[go: up one dir, main page]
More Web Proxy on the site http://driver.im/
Next Issue
Volume 21, May-2
Previous Issue
Volume 21, April-2
You seem to have javascript disabled. Please note that many of the page functionalities won't work as expected without javascript enabled.
 
 
sensors-logo

Journal Browser

Journal Browser

Sensors, Volume 21, Issue 9 (May-1 2021) – 413 articles

Cover Story (view full-size image): Admittance control of robotic manipulators requires force/torque measurement at the end effector (EE). This is usually pursued by means of direct force/torque sensors located at the EE. While this allows accurate measurement, it is a very expensive approach, and the size of the sensor is further problematic. In this paper, a novel approach is presented where the EE forces/torques are indirectly deduced from the force/torque measurement at the base of the robot. To this end, a dedicated sensor is developed that measures the ground reaction wrench at the base. The sensor concept relies on a model-based calibration that accounts for the non-linear dynamics of the robot and the pose-dependent transmission characteristics. Two sensor setups are investigated, one using ground-fixed load cells, and the other using a tailored four-spoke structure equipped with strain gauges. View this paper.
  • Issues are regarded as officially published after their release is announced to the table of contents alert mailing list.
  • You may sign up for e-mail alerts to receive table of contents of newly released issues.
  • PDF is the official format for papers published in both, html and pdf forms. To view the papers in pdf format, click on the "PDF Full-text" link, and use the free Adobe Reader to open them.
Order results
Result details
Section
Select all
Export citation of selected articles as:
26 pages, 22394 KiB  
Review
3D Printing Techniques and Their Applications to Organ-on-a-Chip Platforms: A Systematic Review
by Violeta Carvalho, Inês Gonçalves, Teresa Lage, Raquel O. Rodrigues, Graça Minas, Senhorinha F. C. F. Teixeira, Ana S. Moita, Takeshi Hori, Hirokazu Kaji and Rui A. Lima
Sensors 2021, 21(9), 3304; https://doi.org/10.3390/s21093304 - 10 May 2021
Cited by 83 | Viewed by 12126
Abstract
Three-dimensional (3D) in vitro models, such as organ-on-a-chip platforms, are an emerging and effective technology that allows the replication of the function of tissues and organs, bridging the gap amid the conventional models based on planar cell cultures or animals and the complex [...] Read more.
Three-dimensional (3D) in vitro models, such as organ-on-a-chip platforms, are an emerging and effective technology that allows the replication of the function of tissues and organs, bridging the gap amid the conventional models based on planar cell cultures or animals and the complex human system. Hence, they have been increasingly used for biomedical research, such as drug discovery and personalized healthcare. A promising strategy for their fabrication is 3D printing, a layer-by-layer fabrication process that allows the construction of complex 3D structures. In contrast, 3D bioprinting, an evolving biofabrication method, focuses on the accurate deposition of hydrogel bioinks loaded with cells to construct tissue-engineered structures. The purpose of the present work is to conduct a systematic review (SR) of the published literature, according to the guidelines of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses, providing a source of information on the evolution of organ-on-a-chip platforms obtained resorting to 3D printing and bioprinting techniques. In the literature search, PubMed, Scopus, and ScienceDirect databases were used, and two authors independently performed the search, study selection, and data extraction. The goal of this SR is to highlight the importance and advantages of using 3D printing techniques in obtaining organ-on-a-chip platforms, and also to identify potential gaps and future perspectives in this research field. Additionally, challenges in integrating sensors in organs-on-chip platforms are briefly investigated and discussed. Full article
(This article belongs to the Special Issue Organ-on-a-Chip and Biosensors)
Show Figures

Graphical abstract

Graphical abstract
Full article ">Figure 1
<p>Schematic diagram showing the preclinical models used in biomedical research.</p>
Full article ">Figure 2
<p>PRISMA flow diagram displaying the procedure of study selection.</p>
Full article ">Figure 3
<p>Number of papers and the respective year of publication included in the SR.</p>
Full article ">Figure 4
<p>Representation of the biosensors for a human-on-a-chip platform. Reprinted from ref. [<a href="#B70-sensors-21-03304" class="html-bibr">70</a>].</p>
Full article ">
22 pages, 762 KiB  
Article
LSTMs and Deep Residual Networks for Carbohydrate and Bolus Recommendations in Type 1 Diabetes Management
by Jeremy Beauchamp, Razvan Bunescu, Cindy Marling, Zhongen Li and Chang Liu
Sensors 2021, 21(9), 3303; https://doi.org/10.3390/s21093303 - 10 May 2021
Cited by 8 | Viewed by 3748
Abstract
To avoid serious diabetic complications, people with type 1 diabetes must keep their blood glucose levels (BGLs) as close to normal as possible. Insulin dosages and carbohydrate consumption are important considerations in managing BGLs. Since the 1960s, models have been developed to forecast [...] Read more.
To avoid serious diabetic complications, people with type 1 diabetes must keep their blood glucose levels (BGLs) as close to normal as possible. Insulin dosages and carbohydrate consumption are important considerations in managing BGLs. Since the 1960s, models have been developed to forecast blood glucose levels based on the history of BGLs, insulin dosages, carbohydrate intake, and other physiological and lifestyle factors. Such predictions can be used to alert people of impending unsafe BGLs or to control insulin flow in an artificial pancreas. In past work, we have introduced an LSTM-based approach to blood glucose level prediction aimed at “what-if” scenarios, in which people could enter foods they might eat or insulin amounts they might take and then see the effect on future BGLs. In this work, we invert the “what-if” scenario and introduce a similar architecture based on chaining two LSTMs that can be trained to make either insulin or carbohydrate recommendations aimed at reaching a desired BG level in the future. Leveraging a recent state-of-the-art model for time series forecasting, we then derive a novel architecture for the same recommendation task, in which the two LSTM chain is used as a repeating block inside a deep residual architecture. Experimental evaluations using real patient data from the OhioT1DM dataset show that the new integrated architecture compares favorably with the previous LSTM-based approach, substantially outperforming the baselines. The promising results suggest that this novel approach could potentially be of practical use to people with type 1 diabetes for self-management of BGLs. Full article
(This article belongs to the Special Issue Sensor Technologies: Artificial Intelligence for Diabetes Management)
Show Figures

Figure 1

Figure 1
<p>The neural network architecture for the carbohydrate recommendation scenario. The dashed blue line plots BG levels, while the solid red line represents the basal rate of insulin. The gray star represents the meal event at time <math display="inline"><semantics> <mrow> <mi>t</mi> <mo>+</mo> <mn>10</mn> </mrow> </semantics></math>. Other meals are represented by squares, whereas boluses are represented by circles. Meals and boluses with a red outline can appear only in unrestricted examples. The blue <math display="inline"><semantics> <msub> <mi>LSTM</mi> <mn>1</mn> </msub> </semantics></math> units receive input from time steps in the past. The green <math display="inline"><semantics> <msub> <mi>LSTM</mi> <mn>2</mn> </msub> </semantics></math> units receive input from time steps in the prediction window. The purple block stands for the fully connected layers of the FCN that computes the prediction.</p>
Full article ">Figure 2
<p>The general neural network architecture for the bolus and bolus given carbs recommendation scenarios. The architecture itself is similar to that shown in <a href="#sensors-21-03303-f001" class="html-fig">Figure 1</a>. The gray star now represents the bolus at time <math display="inline"><semantics> <mrow> <mi>t</mi> <mo>+</mo> <mn>10</mn> </mrow> </semantics></math>. For the bolus recommendation scenario, the events outlined in red or orange are not allowed in inertial examples. However, in the bolus given carbs scenario, the meal event <math display="inline"><semantics> <msub> <mi>C</mi> <mrow> <mi>t</mi> <mo>+</mo> <mn>20</mn> </mrow> </msub> </semantics></math> shown with the yellow outline is an important part of each example, be it inertial or unrestricted. As such, in this scenario, the dashed <math display="inline"><semantics> <msub> <mi>C</mi> <mrow> <mi>t</mi> <mo>+</mo> <mn>20</mn> </mrow> </msub> </semantics></math> becomes part of the input to the FCN.</p>
Full article ">Figure 3
<p>The N-BEATS inspired deep residual architecture for carbohydrate recommendation. A similar architecture is used for bolus and bolus given carbs recommendations.</p>
Full article ">Figure 4
<p>Boxplots showing the absolute error per subject for each recommendation scenario achieved by the N-BEATS.best model in the inertial scenario. The orange lines within each box represent the median absolute errors, while the red crosses represent the average absolute errors. The green circles represent outliers. To avoid stretching the figures, outliers were clipped at 25 for the Carbs<math display="inline"><semantics> <msup> <mrow/> <mrow> <mo>(</mo> <mo>±</mo> <mi>b</mi> <mo>)</mo> </mrow> </msup> </semantics></math> scenario, 8 for the Bolus<math display="inline"><semantics> <msup> <mrow/> <mrow> <mo>(</mo> <mo>±</mo> <mi>c</mi> <mo>)</mo> </mrow> </msup> </semantics></math> scenario, and 5 for the Bolus<math display="inline"><semantics> <msup> <mrow/> <mrow> <mo>(</mo> <mo>+</mo> <mi>c</mi> <mo>)</mo> </mrow> </msup> </semantics></math> scenario. The number of clipped outliers is shown next to the subject’s largest outlier. Above the top line is shown, for each subject, the percentage of test examples that are outliers.</p>
Full article ">
17 pages, 25394 KiB  
Article
Multiphase Identification Algorithm for Fall Recording Systems Using a Single Wearable Inertial Sensor
by Chia-Yeh Hsieh, Hsiang-Yun Huang, Kai-Chun Liu, Chien-Pin Liu, Chia-Tai Chan and Steen Jun-Ping Hsu
Sensors 2021, 21(9), 3302; https://doi.org/10.3390/s21093302 - 10 May 2021
Cited by 4 | Viewed by 2817
Abstract
Fall-related information can help clinical professionals make diagnoses and plan fall prevention strategies. The information includes various characteristics of different fall phases, such as falling time and landing responses. To provide the information of different phases, this pilot study proposes an automatic multiphase [...] Read more.
Fall-related information can help clinical professionals make diagnoses and plan fall prevention strategies. The information includes various characteristics of different fall phases, such as falling time and landing responses. To provide the information of different phases, this pilot study proposes an automatic multiphase identification algorithm for phase-aware fall recording systems. Seven young adults are recruited to perform the fall experiment. One inertial sensor is worn on the waist to collect the data of body movement, and a total of 525 trials are collected. The proposed multiphase identification algorithm combines machine learning techniques and fragment modification algorithm to identify pre-fall, free-fall, impact, resting and recovery phases in a fall process. Five machine learning techniques, including support vector machine, k-nearest neighbor (kNN), naïve Bayesian, decision tree and adaptive boosting, are applied to identify five phases. Fragment modification algorithm uses the rules to detect the fragment whose results are different from the neighbors. The proposed multiphase identification algorithm using the kNN technique achieves the best performance in 82.17% sensitivity, 85.74% precision, 73.51% Jaccard coefficient, and 90.28% accuracy. The results show that the proposed algorithm has the potential to provide automatic fine-grained fall information for clinical measurement and assessment. Full article
(This article belongs to the Special Issue Wearable Sensors for Gait and Falls Monitoring)
Show Figures

