic-scale. This scale also sets the threshold which differentiates between a sharp-inlet regime, accompanied by a separation bubble, and a rounded-inlet one without it. The latter regime relaxes more rapidly to the monotonic profile—captured already beyond x = 2R. Finally, the modified idealized theory is demonstrated as a useful design tool—explicitly relating nozzle length to characteristics of emerging free-surface and submerged jets.

Detecting small faces in the wild based on generative adversarial network and contextual information

Face detection techniques have been developed for decades, and one of the remaining open challenges is detecting small faces in unconstrained conditions. The reason is that tiny faces are often lacking detailed information and blurry. In this paper, we proposed an algorithm to directly generate a clear high-resolution face from a small blurry one by adopting a generative adversarial network (GAN). Toward this end, the basic GAN formulation achieves it by super-resolving and refining sequentially (e.g. SR-GAN and Cycle-GAN). However, we design a novel network to address the problem of super-resolving and refining jointly. Moreover, we also introduce new training losses (i.e. classification loss and regression loss) to promote the generator network to recover fine details of the small faces and to guide the discriminator network to distinguish face vs. non-face and to refine location simultaneously. Additionally, considering the importance of contextual information when detecting tiny faces in crowded cases, the context around face regions is combined to train the proposed GAN-based network for mining those very small faces from unconstrained scenarios. Extensive experiments on the challenging datasets WIDER FACE and FDDB demonstrate the effectiveness of the proposed method in restoring a clear high-resolution face from a small blurry one, and show that the achieved performance outperforms previous state-of-the-art methods by a large margin.

Theoretical and numerical analysis of the mechanical responses of BCC and FCC lattice structures

Analytical and numerical analyses for predicting the mechanical responses of BCC and FCC lattice structures under compressive loading were performed and verified by comparing them to experimental data. The analytical and numerical results are in excellent correlation, while the error between the experimental and numerical results is about 5.8 ~ 15.3 %, which was caused by the inconsistent lattice strut diameter and the material elastic modulus. The measured lattice strut diameter is smaller than the designed diameter by thermal shrinkage, so the thermal shrinkage should be taken into the lattice structure design in additive manufacturing processes. Based on the same analytical and numerical techniques, a parametric investigation of the elastic moduli of BCC and FCC lattice unit cells with respect to the strut aspect ratio, specific density and strut angle was performed. Finally, the elastic moduli of multi-cell BCC and FCC lattice structures with the same density were investigated by varying the number of rows and columns of lattice unit cells in both unconstrained and constrained boundary conditions. As the strut aspect ratio, density and angle increase, the elastic moduli of both BCC and FCC lattice increase. FCC elastic modulus is higher than that of BCC lattice with respect to the strut aspect ratio, density and angle, while BCC elastic modulus could be found to be higher than of FCC lattice for a constrained boundary condition.

Assessment of human motoneuron excitability during functional motor tasks

Detailed investigations of the human motor system are often performed under isometric conditions, although few real‐world activities incorporate isometric contractions. Because of their one‐to‐one relationship with muscle fibers and high safety factor, spinal motoneurons are the only cells in the central nervous system from which firing patterns can be readily accessed in humans. Recent advances in motor unit action potential (MUAP) decomposition techniques using high‐density surface array electrodes allow large numbers of MUAPs to be discriminated relatively non‐invasively. The activity of multiple motor units can allow for the quantification of motoneuron excitability. For example, the paired motor unit analysis technique compares the onset and offset of a high‐threshold motor unit with respect to a low‐threshold unit and quantifies discharge rate hysteresis (ΔF), which is proportional to the intrinsic excitability of the spinal motoneuron. During movements where joint torque measurements are not feasible (i.e. unconstrained by the dynamometer) electromyography (EMG) is often used as a surrogate for the level of muscle activity. However, whether activity level feedback provided by EMG measurements or torque during voluntary contractions elicits similar ΔF values is unknown. We hypothesized that during voluntary isometric ramp contractions, ΔF is similar whether feedback is provided with a torque signal or conditioned EMG. We further hypothesized that standing postural adjustments using EMG feedback in unconstrained conditions will produce similar ΔF values to those calculated during the isometric ramps. We expected that although the mean values across conditions will be similar, the inter‐subject coefficient of variation of ΔF will be greater in the less constrained condition. Accordingly, surface EMG using high‐density array electrodes was collected from five subjects under three conditions: 1) seated in a dynamometer in the isometric mode with activity level feedback provided by the torque (ISO‐TOR), 2) in the isometric dynamometer with feedback provided by EMG (ISO‐EMG), and 3) during postural adjustments while standing freely with feedback provided by EMG (POS‐EMG). In agreement with our hypothesis, we found that ΔF values across all three conditions were similar. In the ISO‐TOR and ISO‐EMG conditions, ΔF was 2.70 ± 0.70 imp/s and 3.40 ± 0.75 imp/s, respectively; during POS‐EMG it was 3.08 ± 0.73 imp/s (all n = 5, mean ± SE). Contrary to our expectations, the coefficient of variation was not greater in the unconstrained standing condition. These results demonstrate that real‐time EMG feedback is sufficiently similar to torque feedback during both isometric and near‐isometric, unconstrained contractions, suggesting that detailed investigations of motor unit activity may be expanded to more functional movements.Support or Funding InformationFunded by R01 NS098509‐01 and T32 HD007418‐27This abstract is from the Experimental Biology 2019 Meeting. There is no full text article associated with this abstract published in The FASEB Journal.