Figure 1

Figure 1
<p>Diagram of an acceleration-based signal of a fall process.</p>
Full article ">Figure 2
<p>Functional diagram of proposed multiphase identification algorithm.</p>
Full article ">Figure 3
<p>Diagram of the experimental environment setting.</p>
Full article ">Figure 4
<p>The sensor orientation, wearing position of a sensor, and the subject has worn protectors in the experiment. (<b>a</b>) Sensor orientation; (<b>b</b>) The sensor was worn on the waist (lower back); (<b>c</b>,<b>d</b>) The front and back view of the subject worn protectors, respectively.</p>
Full article ">Figure 5
<p>Diagram of proposed fragment modification algorithm. An example to modify one (situation 1), two (situation 2), or three (situation 3) segments that are different from previous and following segments. These segments (misclassified segments) should be modified to the same with the previous or following segments.</p>
Full article ">Figure 6
<p>An example of a process in the multiphase identification. The fragment modified results were compared against the ground truth in terms of <span class="html-italic">TN</span>, <span class="html-italic">TP</span>, <span class="html-italic">FP</span> and <span class="html-italic">FN</span>.</p>
Full article ">Figure 7
<p>The average performance using different machine learning techniques.</p>
Full article ">Figure 8
<p>The average performance using different window sizes.</p>
Full article ">Figure 9
<p>The average performance of proposed algorithm using the kNN technique with each window size.</p>
Full article ">
15 pages, 2352 KiB  
Article
Thermal Face Verification through Identification
by Artur Grudzień, Marcin Kowalski and Norbert Pałka
Sensors 2021, 21(9), 3301; https://doi.org/10.3390/s21093301 - 10 May 2021
Cited by 4 | Viewed by 2648
Abstract
This paper reports on a new approach to face verification in long-wavelength infrared radiation. Two face images were combined into one double image, which was then used as an input for a classification based on neural networks. For testing, we exploited two external [...] Read more.
This paper reports on a new approach to face verification in long-wavelength infrared radiation. Two face images were combined into one double image, which was then used as an input for a classification based on neural networks. For testing, we exploited two external and one homemade thermal face databases acquired in various variants. The method is reported to achieve a true acceptance rate of about 83%. We proved that the proposed method outperforms other studied baseline methods by about 20 percentage points. We also analyzed the issue of extending the performance of algorithms. We believe that the proposed double image method can also be applied to other spectral ranges and modalities different than the face. Full article
(This article belongs to the Section Sensing and Imaging)
Show Figures

Figure 1

Figure 1
<p>Gallery of thermal face images from (<b>a</b>) PROTECT dataset, (<b>b</b>) CARL dataset, (<b>c</b>) In-House (FLIR A65), (<b>d</b>) and In-House (FLIR P640).</p>
Full article ">Figure 2
<p>Scheme of combining two images into genuine and impostor classes.</p>
Full article ">Figure 3
<p>Scheme of the verification through identification method.</p>
Full article ">Figure 4
<p>ROC curves for individual databases: (<b>A</b>) CNNs with metrics, (<b>B</b>) CNNs with SVM, (<b>C</b>) local descriptor methods with metrics, and (<b>D</b>) local descriptor methods with SVM.</p>
Full article ">Figure 5
<p>ROC curves for individual databases for the VTI method.</p>
Full article ">Figure 6
<p>The best ROC curves for five face verification methods.</p>
Full article ">
11 pages, 1965 KiB  
Article
Developing Sidewalk Inventory Data Using Street View Images
by Bumjoon Kang, Sangwon Lee and Shengyuan Zou
Sensors 2021, 21(9), 3300; https://doi.org/10.3390/s21093300 - 10 May 2021
Cited by 20 | Viewed by 4221
Abstract
(1) Background: Public sidewalk GIS data are essential for smart city development. We developed an automated street-level sidewalk detection method with image-processing Google Street View data. (2) Methods: Street view images were processed to produce graph-based segmentations. Image segment regions were manually labeled [...] Read more.
(1) Background: Public sidewalk GIS data are essential for smart city development. We developed an automated street-level sidewalk detection method with image-processing Google Street View data. (2) Methods: Street view images were processed to produce graph-based segmentations. Image segment regions were manually labeled and a random forest classifier was established. We used multiple aggregation steps to determine street-level sidewalk presence. (3) Results: In total, 2438 GSV street images and 78,255 segmented image regions were examined. The image-level sidewalk classifier had an 87% accuracy rate. The street-level sidewalk classifier performed with nearly 95% accuracy in most streets in the study area. (4) Conclusions: Highly accurate street-level sidewalk GIS data can be successfully developed using street view images. Full article
Show Figures

Figure 1

Figure 1
<p>Street-level sidewalk classification process.</p>
Full article ">Figure 2
<p>Custom-built software labels sidewalks by visually inspecting the original and segmented images side-by-side. The software asks an analyst to put the pointer where the sidewalk is present and labels the associated segment.</p>
Full article ">Figure 3
<p>Street-level sidewalk classification illustration.</p>
Full article ">Figure 4
<p>Sidewalk street image points and the study area.</p>
Full article ">Figure 5
<p>Importance of variables in the random forest model.</p>
Full article ">
22 pages, 3508 KiB  
Article
Probabilistic Load Forecasting Optimization for Building Energy Models via Day Characterization
by Eva Lucas Segarra, Germán Ramos Ruiz and Carlos Fernández Bandera
Sensors 2021, 21(9), 3299; https://doi.org/10.3390/s21093299 - 10 May 2021
Cited by 5 | Viewed by 2711
Abstract
Accurate load forecasting in buildings plays an important role for grid operators, demand response aggregators, building energy managers, owners, customers, etc. Probabilistic load forecasting (PLF) becomes essential to understand and manage the building’s energy-saving potential. This research explains a methodology to optimize the [...] Read more.
Accurate load forecasting in buildings plays an important role for grid operators, demand response aggregators, building energy managers, owners, customers, etc. Probabilistic load forecasting (PLF) becomes essential to understand and manage the building’s energy-saving potential. This research explains a methodology to optimize the results of a PLF using a daily characterization of the load forecast. The load forecast provided by a calibrated white-box model and a real weather forecast was classified and hierarchically selected to perform a kernel density estimation (KDE) using only similar days from the database characterized quantitatively and qualitatively. A real case study is presented to show the methodology using an office building located in Pamplona, Spain. The building monitoring, both inside—thermal sensors—and outside—weather station—is key when implementing this PLF optimization technique. The results showed that thanks to this daily characterization, it is possible to optimize the accuracy of the probabilistic load forecasting, reaching values close to 100% in some cases. In addition, the methodology explained is scalable and can be used in the initial stages of its implementation, improving the values obtained daily as the database increases with the information of each new day. Full article
Show Figures

Figure 1

Figure 1
<p>Components and steps of the probabilistic load forecasting procedure based on white-box models (building energy model (BEM)) [<a href="#B53-sensors-21-03299" class="html-bibr">53</a>].</p>
Full article ">Figure 2
<p>Simulation process methodology [<a href="#B53-sensors-21-03299" class="html-bibr">53</a>].</p>
Full article ">Figure 3
<p>Process of the probabilistic load forecast [<a href="#B53-sensors-21-03299" class="html-bibr">53</a>].</p>
Full article ">Figure 4
<p>Hierarchy of filters proposed for the selection of the training days.</p>
Full article ">Figure 5
<p>Building’s outdoor photograph (<b>top left</b>), simulation model image from OpenStudio [<a href="#B60-sensors-21-03299" class="html-bibr">60</a>] (<b>bottom left</b>), and the weather station located on the roof of the building (<b>right</b>).</p>
Full article ">Figure 6
<p>PICP, MPIW, and training days for the different filters and days’ characterization. <b>Top</b>: 21 October 2019. <b>Bottom</b>: 21 November 2019.</p>
Full article ">Figure 7
<p>Graphical results of the PLF for the four filters for the last day of the test period, 31 January 2020.</p>
Full article ">
18 pages, 3061 KiB  
Article
Relevance of Drift Components and Unit-to-Unit Variability in the Predictive Maintenance of Low-Cost Electrochemical Sensor Systems in Air Quality Monitoring
by Georgi Tancev
Sensors 2021, 21(9), 3298; https://doi.org/10.3390/s21093298 - 10 May 2021
Cited by 15 | Viewed by 4074
Abstract
As key components of low-cost sensor systems in air quality monitoring, electrochemical gas sensors have recently received a lot of interest but suffer from unit-to-unit variability and different drift components such as aging and concept drift, depending on the calibration approach. Magnitudes of [...] Read more.
As key components of low-cost sensor systems in air quality monitoring, electrochemical gas sensors have recently received a lot of interest but suffer from unit-to-unit variability and different drift components such as aging and concept drift, depending on the calibration approach. Magnitudes of drift can vary across sensors of the same type, and uniform recalibration intervals might lead to insufficient performance for some sensors. This publication evaluates the opportunity to perform predictive maintenance solely by the use of calibration data, thereby detecting the optimal moment for recalibration and improving recalibration intervals and measurement results. Specifically, the idea is to define confidence regions around the calibration data and to monitor the relative position of incoming sensor signals during operation. The emphasis lies on four algorithms from unsupervised anomaly detection—namely, robust covariance, local outlier factor, one-class support vector machine, and isolation forest. Moreover, the behavior of unit-to-unit variability and various drift components on the performance of the algorithms is discussed by analyzing published field experiments and by performing Monte Carlo simulations based on sensing and aging models. Although unsupervised anomaly detection on calibration data can disclose the reliability of measurement results, simulation results suggest that this does not translate to every sensor system due to unfavorable arrangements of baseline drifts paired with sensitivity drift. Full article
(This article belongs to the Section Chemical Sensors)
Show Figures

Figure 1

Figure 1
<p>Distribution of sensor currents for a system consisting of two sensors. (<b>a</b>) Classification of anomalies in signal space with supervised anomaly detection. (<b>b</b>) Classification of anomalies in signal space with unsupervised anomaly detection.</p>
Full article ">Figure 2
<p>(<b>a</b>) Fraction of anomalies (weekly moving average) over four months for different <span class="html-italic">ν</span> across algorithms after two weeks of field calibration for sensor data of device 1. (<b>b</b>) Fraction of anomalies (weekly moving average) over four months for different <span class="html-italic">ν</span> across algorithms after two weeks of field calibration for reference data.</p>
Full article ">Figure 3
<p>Distribution of the three sensor populations (q<sub>0.50</sub>, q<sub>0.25–0.75</sub>, q<sub>0.05–0.95</sub>) with respect to zero (baseline) and sensitivity by exposure to 200 evenly spaced gas fractions (CO: 0–1000 ppb, NO<sub>2</sub>: 0–200 ppb, O<sub>3</sub>: 0–200 ppb) without mixing.</p>
Full article ">Figure 4
<p>Distribution of sensor signals (q<sub>0.50</sub>, q<sub>0.25–0.75</sub>, q<sub>0.05–0.95</sub>) over time under constant conditions. (<b>a</b>) At low fractions (CO: 50 ppb, NO<sub>2</sub>: 50 ppb, O<sub>3</sub>: 50 ppb), the baseline drift is mildly recognizable (particularly for CO-B4). (<b>b</b>) At high fractions (CO: 600 ppb, NO<sub>2</sub>: 600 ppb, O<sub>3</sub>: 600 ppb), the curvature due to the sensitivity drift is clearly visible (particularly for NO<sub>2</sub>-B4 and OX-B4); this drift leads to convergence towards the baseline.</p>
Full article ">Figure 5
<p>Distribution of sensor currents for a system consisting of two sensors. (<b>a</b>) Direction of aging drift (blue) is opposite to trajectory of atmosphere (green); trajectories are along largest expansion of the envelope. Signals decrease due to aging, but atmospheric amount fractions increase. Thus, the aging process is masked. (<b>b</b>) After some time, the generated signal intensities will be low enough to be outside the envelope and identified.</p>
Full article ">Figure 6
<p>Distribution of sensor currents for a system consisting of two sensors. (<b>a</b>) Zero drift (blue) moves the baseline (lowest possible signal intensity) in any direction, while sensitivity drift (green) causes the maximum possible signal intensity to decay exponentially. (<b>b</b>) Example of a signal distribution that has shifted inside the envelope, thus no anomalies were detected.</p>
Full article ">Figure 7
<p>Performance of the four ML algorithms (<span class="html-italic">ν</span> = 0.10) as distribution of anomalies (monthly moving average) from the population of low-cost sensor systems (q<sub>0.50</sub>, q<sub>0.25–0.75</sub>, q<sub>0.05–0.95</sub>), computed in Monte Carlo simulations including laboratory calibration followed by aging.</p>
Full article ">
14 pages, 3491 KiB  
Communication
A Two Joint Neck Model to Identify Malposition of the Head Relative to the Thorax
by Philipp M. Schmid, Christoph M. Bauer, Markus J. Ernst, Bettina Sommer, Lars Lünenburger and Martin Weisenhorn
Sensors 2021, 21(9), 3297; https://doi.org/10.3390/s21093297 - 10 May 2021
Cited by 3 | Viewed by 3717
Abstract
Neck pain is a frequent health complaint. Prolonged protracted malpositions of the head are associated with neck pain and headaches and could be prevented using biofeedback systems. A practical biofeedback system to detect malpositions should be realized with a simple measurement setup. To [...] Read more.
Neck pain is a frequent health complaint. Prolonged protracted malpositions of the head are associated with neck pain and headaches and could be prevented using biofeedback systems. A practical biofeedback system to detect malpositions should be realized with a simple measurement setup. To achieve this, a simple biomechanical model representing head orientation and translation relative to the thorax is introduced. To identify the parameters of this model, anthropometric data were acquired from eight healthy volunteers. In this work we determine (i) the accuracy of the proposed model when the neck length is known, (ii) the dependency of the neck length on the body height, and (iii) the impact of a wrong neck length on the models accuracy. The resulting model is able to describe the motion of the head with a maximum uncertainty of 5 mm only. To achieve this high accuracy the effective neck length must be known a priory. If however, this parameter is assumed to be a linear function of the palpable neck length, the measurement error increases. Still, the resulting accuracy can be sufficient to identify and monitor a protracted malposition of the head relative to the thorax. Full article
(This article belongs to the Special Issue Impact of Sensors in Biomechanics, Health Disease and Rehabilitation)
Show Figures