Robust Authentication Using Dorsal Hand Vein Images

This paper presents a robust dorsal hand vein authentication system. A new method is proposed for the region of interest extraction using fingertips and finger valley key points. Some new features and a new classifier are proposed based on information set theory. Information set stems from a fuzzy set on representing the uncertainty in its attribute/information source values using the information-theoretic entropy function. The new feature types include vein effective information, vein energy feature, vein sigmoid feature, Shannon transform feature, and composite transform feature. A classifier called the improved Hanman classifier is formulated from training and test feature vectors using Frank t-norm and the entropy function. The performance and robustness are evaluated on GPDS and BOSPHORUS palm dorsal vein database under both the constrained and unconstrained conditions.

Food properties influence grasping strategies in strepsirrhines

Although hand grasping is ubiquitous in primate species, its origins remain uncertain. This is in part because uncertainty about hand skills and grasping strategies persists in strepsirrhines, a monophyletic group of primates located near the base of the primate tree. In this study, we report and discuss our observations of the different grasping strategies adopted by 85 captive individuals belonging to 22 species of strepsirrhines during the grasping of food items of different sizes and consistencies. Our results indicate that although strepsirrhines do not present variability in their hand-grip types (sole whole-hand power grip), they are able to adjust their grasping strategy depending on the properties of the food. Notably, they use the mouth when more precision is needed (i.e. to grasp small items). Moreover, grasping strategies adopted for big items differ depending on food consistency, revealing a new and potentially essential factor to consider in future research on grasping abilities. We believe that by looking across this important set of species in unconstrained standardized conditions, this study provides valuable insight for further comparative research on the potential selective pressures involved in the evolution of hand grasping.

Weighted similarity and distance metric learning for unconstrained face verification with 3D frontalisation

In this study, the authors focus on the challenging problem of verifying faces captured under unconstrained conditions. Unconstrained face images often vary largely in poses, illuminations, expressions, occlusions, and ages. To address these challenges, they combine face frontalisation method with metric learning. To deal with the variations of poses, they apply an improved 3D face frontalisation method to generate the frontal view of the face images. Recent studies observed that bilinear similarity and Mahalanobis distance have a promising performance on measuring the similarity of two images. Based on these studies, they propose a weighted similarity and distance metric learning method which balances the role of bilinear similarity and Mahalanobis distance to better measure the similarity of an image pair. All the experiments are conducted based on the labelled faces in the wild database and the experimental results show the effectiveness of their method.

Cigarette Smoking Detection with An Inertial Sensor and A Smart Lighter.

In recent years, a number of wearable approaches have been introduced for objective monitoring of cigarette smoking based on monitoring of hand gestures, breathing or cigarette lighting events. However, non-reactive, objective and accurate measurement of everyday cigarette consumption in the wild remains a challenge. This study utilizes a wearable sensor system (Personal Automatic Cigarette Tracker 2.0, PACT2.0) and proposes a method that integrates information from an instrumented lighter and a 6-axis Inertial Measurement Unit (IMU) on the wrist for accurate detection of smoking events. The PACT2.0 was utilized in a study of 35 moderate to heavy smokers in both controlled (1.5–2 h) and unconstrained free-living conditions (~24 h). The collected dataset contained approximately 871 h of IMU data, 463 lighting events, and 443 cigarettes. The proposed method identified smoking events from the cigarette lighter data and estimated puff counts by detecting hand-to-mouth gestures (HMG) in the IMU data by a Support Vector Machine (SVM) classifier. The leave-one-subject-out (LOSO) cross-validation on the data from the controlled portion of the study achieved high accuracy and F1-score of smoking event detection and estimation of puff counts (97%/98% and 93%/86%, respectively). The results of validation in free-living demonstrate 84.9% agreement with self-reported cigarettes. These results suggest that an IMU and instrumented lighter may potentially be used in studies of smoking behavior under natural conditions.

Can runners maintain a newly learned gait pattern outside a laboratory environment following gait retraining?

BackgroundPrevious peak tibial shock gait retraining programs, which were usually conducted on a treadmill, were reported to be effective on impact loading reduction in runners. However, whether the trained runners can translate the training effect at different running modes (treadmill/overground), or running slopes (uphill/downhill), remains unknown. Research questionIs the training effect from a treadmill-based gait retraining translatable to unconstrained running conditions, including overground and uphill/downhill running? MethodsThe peak tibial shock was measured during treadmill/overground running, as well as level/uphill/downhill running before and after a course of treadmill-based gait retraining. The 8-session training aimed to soften footfalls using real-time biofeedback of tibial shock data. Repeated measures ANOVA was used to examine the effect of training, running mode, and running slope, on a group level. Reliable change index of each participant was used to assess the individual response to the training protocol used in this study. ResultsEighty percent of the participants were responsive to the gait retraining and managed to reduce their peak tibial shock following training. They managed to translate the training effect to treadmill slope running (Level: p < 0.05, Cohen’s d = 1.65; Uphill: p = 0.001, Cohen’s d = 0.91; Downhill: p < 0.05; Cohen’s d = 1.29) and overground level running (p = 0.014, Cohen’s d = 0.85). However, their peak tibial shock were not reduced during overground slope running (Uphill: p = 0.054; Cohen’s d = 0.62; Downhill p = 0.12; Cohen’s d = 0.48). SignificanceOur findings indicated that a newly learned gait pattern may not fully translate to running outside of the laboratory environment.

Robust Single-Sample Face Recognition by Sparsity-Driven Sub-Dictionary Learning Using Deep Features.

Face recognition using a single reference image per subject is challenging, above all when referring to a large gallery of subjects. Furthermore, the problem hardness seriously increases when the images are acquired in unconstrained conditions. In this paper we address the challenging Single Sample Per Person (SSPP) problem considering large datasets of images acquired in the wild, thus possibly featuring illumination, pose, face expression, partial occlusions, and low-resolution hurdles. The proposed technique alternates a sparse dictionary learning technique based on the method of optimal direction and the iterative -norm minimization algorithm called k-LiMapS. It works on robust deep-learned features, provided that the image variability is extended by standard augmentation techniques. Experiments show the effectiveness of our method against the hardness introduced above: first, we report extensive experiments on the unconstrained LFW dataset when referring to large galleries up to 1680 subjects; second, we present experiments on very low-resolution test images up to pixels; third, tests on the AR dataset are analyzed against specific disguises such as partial occlusions, facial expressions, and illumination problems. In all the three scenarios our method outperforms the state-of-the-art approaches adopting similar configurations.

Acoustic radiation pressure in laterally unconfined plane wave beams

The acoustic radiation pressure in laterally unconfined, plane wave beams in inviscid fluids is derived via the direct application of finite deformation theory for which an analytical accounting is made ab initio that the radiation pressure is established under static, laterally unconstrained conditions, while the acoustic wave that generates the radiation pressure propagates under dynamic (sinusoidal), laterally constrained conditions. The derivation reveals that the acoustic radiation pressure for laterally unconfined, plane waves along the propagation direction is equal to (3/4)〈2K〉, where is the mean kinetic energy density of the wave, and zero in directions normal to the propagation direction. The results hold for both Lagrangian and Eulerian coordinates. The value differs from the value traditionally used in the assessment of acoustic radiation pressure, obtained from the Langevin theory or from the momentum flux density in the Brillouin stress tensor. Errors in traditional derivations leading to the Brillouin stress tensor and the Langevin radiation pressure are pointed out and a long-standing misunderstanding of the relationship between Lagrangian and Eulerian quantities is corrected. The present theory predicts a power output from the transducer that is 4/3 times larger than that predicted from the Langevin theory. Tentative evidence for the validity of the present theory is provided from measurements previously reported in the literature, revealing the need for more accurate and precise measurements for experimental confirmation of the present theory.