Figure 1

Figure 1
<p>The neck model shown in blue contains the neck-stick which is assumed to be rigidly connected to the neck rigid body and the sternum rigid body. The extended model comprises one pose sensing tripod at the sternum and one at the forehead. These two tripods are assumed to be rigidly connected to head-stick and the thorax-stick. The joints <math display="inline"><semantics> <msub> <mi>J</mi> <mn>1</mn> </msub> </semantics></math> and <math display="inline"><semantics> <msub> <mi>J</mi> <mn>2</mn> </msub> </semantics></math> are ball joints. The extended model, comprising a head-stick and head rigid body is introduced for identification of the neck-stick <math display="inline"><semantics> <msub> <mi>l</mi> <mn>12</mn> </msub> </semantics></math>.</p>
Full article ">Figure 2
<p>Determination of the protraction <span class="html-italic">p</span> from the orientation of the neck relative to the orientation of the thorax. The red arrow is fixed to the thorax tripod. It is defined to be exactly vertical when the upright posture is assumed during the calibration phase. The distance between C2 and C7, measurable by palpation is denoted as <math display="inline"><semantics> <msub> <mi>l</mi> <mrow> <mi>C</mi> <mn>2</mn> <mi>C</mi> <mn>7</mn> </mrow> </msub> </semantics></math>.</p>
Full article ">Figure 3
<p>Measurement setup—one participant seated on a stool. The head pose tripoid and the sternum pose tripoid with reflective markers are visible. The neck orientation tripoid is fixated to the black collar. The illustration of the movement pattern is attached to the vertical bar in a way that is clearly visible to the participant.</p>
Full article ">Figure 4
<p>Linear model to describe the fitted length of vector <math display="inline"><semantics> <msub> <mi>l</mi> <mn>12</mn> </msub> </semantics></math> depending on the distance between C2 and C7. Data points are labeled with the subject number. The linear model describes the distance <math display="inline"><semantics> <msub> <mi>l</mi> <mn>12</mn> </msub> </semantics></math> with a large residual error.</p>
Full article ">Figure 5
<p>Influence of position noise on the parameter estimation.</p>
Full article ">Figure 6
<p>Influence of orientation noise on parameter estimation: A bias towards smaller neck lengths is introduced.</p>
Full article ">Figure 7
<p>Influence of position and orientation noise on parameter estimation.</p>
Full article ">
11 pages, 513 KiB  
Communication
CIR-Based Device-Free People Counting via UWB Signals
by Mauro De Sanctis, Aleandro Conte, Tommaso Rossi, Simone Di Domenico and Ernestina Cianca
Sensors 2021, 21(9), 3296; https://doi.org/10.3390/s21093296 - 10 May 2021
Cited by 7 | Viewed by 3775
Abstract
The outbreak of COVID-19 has resulted in many different policies being adopted across the world to reduce the spread of the virus. These policies include wearing surgical masks, hand hygiene practices, increased social distancing and full country-wide lockdown. Specifically, social distancing involves keeping [...] Read more.
The outbreak of COVID-19 has resulted in many different policies being adopted across the world to reduce the spread of the virus. These policies include wearing surgical masks, hand hygiene practices, increased social distancing and full country-wide lockdown. Specifically, social distancing involves keeping a certain distance from others and avoiding gathering together in large groups. Automatic crowd density estimation is a technological solution that could help in guaranteeing social distancing by reducing the probability that two persons in a public area come in close proximity to each other while moving around. This paper proposes a novel low complexity RF sensing system for automatic people counting based on low cost UWB transceivers. The proposed system is based on an ordinary classifier that exploits features extracted from the channel impulse response of UWB communication signals. Specifically, features are extracted from the sorted list of singular values obtained from the singular value decomposition applied to the matrix of the channel impulse response vector differences. Experimental results achieved in two different environments show that the proposed system is a promising candidate for future automatic crowd density monitoring systems. Full article
(This article belongs to the Special Issue Communications Signal Processing and Networking in the Pandemic)
Show Figures

Figure 1

Figure 1
<p>Singular values for different number of people in room A.</p>
Full article ">Figure 2
<p>Layout of Room A and Room B.</p>
Full article ">Figure 3
<p>Normalized Channel Impulse Response, Room A, empty.</p>
Full article ">Figure 4
<p>Normalized Channel Impulse Response, Room A with 5 People moving in the room.</p>
Full article ">Figure 5
<p>Scatter plot of the features <span class="html-italic">Area under the curve</span> vs. <span class="html-italic">Center of gravity</span>-for the different classes in Room A with maximum 5 people.</p>
Full article ">Figure 6
<p>Scatter plot of the features <span class="html-italic">Area under the curve</span> vs. <span class="html-italic">Center of gravity</span>-for the different classes in Room B with maximum 8 people.</p>
Full article ">
29 pages, 1469 KiB  
Article
Enabling Reliable UAV Control by Utilizing Multiple Protocols and Paths for Transmitting Duplicated Control Packets
by Woonghee Lee
Sensors 2021, 21(9), 3295; https://doi.org/10.3390/s21093295 - 10 May 2021
Cited by 5 | Viewed by 3087
Abstract
In the last ten years, supported by the advances in technologies for unmanned aerial vehicles (UAVs), UAVs have developed rapidly and are utilized for a wide range of applications. To operate UAVs safely, by exchanging control packets continuously, operators should be able to [...] Read more.
In the last ten years, supported by the advances in technologies for unmanned aerial vehicles (UAVs), UAVs have developed rapidly and are utilized for a wide range of applications. To operate UAVs safely, by exchanging control packets continuously, operators should be able to monitor UAVs in real-time and deal with any problems immediately. However, due to any networking problems or unstable wireless communications, control packets can be lost or transmissions can be delayed, which causes the unstable drone control. To overcome this limitation, in this paper, we propose MuTran for enabling reliable UAV control. MuTran considers the packet type and duplicates only control packets, not data packets. After that, MuTran transmits the original and duplicate packets through multiple protocols and paths to improve the reliability of control packet transmissions. We designed MuTran and conducted a lot of theoretical analyses to demonstrate the validity of MuTran and analyze it from various aspects. We implemented MuTran on real devices and evaluated MuTran using the devices. We conducted experiments to verify the limitations of the existing systems and demonstrate that control packets can be transmitted more stably by using MuTran. Through the analysis and experimental results, we confirmed that MuTran reduces the control packet transfer delay, which improves the reliability and stability of controlling UAVs. Full article
(This article belongs to the Special Issue Time-Sensitive Networks for Unmanned Aircraft Systems)
Show Figures

Figure 1

Figure 1
<p>The packet transmissions of various cases. (This figure is drawn to distinctly present the difference between cases, so the ratio of control packets to data packets is exaggerated.)</p>
Full article ">Figure 2
<p>The difference between TCP and UDP modes of <span class="html-italic">MuTran</span>.</p>
Full article ">Figure 3
<p>The structure and operation flow of <span class="html-italic">MuTran</span>.</p>
Full article ">Figure 4
<p>Pointers related to segment in kernel.</p>
Full article ">Figure 5
<p>The ratio of delay when using <span class="html-italic">MuTran</span> in TCP-mode to that when using the existing system.</p>
Full article ">Figure 6
<p>Periodical transmissions of control packet in the UDP mode.</p>
Full article ">Figure 7
<p>Analysis of the delay when using <span class="html-italic">MuTran</span> in UDP-mode.</p>
Full article ">Figure 8
<p>The analytic delay results in the practical cases.</p>
Full article ">Figure 8 Cont.
<p>The analytic delay results in the practical cases.</p>
Full article ">Figure 9
<p>The experiment setup for evaluating the existing system in the problem situation.</p>
Full article ">Figure 10
<p>The limitation of existing system in the problem situation.</p>
Full article ">Figure 11
<p>The experiment setup for evaluating packet transmissions in a round-robin fashion. (In order to use MPTCP, it is required that the routing tables in the sender and receiver are configured appropriately and the multiple interfaces are connected to the receiver’s board via a router [<a href="#B30-sensors-21-03295" class="html-bibr">30</a>].)</p>
Full article ">Figure 12
<p>The round-robin scheduler’s problem in a low-quality communication situation.</p>
Full article ">Figure 13
<p>The experiment setup for evaluating the performance of <span class="html-italic">MuTran</span></p>
Full article ">Figure 14
<p>Performance comparison among <span class="html-italic">MuTran</span> and existing systems.</p>
Full article ">Figure 15
<p>The comparison between the related techniques and <span class="html-italic">MuTran</span>. (<b>a</b>) The goodput performances of the data packet transmission of the related techniques [<a href="#B14-sensors-21-03295" class="html-bibr">14</a>,<a href="#B16-sensors-21-03295" class="html-bibr">16</a>] and <span class="html-italic">MuTran</span>. (<b>b</b>) Arrival moments of control packets in a low-quality communication situation when using the related technique [<a href="#B15-sensors-21-03295" class="html-bibr">15</a>] and <span class="html-italic">MuTran</span>.</p>
Full article ">
26 pages, 54885 KiB  
Article
Fatigue Crack Monitoring of T-Type Joints in Steel Offshore Oil and Gas Jacket Platform
by Liaqat Ali, Sikandar Khan, Salem Bashmal, Naveed Iqbal, Weishun Dai and Yong Bai
Sensors 2021, 21(9), 3294; https://doi.org/10.3390/s21093294 - 10 May 2021
Cited by 22 | Viewed by 7157
Abstract
Several approaches have been used in the past to predict fatigue crack growth rates in T-joints of the offshore structures, but there are relatively few cases of applying structural health monitoring during the non-destructive testing of jacket platforms. This paper presents an experimental [...] Read more.
Several approaches have been used in the past to predict fatigue crack growth rates in T-joints of the offshore structures, but there are relatively few cases of applying structural health monitoring during the non-destructive testing of jacket platforms. This paper presents an experimental method based on the sensing of the piezoelectric sensors and finite element analysis method for studying the fatigue cracks in the offshore steel jacket structure. Three types of joints are selected in the current research work: T-type plate, T-type tube-plate, and T-type tube joints. The finite element analysis model established in the current study computes and analyzes the high stress and high strain regions in the T-type joints. The fatigue damage in the T-type joints was successfully detected by utilizing both the finite element analysis and experimental methods. The results showed that fatigue cracks of the three types of joints are prone to appear at the weld toe and spread in the welding direction. The fatigue damage location of T-type plate and T-type tube-plate joints is more concentrated in the upper weld toe area, and the fatigue damage location of the T-type tube joint is closer to the lower weld toe area. Full article
(This article belongs to the Special Issue Sensors for Structural Damage Identification)
Show Figures

Figure 1

Figure 1
<p>(<b>a</b>) Schematic diagram of JZ25-1S jacket; (<b>b</b>) real platform.</p>
Full article ">Figure 2
<p>Geometry of the three types of specimens used in laboratory measurements. (<b>a</b>) T-type plate joint. (<b>b</b>) T-type tube-plate joint. (<b>c</b>) T-type tube joint.</p>
Full article ">Figure 3
<p>Supports and the points and directions of the testing force (<b>a</b>) for static structures (<b>b</b>) for pipe plate fatigue (<b>c</b>) for same size sample ratio.</p>
Full article ">Figure 4
<p>Experimental setup.</p>
Full article ">Figure 5
<p>PZT transducers mounted on the specimens.</p>
Full article ">Figure 6
<p>Cracks locations on T-type welded metallic plate (<b>a</b>) at upper welding toe (<b>b</b>) near the edge corner (<b>c</b>) on the weld toe (<b>d</b>) near side edge corner.</p>
Full article ">Figure 6 Cont.
<p>Cracks locations on T-type welded metallic plate (<b>a</b>) at upper welding toe (<b>b</b>) near the edge corner (<b>c</b>) on the weld toe (<b>d</b>) near side edge corner.</p>
Full article ">Figure 7
<p>Microscopic observation of fatigue crack in T-type plate joint (<b>a</b>) cracks initiated at edges of the base metal (<b>b</b>) observed crack of 0.463 mm length (<b>c</b>) observed crack of 0.687 mm length (<b>d</b>) observed cracks of 0.561 mm and 1.285 mm lengths.</p>
Full article ">Figure 7 Cont.
<p>Microscopic observation of fatigue crack in T-type plate joint (<b>a</b>) cracks initiated at edges of the base metal (<b>b</b>) observed crack of 0.463 mm length (<b>c</b>) observed crack of 0.687 mm length (<b>d</b>) observed cracks of 0.561 mm and 1.285 mm lengths.</p>
Full article ">Figure 8
<p>Cracks of T-type tube-plate joint (<b>a</b>) observed crack of 0.569 mm length (<b>b</b>) observed crack of 0.929 mm length (<b>c</b>) observed crack of 2.424 mm length (<b>d</b>) large crack initiated at edges of the base metal.</p>
Full article ">Figure 9
<p>Microscopic observation of fatigue crack in T-type tube-plate joint (<b>a</b>) observed crack of 0.559 mm length (<b>b</b>) observed crack of 0.929 mm length (<b>c</b>) observed crack of 2.424 mm length (<b>d</b>) cracks initiated at edges of the base metal.</p>
Full article ">Figure 10
<p>Cracks of T-type tube joint (<b>a</b>) observed crack of 0.864 mm length (<b>b</b>) observed crack of 0.970 mm length.</p>
Full article ">Figure 11
<p>Microscopic observation of fatigue crack in T-type tube joint (<b>a</b>) observed crack of 0.864 mm length (<b>b</b>) cracks initiated at edges of the base metal.</p>
Full article ">Figure 12
<p>Finite element models of three typical joints (<b>a</b>) T-type plate joint (<b>b</b>) T-type tube-plate joint (<b>c</b>) T-type tube joint.</p>
Full article ">Figure 13
<p>Boundary condition of each joint (<b>a</b>) T-type plate joint (<b>b</b>) T-type tube-plate joint (<b>c</b>) T-type tube joint.</p>
Full article ">Figure 13 Cont.
<p>Boundary condition of each joint (<b>a</b>) T-type plate joint (<b>b</b>) T-type tube-plate joint (<b>c</b>) T-type tube joint.</p>
Full article ">Figure 14
<p>S-N curves for three typical joints. (<b>a</b>) Curve E for T-type plate joint. (<b>b</b>) Curve G for T-type tube-plate joint. (<b>c</b>) Curve F3 for T-type tube joint.</p>
Full article ">Figure 15
<p>Comparison of fatigue damage in T-type plate joint (<b>a</b>) Finite element result (<b>b</b>) Test result.</p>
Full article ">Figure 16
<p>Comparison of fatigue damage in T-type tube-plate joint (<b>a</b>) Test result (<b>b</b>) Finite element result.</p>
Full article ">Figure 17
<p>Comparison of fatigue damage in T-type tube joint (<b>a</b>) Test result (<b>b</b>) Finite element result.</p>
Full article ">
21 pages, 7855 KiB  
Article
A Low-Cost IoT System for Real-Time Monitoring of Climatic Variables and Photovoltaic Generation for Smart Grid Application
by Gustavo Costa Gomes de Melo, Igor Cavalcante Torres, Ícaro Bezzera Queiroz de Araújo, Davi Bibiano Brito and Erick de Andrade Barboza
Sensors 2021, 21(9), 3293; https://doi.org/10.3390/s21093293 - 10 May 2021
Cited by 42 | Viewed by 6194
Abstract
Monitoring and data acquisition are essential to recognize the renewable resources available on-site, evaluate electrical conversion efficiency, detect failures, and optimize electrical production. Commercial monitoring systems for the photovoltaic system are generally expensive and closed for modifications. This work proposes a low-cost real-time [...] Read more.
Monitoring and data acquisition are essential to recognize the renewable resources available on-site, evaluate electrical conversion efficiency, detect failures, and optimize electrical production. Commercial monitoring systems for the photovoltaic system are generally expensive and closed for modifications. This work proposes a low-cost real-time internet of things system for micro and mini photovoltaic generation systems that can monitor continuous voltage, continuous current, alternating power, and seven meteorological variables. The proposed system measures all relevant meteorological variables and directly acquires photovoltaic generation data from the plant (not from the inverter). The system is implemented using open software, connects to the internet without cables, stores data locally and in the cloud, and uses the network time protocol to synchronize the devices’ clocks. To the best of our knowledge, no work reported in the literature presents these features altogether. Furthermore, experiments carried out with the proposed system showed good effectiveness and reliability. This system enables fog and cloud computing in a photovoltaic system, creating a time series measurements data set, enabling the future use of machine learning to create smart photovoltaic systems. Full article
(This article belongs to the Special Issue Smart IoT System for Renewable Energy Resource)
Show Figures

Figure 1

Figure 1
<p>Simplified diagram of the proposed system.</p>
Full article ">Figure 2
<p>Simplified diagram of the solarimetric station data logger, with emphasis on the components and connections.</p>
Full article ">Figure 3
<p>Simplified diagram of the PV generation data logger, with emphasis on the components and connections.</p>
Full article ">Figure 4
<p>Simplified diagram representing the operation of the data logger devices.</p>
Full article ">Figure 5
<p>Proposed LoRA payload structure.</p>
Full article ">Figure 6
<p>Diagram representing the IoT architecture.</p>
Full article ">Figure 7
<p>Web application home page, displaying the drop-down sub-menu.</p>
Full article ">Figure 8
<p>Web application page for real-time monitoring of the solarimetric station.</p>
Full article ">Figure 9
<p>Web application page to consult the history of PV generation data.</p>
Full article ">Figure 10
<p>Proposed system installed in a PV plant. (<b>a</b>) Solarimetric station, with emphasis on its data logger and sensors. (<b>b</b>) Cabinet with the transducers and the PV generation data logger.</p>
Full article ">Figure 11
<p>Graphical comparison of the data obtained during one week (1 March to 7 March 2021) by our proposed system (blue) and the CR1000 (red). The graphs show the following measurements: (<b>a</b>) ambient temperature, (<b>b</b>) PV module temperature, (<b>c</b>) irradiance, (<b>d</b>) AC power.</p>
Full article ">Figure 12
<p>Graphical comparison of the data obtained during one week (1 March to 7 March 2021) by our proposed system (blue) and the CR1000 (red). The graphs show the following measurements: (<b>a</b>) string 1 current, (<b>b</b>) string 2 current, (<b>c</b>) string 1 voltage, (<b>d</b>) string 2 voltage.</p>
Full article ">
16 pages, 7139 KiB  
Article
Horizontal-to-Vertical Spectral Ratio of Ambient Vibration Obtained with Hilbert–Huang Transform
by Maik Neukirch, Antonio García-Jerez, Antonio Villaseñor, Francisco Luzón, Mario Ruiz and Luis Molina
Sensors 2021, 21(9), 3292; https://doi.org/10.3390/s21093292 - 10 May 2021
Cited by 5 | Viewed by 3254
Abstract
The Horizontal-to-Vertical Spectral Ratio (HVSR) of ambient vibration measurements is a common tool to explore near surface shear wave velocity (Vs) structure. HVSR is often applied for earthquake risk assessments and civil engineering projects. Ambient vibration signal originates from the combination of a [...] Read more.
The Horizontal-to-Vertical Spectral Ratio (HVSR) of ambient vibration measurements is a common tool to explore near surface shear wave velocity (Vs) structure. HVSR is often applied for earthquake risk assessments and civil engineering projects. Ambient vibration signal originates from the combination of a multitude of natural and man-made sources. Ambient vibration sources can be any ground motion inducing phenomena, e.g., ocean waves, wind, industrial activity or road traffic, where each source does not need to be strictly stationary even during short times. Typically, the Fast Fourier Transform (FFT) is applied to obtain spectral information from the measured time series in order to estimate the HVSR, even though possible non-stationarity may bias the spectra and HVSR estimates. This problem can be alleviated by employing the Hilbert–Huang Transform (HHT) instead of FFT. Comparing 1D inversion results for FFT and HHT-based HVSR estimates from data measured at a well studied, urban, permanent station, we find that HHT-based inversion models may yield a lower data misfit χ2 by up to a factor of 25, a more appropriate Vs model according to available well-log lithology, and higher confidence in the achieved model. Full article
(This article belongs to the Special Issue Data Acquisition and Analysis of Seismic Noise)
Show Figures