PGC-Net: A Light Weight Convolutional Sequence Network for Digital Pressure Gauge Calibration

Automatic digital pressure gauge calibration is challenging due to various unconstrained conditions. Although existing CNN-RNN based methods have been almost perfect on scene text recognition, they fail to perform well on digital pressure gauge calibration that requires to be extremely computation-efficient and accurate. In this paper, we propose a light weight fully convolutional sequence recognition network for fast and accurate digital P ressure G auge C alibration (PGC-Net). PGC-Net integrates feature extraction, sequence modelling and transcription into a unified framework. Experimental results show that PGC-Net runs 28 fps on CPU with 97.41% accuracy. Compared with previous methods, PGC-Net achieves better or comparable performance at lower inference time. Without bells and whistles, PGC-Net is capable of recognizing decimal points that usually appear in pressure gauge images, which evidently verifies the feasibility of PGC-Net. We collected a dataset that contains 17, 240 gauge images with annotated labels for automatic digital pressure gauge calibration. The dataset has been public for future research.

An Adaptive CNNs Technology for Robust Iris Segmentation

Iris segmentation algorithms are of great significance in complete iris recognition systems, and directly affect the iris verification and recognition results. However, the conventional iris segmentation algorithms have poor adaptability and are not sufficiently robust when applied to noisy iris databases captured under unconstrained conditions. In addition, there are currently no large iris databases; thus, the iris segmentation algorithms cannot maximize the benefits of convolutional neural networks (CNNs). The main work of this paper is as follows: first, we propose an architecture based on CNNs combined with dense blocks for iris segmentation, referred to as a dense-fully convolutional network (DFCN), and adopt some popular optimizer methods, such as batch normalization (BN) and dropout. Second, because the public ground-truth masks of the CASIA-Interval-v4 and IITD iris databases do not include the labeled eyelash regions, we label these regions that occlude the iris regions using the Labelme software package. Finally, the promising results of experiments based on the CASIA-Interval-v4, IITD, and UBIRIS.V2 iris databases captured under different conditions reveal that the iris segmentation network proposed in this paper outperforms all of the conventional and most of the CNN-based iris segmentation algorithms with which we compared our algorithm’s results in terms of various metrics, including the accuracy, precision, recall, f1 score, and nice1 and nice2 error scores, reflecting the robustness of our proposed network.

Non-Contact Infrared-Depth Camera-Based Method for Respiratory Rhythm Measurement While Driving

This work proposes a new non-contact method based on Infrared and Depth cameras to measure respiratory rhythm in real-life situations. The proposed algorithm consists on using both video feeds to track the movements of the subject in real-time, compute the location of the face and calculate the most suitable ROI to extract the respiratory signal in adult population. 20 subjects were measured while driving in a car simulator with no constraints other than the simulator itself. The algorithm has been validated using a commercial thorax plethysmography system. An opportunistic approach has been used to obtain pieces from the tests for each subject, thus making a more realistic approach to the real-life situations where the signal is likely to contain errors. The breath-to-breath respiratory signal and the instantaneous frequency from both methods has been computed from each piece to characterise the error between the proposed method and the reference system. The results show a high correlation between the measured rhythm from the reference method and the proposed method, with relatively low error results and good sensitivity in cycle detection and low errors for the instantaneous frequency between methods. The error results have also been compared with the ones obtained in previous studies showing a good agreement between the obtained results and the ones presented in the previous studies. No relationship between the length of the pieces and the error has been found either for the respiratory cycle signal or the instantaneous frequency signal. The proposed algorithm can be used to measure respiratory rhythm in unconstrained conditions and with opportunistic measurements, thus making it suitable to perform in real-life situations while driving. Further studies taking into account vibrations or light changing conditions are needed to confirm that the proposed method performs with the same accuracy with these constraints.

Occlusion aware facial expression recognition using CNN with attention mechanism.