Graphical abstract

Graphical abstract
Full article ">Figure 1
<p>Basic Empirical Mode Decomposition. Note that the sifting stop criteria (point 5) above is given in its original form and various alternatives have been discussed in the literature [<a href="#B27-sensors-21-03292" class="html-bibr">27</a>]. However, the exact formulation of the stopping criteria for the sifting process (points 3 to 5) is not central to our work as the EMD algorithm performs with any chosen criteria.</p>
Full article ">Figure 2
<p>ICJA results for FFT- and MEMD-based processing.</p>
Full article ">Figure 3
<p>Weighted distribution of tested ICJA inversion models’ <math display="inline"><semantics> <msup> <mi>χ</mi> <mn>2</mn> </msup> </semantics></math> for FFT and MEMD curves.</p>
Full article ">Figure 4
<p>Well-log, models and data for ICJA station obtained with FFT. Well-log column taken from [<a href="#B35-sensors-21-03292" class="html-bibr">35</a>].</p>
Full article ">Figure 5
<p>Well-log, models and data for ICJA station obtained with MEMD. Well-log column taken from [<a href="#B35-sensors-21-03292" class="html-bibr">35</a>].</p>
Full article ">Figure 6
<p>EJDN results for FFT- and MEMD-based processing.</p>
Full article ">Figure 7
<p>Weighted distribution of tested EJDN inversion models’ <math display="inline"><semantics> <msup> <mi>χ</mi> <mn>2</mn> </msup> </semantics></math> for FFT and MEMD curves.</p>
Full article ">Figure 8
<p>Model results and data fit for the inversion of EJDN data processed by MEMD and FFT.</p>
Full article ">
27 pages, 2499 KiB  
Review
Structure–Function Relationships of Nanocarbon/Polymer Composites for Chemiresistive Sensing: A Review
by Maryam Ehsani, Parvaneh Rahimi and Yvonne Joseph
Sensors 2021, 21(9), 3291; https://doi.org/10.3390/s21093291 - 10 May 2021
Cited by 29 | Viewed by 4700
Abstract
Composites of organic compounds and inorganic nanomaterials provide novel sensing platforms for high-performance sensor applications. The combination of the attractive functionalities of nanomaterials with polymers as an organic matrix offers promising materials with tunable electrical, mechanical, and chemisensitive properties. This review mainly focuses [...] Read more.
Composites of organic compounds and inorganic nanomaterials provide novel sensing platforms for high-performance sensor applications. The combination of the attractive functionalities of nanomaterials with polymers as an organic matrix offers promising materials with tunable electrical, mechanical, and chemisensitive properties. This review mainly focuses on nanocarbon/polymer composites as chemiresistors. We first describe the structure and properties of carbon nanofillers as reinforcement agents used in the manufacture of polymer composites and the sensing mechanism of developed nanocomposites as chemiresistors. Then, the design and synthesizing methods of polymer composites based on carbon nanofillers are discussed. The electrical conductivity, mechanical properties, and the applications of different nanocarbon/polymer composites for the detection of different analytes are reviewed. Lastly, challenges and the future vision for applications of such nanocomposites are described. Full article
(This article belongs to the Special Issue Chemiresistive Sensors: Materials and Applications)
Show Figures

Figure 1

Figure 1
<p>Various types of nanofillers for polymer nanocomposites.</p>
Full article ">Figure 2
<p>Schematic representation of chemiresistor mechanism based on swelling effect. The red dashed line gives the electrical current along the percolation pathway (<b>A</b>) before and (<b>B</b>) after swelling.</p>
Full article ">Figure 3
<p>Schematic representation of solution/emulsion processing method.</p>
Full article ">Figure 4
<p>Scheme representation of synthesis of nanocarbon/polymer nanocomposite using self-assembly process.</p>
Full article ">Figure 5
<p>Schematic representation of in situ technique.</p>
Full article ">Figure 6
<p>SEM images of PS nanocomposites with (<b>a</b>) 1.5 wt.% CNT and (<b>b</b>) 1.5 wt.% G. TEM images of PS nanocomposites with (<b>c</b>) 1.5 wt.% CNT and (<b>d</b>) 1.5 wt.% G [<a href="#B80-sensors-21-03291" class="html-bibr">80</a>], reprinted with permission from American Chemical Society.</p>
Full article ">Figure 7
<p>SEM images of rGO (RGO)/PEDOT film and conduction pathway [<a href="#B92-sensors-21-03291" class="html-bibr">92</a>], reprinted with permission from Elsevier.</p>
Full article ">Figure 8
<p>(<b>a</b>) SEM images of agglomerated edge-functionalized GNR and (<b>b</b>) TEM image of nanocomposite loaded by 0.15 wt.% edge-functionalized GNRs (white arrows mark the individually dispersed GNRs) [<a href="#B95-sensors-21-03291" class="html-bibr">95</a>], reprinted with permission from Elsevier.</p>
Full article ">Figure 9
<p>SEM micrographs (<b>a</b>) 0.5 wt.% SWCNT/iPP and (<b>b</b>) 1.0 wt.% SWCNT/iPP composites [<a href="#B100-sensors-21-03291" class="html-bibr">100</a>], reprinted with permission from Elsevier.</p>
Full article ">Figure 10
<p>TEM images of (<b>a</b>,<b>a</b>’) PANI and (<b>b</b>,<b>b</b>’) FLN/PANI [<a href="#B118-sensors-21-03291" class="html-bibr">118</a>], reprinted with permission from Elsevier.</p>
Full article ">
18 pages, 4413 KiB  
Article
A New Stochastic Model Updating Method Based on Improved Cross-Model Cross-Mode Technique
by Hui Chen, Bin Huang, Kong Fah Tee and Bo Lu
Sensors 2021, 21(9), 3290; https://doi.org/10.3390/s21093290 - 10 May 2021
Cited by 3 | Viewed by 2486
Abstract
This paper proposes a new stochastic model updating method to update structural models based on the improved cross-model cross-mode (ICMCM) technique. This new method combines the stochastic hybrid perturbation-Galerkin method with the ICMCM method to solve the model updating problems with limited measurement [...] Read more.
This paper proposes a new stochastic model updating method to update structural models based on the improved cross-model cross-mode (ICMCM) technique. This new method combines the stochastic hybrid perturbation-Galerkin method with the ICMCM method to solve the model updating problems with limited measurement data and uncertain measurement errors. First, using the ICMCM technique, a new stochastic model updating equation with an updated coefficient vector is established by considering the uncertain measured modal data. Then, the stochastic model updating equation is solved by the stochastic hybrid perturbation-Galerkin method so as to obtain the random updated coefficient vector. Following that, the statistical characteristics of the updated coefficients can be determined. Numerical results of a continuous beam show that the proposed method can effectively cope with relatively large uncertainty in measured data, and the computational efficiency of this new method is several orders of magnitude higher than that of the Monte Carlo simulation method. When considering the rank deficiency, the proposed stochastic ICMCM method can achieve more accurate updating results compared with the cross-model cross-mode (CMCM) method. An experimental example shows that the new method can effectively update the structural stiffness and mass, and the statistics of the frequencies of the updated model are consistent with the measured results, which ensures that the updated coefficients are of practical significance. Full article
(This article belongs to the Special Issue Structural Health Monitoring for Smart Structures)
Show Figures

Figure 1

Figure 1
<p>Flowchart of stochastic model updating by means of the HPG-ICMCM method.</p>
Full article ">Figure 2
<p>A two-span continuous beam.</p>
Full article ">Figure 3
<p>The probability density functions of the 24 updated parameters.</p>
Full article ">Figure 3 Cont.
<p>The probability density functions of the 24 updated parameters.</p>
Full article ">Figure 4
<p>PDFs of the first five updated frequencies.</p>
Full article ">Figure 5
<p>PDFs of the updated parameters of all elements in the three cases.</p>
Full article ">Figure 5 Cont.
<p>PDFs of the updated parameters of all elements in the three cases.</p>
Full article ">Figure 6
<p>A steel cantilever beam.</p>
Full article ">Figure 7
<p>Means of the first four order measured modal shapes.</p>
Full article ">Figure 8
<p>The means of the updated coefficients of the 12 updated parameters including the six elastic moduli and the six mass values.</p>
Full article ">Figure 9
<p>The standard deviations of the updated coefficients of the 12 updated parameters including the six elastic moduli and the six mass values.</p>
Full article ">Figure 10
<p>PDFs of the first three updated frequencies.</p>
Full article ">
16 pages, 7256 KiB  
Article
Deep Supervised Residual Dense Network for Underwater Image Enhancement
by Yanling Han, Lihua Huang, Zhonghua Hong, Shouqi Cao, Yun Zhang and Jing Wang
Sensors 2021, 21(9), 3289; https://doi.org/10.3390/s21093289 - 10 May 2021
Cited by 28 | Viewed by 3648
Abstract
Underwater images are important carriers and forms of underwater information, playing a vital role in exploring and utilizing marine resources. However, underwater images have characteristics of low contrast and blurred details because of the absorption and scattering of light. In recent years, deep [...] Read more.
Underwater images are important carriers and forms of underwater information, playing a vital role in exploring and utilizing marine resources. However, underwater images have characteristics of low contrast and blurred details because of the absorption and scattering of light. In recent years, deep learning has been widely used in underwater image enhancement and restoration because of its powerful feature learning capabilities, but there are still shortcomings in detailed enhancement. To address the problem, this paper proposes a deep supervised residual dense network (DS_RD_Net), which is used to better learn the mapping relationship between clear in-air images and synthetic underwater degraded images. DS_RD_Net first uses residual dense blocks to extract features to enhance feature utilization; then, it adds residual path blocks between the encoder and decoder to reduce the semantic differences between the low-level features and high-level features; finally, it employs a deep supervision mechanism to guide network training to improve gradient propagation. Experiments results (PSNR was 36.2, SSIM was 96.5%, and UCIQE was 0.53) demonstrated that the proposed method can fully retain the local details of the image while performing color restoration and defogging compared with other image enhancement methods, achieving good qualitative and quantitative effects. Full article
(This article belongs to the Special Issue Image Sensing and Processing with Convolutional Neural Networks)
Show Figures

Figure 1

Figure 1
<p>The proposed general framework for underwater image enhancement (three major parts: synthetic model of underwater images based on UWGAN, underwater image enhancement network based on DS_RD_Net, and evaluation of underwater image enhancement methods).</p>
Full article ">Figure 2
<p>Physical model of underwater imaging.</p>
Full article ">Figure 3
<p>UWGAN architecture. UWGAN takes in-air images and its depth maps as input; then, it synthesizes underwater degraded images on the basis of underwater optical imaging model by generative adversarial training.</p>
Full article ">Figure 4
<p>DS_RD_Net architecture. DS_RD_Net adds residual dense blocks, residual path blocks, and a deep supervision mechanism to learn the mapping relationship between clear in-air images and synthetic underwater degraded images.</p>
Full article ">Figure 5
<p>Details of several blocks: (<b>a</b>) residual dense encoder block; (<b>b</b>) residual dense decoder block; (<b>c</b>) residual path block.</p>
Full article ">Figure 6
<p>Qualitative comparisons for samples from synthetic underwater datasets. (<b>a</b>) Synthetic underwater degraded images. (<b>b</b>) Results of UCM. (<b>c</b>) Results of UDCP. (<b>d</b>) Results of Unet3. (<b>e</b>) Results of UGAN. (<b>f</b>) Results of FunieGAN. (<b>g</b>) Our results. (<b>h</b>) Ground truth.</p>
Full article ">Figure 7
<p>Qualitative comparisons for samples from real underwater datasets. (<b>a</b>) Real underwater images. (<b>b</b>) Results of UCM. (<b>c</b>) Results of UDCP. (<b>d</b>) Results of Unet3. (<b>e</b>) Results of UGAN. (<b>f</b>) Results of FunieGAN. (<b>g</b>) Our results.</p>
Full article ">Figure 8
<p>Underwater object detection results before and after enhancement. (<b>a</b>) Results with labels. (<b>b</b>) Results before enhancement. (<b>c</b>) Results after enhancement. Red boxes represent holothurians, blue boxes represent starfishes, yellow boxes represent echinus, and green boxes represent scallops.</p>
Full article ">
12 pages, 21757 KiB  
Communication
Tilted-Beam Antenna Based on SSPPs-TL with Stable Gain
by Dujuan Wei, Youlin Geng, Pengquan Zhang, Zhonghai Zhang and Chuan Yin
Sensors 2021, 21(9), 3288; https://doi.org/10.3390/s21093288 - 10 May 2021
Cited by 1 | Viewed by 3078
Abstract
In this paper, a titled-beam antenna based on spoof surface plasmon polaritons (SSPPs) transmission lines (TLs) is proposed. The parallel SSPPs-TL is a slow-wave TL, which is able to limit waves in the TL strictly. By periodically introducing a set of tapered stubs [...] Read more.
In this paper, a titled-beam antenna based on spoof surface plasmon polaritons (SSPPs) transmission lines (TLs) is proposed. The parallel SSPPs-TL is a slow-wave TL, which is able to limit waves in the TL strictly. By periodically introducing a set of tapered stubs along the SSPPs-TL, the backward endfire beams are formed by the surface waves in the slow-wave radiation region. Then, through the placement of a big metal plate below the endfire antenna, the backward endfire beams are tilted, and the tilted angle of the beams are steered by the distance of the metal plate and antenna. Over the band of 5.7 GHz~7.0 GHz, the tilted antenna performs constant shapes of radiation patterns. The gain keeps stable at around 12 dBi and the 1-dB gain bandwidth is 20%. The measured results of the fabricated prototypes confirm the design theory and simulated results. Full article
(This article belongs to the Section Electronic Sensors)
Show Figures

Figure 1

Figure 1
<p>The profile of the parallel SSPPs-TL. (<b>a</b>) the full view, (<b>b</b>) a unit cell, (<b>c</b>) a transition section. <span class="html-italic">d</span> = 6 mm, <span class="html-italic">a</span> = 3 mm, <span class="html-italic">w</span> = 2.7 mm, <span class="html-italic">h</span> = 4 mm, <span class="html-italic">H</span> = 10.7 mm, <span class="html-italic">l</span><sub>1</sub> = 8.7 mm, <span class="html-italic">l</span><sub>2</sub> = 6.7 mm, <span class="html-italic">l</span><sub>3</sub> = 4.7 mm, <span class="html-italic">w</span><sub>1</sub> = 4.7 mm, <span class="html-italic">w</span><sub>2</sub> = 6.7 mm, <span class="html-italic">w</span><sub>3</sub> = 8.</p>
Full article ">Figure 2
<p>Dispersion diagram of the parallel SSPPs-TL.</p>
Full article ">Figure 3
<p>S parameters of the parallel SSPPs-TL with <span class="html-italic">h</span> = 4 mm.</p>
Full article ">Figure 4
<p>The structure of the proposed antenna with uniform stubs. <span class="html-italic">p</span> = 12 mm, <span class="html-italic">h</span> = 4 mm, <span class="html-italic">b</span> = 3 mm, <span class="html-italic">h<sub>u</sub></span> = 6 mm.</p>
Full article ">Figure 5
<p>Normalized phase constant.</p>
Full article ">Figure 6
<p>Radiation patterns of the antenna.</p>
Full article ">Figure 7
<p>Electric field distribution of the antenna in <a href="#sensors-21-03288-f004" class="html-fig">Figure 4</a> at 6.6 GHz in YOZ plane.</p>
Full article ">Figure 8
<p>The profile of the proposed antenna. (<b>a</b>) top view, (<b>b</b>) full view. <span class="html-italic">w<sub>g</sub></span> = 62 mm, <span class="html-italic">l<sub>g</sub></span> = 130 mm, <span class="html-italic">L</span> = 100 mm, <span class="html-italic">h<sub>g</sub></span> = 8 mm, <span class="html-italic">h</span><sub>1</sub> = 6 mm, <span class="html-italic">h</span><sub>2</sub> = 7.5 mm, <span class="html-italic">h</span><sub>3</sub> = 8.5 mm, <span class="html-italic">h</span><sub>4</sub> = 9.5 mm, <span class="html-italic">h</span><sub>5</sub> = 8.5 mm, <span class="html-italic">h</span><sub>6</sub> = 7.5 mm.</p>
Full article ">Figure 9
<p>Normalized radiation patterns of the proposed antenna without metal plate in H-plane (YOZ plane).</p>
Full article ">Figure 10
<p>Radiation patterns of the proposed antenna with metal plate at 6.4 GHz as <span class="html-italic">h<sub>g</sub></span> varies.</p>
Full article ">Figure 11
<p>Electric field distribution of the proposed antenna in YOZ plane. (<b>a</b>) without the metal plate, (<b>b</b>) with the metal plate (<span class="html-italic">h<sub>g</sub></span> = 8 mm).</p>
Full article ">Figure 12
<p>Theory of tilted-beam formation (<span class="html-italic">f</span> = 5.7 GHz, <span class="html-italic">h<sub>g</sub></span> = 8 mm).</p>
Full article ">Figure 13
<p>Reflective coefficients and gain of the proposed with and without plate.</p>
Full article ">Figure 14
<p>Photos of the proposed antenna. (<b>a</b>) top and bottom surfaces, (<b>b</b>) full view.</p>
Full article ">Figure 15
<p>Reflective coefficients and gain of the proposed antenna.</p>
Full article ">Figure 16
<p>Radiation patterns of the proposed antenna in H-plane (YOZ plane). (<b>a</b>) Simulated results, (<b>b</b>) measured results.</p>
Full article ">Figure 17
<p>Simulated radiation patterns of the proposed antenna in YOZ plane.</p>
Full article ">
23 pages, 635 KiB  
Article
An Appliance Scheduling System for Residential Energy Management
by Hanife Apaydin-Özkan
Sensors 2021, 21(9), 3287; https://doi.org/10.3390/s21093287 - 10 May 2021
Cited by 7 | Viewed by 3393
Abstract
In this work, an Appliance Scheduling-based Residential Energy Management System (AS-REMS) for reducing electricity cost and avoiding peak demand while keeping user comfort is presented. In AS-REMS, based on the effects of starting times of appliances on user comfort and the user attendance [...] Read more.
In this work, an Appliance Scheduling-based Residential Energy Management System (AS-REMS) for reducing electricity cost and avoiding peak demand while keeping user comfort is presented. In AS-REMS, based on the effects of starting times of appliances on user comfort and the user attendance during their operations, appliances are divided into two classes in terms of controllability: MC-controllable (allowed to be scheduled by the Main Controller) and user-controllable (allowed to be scheduled only by a user). Use of all appliances are monitored in the considered home for a while for recording users’ appliance usage preferences and habits on each day of the week. Then, for each MC-controllable appliance, preferred starting times are determined and prioritized according to the recorded user preferences on similar days. When scheduling, assigned priorities of starting times of these appliances are considered for maintaining user comfort, while the tariff rate is considered for reducing electricity cost. Moreover, expected power consumptions of user-controllable appliances corresponding to the recorded user habits and power consumptions of MC-controllable appliances corresponding to the assigned starting times are considered for avoiding peak demand. The corresponding scheduling problem is solved by Brute-Force Closest Pair method. AS-REMS reduces the peak demand levels by 45% and the electricity costs by 39.6%, while provides the highest level of user comfort by 88%. Thus, users’ appliance usage preferences are sustained at a lower cost while their comfort is kept impressively. Full article
(This article belongs to the Special Issue Smart Sensor Networks for Smart Grids)
Show Figures