Facial expression recognition in the wild is challenging due to various un-constrained conditions. Although existing facial expression classifiers have been almost perfect on analyzing constrained frontal faces, they fail to perform well on partially occluded faces that are common in the wild. In this paper, we propose a Convolution Neutral Network with attention mechanism (ACNN) that can perceive the occlusion regions of the face and focus on the most discriminative unoccluded regions. ACNN is an end to end learning framework. It combines the multiple representations from facial regions of interest (ROIs). Each representation is weighed via a proposed Gate Unit that computes an adaptive weight from the region itself according to the unobstructed-ness and importance. Considering different RoIs, we introduce two versions of ACNN: patch based ACNN (pACNN) and global-local based ACNN (gACNN). pACNN only pays attention to local facial patches. gACNN integrates local representations at patch-level with global representation at image-level. The proposed ACNNs are evaluated on both real and synthetic occlusions, including a self-collected facial expression dataset with real-world occlusions (FED-RO), two largest in-the-wild facial expression datasets (RAF-DB and AffectNet) and their modifications with synthesized facial occlusions. Experimental results show that ACNNs improve the recognition accuracy on both the non-occluded faces and occluded faces. Visualization results demonstrate that, compared with the CNN without Gate Unit, ACNNs are capable of shifting the attention from the occluded patches to other related but unobstructed ones. ACNNs also outperform other state-of-the-art methods on several widely used in-the-lab facial expression datasets under the cross-dataset evaluation protocol.

Computational Assessment of Facial Expression Production in ASD Children.

In this paper, a computational approach is proposed and put into practice to assess the capability of children having had diagnosed Autism Spectrum Disorders (ASD) to produce facial expressions. The proposed approach is based on computer vision components working on sequence of images acquired by an off-the-shelf camera in unconstrained conditions. Action unit intensities are estimated by analyzing local appearance and then both temporal and geometrical relationships, learned by Convolutional Neural Networks, are exploited to regularize gathered estimates. To cope with stereotyped movements and to highlight even subtle voluntary movements of facial muscles, a personalized and contextual statistical modeling of non-emotional face is formulated and used as a reference. Experimental results demonstrate how the proposed pipeline can improve the analysis of facial expressions produced by ASD children. A comparison of system’s outputs with the evaluations performed by psychologists, on the same group of ASD children, makes evident how the performed quantitative analysis of children’s abilities helps to go beyond the traditional qualitative ASD assessment/diagnosis protocols, whose outcomes are affected by human limitations in observing and understanding multi-cues behaviors such as facial expressions.

Joint Head Attribute Classifier and Domain-Specific Refinement Networks for Face Alignment

In this article, a two-stage refinement network is proposed for facial landmarks detection on unconstrained conditions. Our model can be divided into two modules, namely the Head Attribude Classifier (HAC) module and the Domain-Specific Refinement (DSR) module. Given an input facial image, HAC adopts multi-task learning mechanism to detect the head pose and obtain an initial shape. Based on the obtained head pose, DSR designs three different CNN-based refinement networks trained by specific domain, respectively, and automatically selects the most approximate network for the landmarks refinement. Different from existing two-stage models, HAC combines head pose prediction with facial landmarks estimation to improve the accuracy of head pose prediction, as well as obtaining a robust initial shape. Moreover, an adaptive sub-network training strategy applied in the DSR module can effectively solve the issue of traditional multi-view methods that an improperly selected sub-network may result in alignment failure. The extensive experimental results on two public datasets, AFLW and 300W, confirm the validity of our model.

Particle transport in gravel‐bed rivers: Revisiting passive tracer data

AbstractData from tracer experiments were compiled and analysed in order to explore the role of geomorphological, hydrological and sedimentological constraints on fluvial gravel transport in gravel‐bed rivers. A large data set from 217 transport episodes of tagged stones were compiled from 33 scientific papers. Our analyses showed that while magnitude of peak discharge is a major control on gravel transport and mobility, tracer travel distances show some scale dependence on the morphological configuration of the channel. Our results also highlight differences in the way tracers are displaced between step–pool and riffle and pool channels. The riffle–pool sequence seems to be a more efficient trap for travelling gravels than the step–pool pair. In addition, in step–pool channels there are clear differences in tracer transport between observations of first displacements after tracer seeding (unconstrained‐stone conditions), and second and subsequent observations of tracer displacements (constrained‐stone conditions). The comparison between tracer experiments under constrained conditions and those under unconstrained conditions also highlights the importance of bed state and structures in gravel mobility. The results of this study confirm that sediment transport in gravel‐bed rivers is a complex process, whereby sedimentological and geomorphological controls are superimposed on the hydraulic forcing. © 2018 John Wiley &amp; Sons, Ltd.

New Approaches for Automatic Face Recognition Based on Deep Learning Models and Local Handcrafted ALTP

Face recognition is one of the most important topics in biometrics, where it achieved great success under controlled scenarios. Still, its accuracy degraded significantly in unconstrained conditions. To meet this challenge, we proposed a handcraft method based on extracting import