Figure 1

Figure 1
<p>AS-REMS Structure.</p>
Full article ">Figure 2
<p>Power consumption profile of a kettle (1000 W and 1.2 lt capacity).</p>
Full article ">Figure 3
<p>Power consumption profile of an air conditioner (6.74 kW cooling and 7.03 kW heating capacity).</p>
Full article ">Figure 4
<p>Power consumption profile of a washing machine (7 kg front-load).</p>
Full article ">Figure 5
<p>An example application of the BFCP method.</p>
Full article ">Figure 6
<p>Three-axis chart for possible triplets of (<math display="inline"><semantics> <mrow> <msubsup> <mi>t</mi> <mi>s</mi> <mi mathvariant="italic">wm</mi> </msubsup> <mo>,</mo> <mspace width="3.33333pt"/> <msubsup> <mi>t</mi> <mi>s</mi> <mi mathvariant="italic">dw</mi> </msubsup> <mo>,</mo> <mspace width="3.33333pt"/> <msubsup> <mi>t</mi> <mi>s</mi> <mi mathvariant="italic">bp</mi> </msubsup> </mrow> </semantics></math>) for Scenario 1.</p>
Full article ">Figure 7
<p>Power consumption graph of the minimum cost solution for Scenario 1.</p>
Full article ">Figure 8
<p>Power consumption graph of the highest priority solution for Scenario 1.</p>
Full article ">Figure 9
<p>Power consumption graph of the optimal solution for Scenario 1.</p>
Full article ">Figure 10
<p>Three-axis chart for possible triplets of (<math display="inline"><semantics> <mrow> <msubsup> <mi>t</mi> <mi>s</mi> <mi mathvariant="italic">wm</mi> </msubsup> <mo>,</mo> <mspace width="3.33333pt"/> <msubsup> <mi>t</mi> <mi>s</mi> <mi mathvariant="italic">dw</mi> </msubsup> <mo>,</mo> <mspace width="3.33333pt"/> <msubsup> <mi>t</mi> <mi>s</mi> <mi mathvariant="italic">bp</mi> </msubsup> </mrow> </semantics></math>) for Scenario 2.</p>
Full article ">Figure 11
<p>Power consumption graph of minimum cost (<b>a</b>) and optimal (<b>b</b>) solution for Scenario 2.</p>
Full article ">Figure 12
<p>Power consumption graph of both without AS-REMS (<b>a</b>) and optimal (<b>b</b>) solution for Scenario 3.</p>
Full article ">
14 pages, 549 KiB  
Article
Localization and Tracking of an Indoor Autonomous Vehicle Based on the Phase Difference of Passive UHF RFID Signals
by Yunlei Zhang, Xiaolin Gong, Kaihua Liu and Shuai Zhang
Sensors 2021, 21(9), 3286; https://doi.org/10.3390/s21093286 - 10 May 2021
Cited by 16 | Viewed by 4375
Abstract
State-of-the-art radio frequency identification (RFID)-based indoor autonomous vehicles localization methods are mostly based on received signal strength indicator (RSSI) measurements. However, the accuracy of these methods is not high enough for real-world scenarios. To overcome this problem, a novel dual-frequency phase difference of [...] Read more.
State-of-the-art radio frequency identification (RFID)-based indoor autonomous vehicles localization methods are mostly based on received signal strength indicator (RSSI) measurements. However, the accuracy of these methods is not high enough for real-world scenarios. To overcome this problem, a novel dual-frequency phase difference of arrival (PDOA) ranging-based indoor autonomous vehicle localization and tracking scheme was developed. Firstly, the method gets the distance between the RFID reader and the tag by dual-frequency PDOA ranging. Then, a maximum likelihood estimation and semi-definite programming (SDP)-based localization algorithm is utilized to calculate the position of the autonomous vehicles, which can mitigate the multipath ranging error and obtain a more accurate positioning result. Finally, vehicle traveling information and the position achieved by RFID localization are fused with a Kalman filter (KF). The proposed method can work in a low-density tag deployment environment. Simulation experiment results showed that the proposed vehicle localization and tracking method achieves centimeter-level mean tracking accuracy. Full article
(This article belongs to the Special Issue Internet of Things for Industrial Applications)
Show Figures

Figure 1

Figure 1
<p>Schematic of the UHF RFID-based indoor autonomous localization and tracking system. The UHF RFID reader is placed on the vehicle, and tags are placed on the floor.</p>
Full article ">Figure 2
<p>The vehicle state transition of two adjacent points on the trajectory.</p>
Full article ">Figure 3
<p>Multipath propagation scenario of the indoor autonomous vehicle navigation system.</p>
Full article ">Figure 4
<p>Phase diagram for narrowband signaling propagation on a multipath channel.</p>
Full article ">Figure 5
<p>The localization and tracking results when RFID reader could read four neighboring tag. (<b>a</b>) The localization and tracking results of the line trajectory. (<b>b</b>) The localization and tracking results of the circular trajectory.</p>
Full article ">Figure 6
<p>The CDF curves of the circle trajectory tracking error with different numbers of tags.</p>
Full article ">Figure 7
<p>The CDF curves of the circle trajectory tracking and localization error with the RSS and phase.</p>
Full article ">
19 pages, 1616 KiB  
Article
Linear Matrix Inequalities for an Iterative Solution of Robust Output Feedback Control of Systems with Bounded and Stochastic Uncertainty
by Andreas Rauh and Swantje Romig
Sensors 2021, 21(9), 3285; https://doi.org/10.3390/s21093285 - 10 May 2021
Cited by 9 | Viewed by 3537
Abstract
Linear matrix inequalities (LMIs) have gained much importance in recent years for the design of robust controllers for linear dynamic systems, for the design of state observers, as well as for the optimization of both. Typical performance criteria that are considered in these [...] Read more.
Linear matrix inequalities (LMIs) have gained much importance in recent years for the design of robust controllers for linear dynamic systems, for the design of state observers, as well as for the optimization of both. Typical performance criteria that are considered in these cases are either H2 or H measures. In addition to bounded parameter uncertainty, included in the LMI-based design by means of polytopic uncertainty representations, the recent work of the authors showed that state observers can be optimized with the help of LMIs so that their error dynamics become insensitive against stochastic noise. However, the joint optimization of the parameters of the output feedback controllers of a proportional-differentiating type with a simultaneous optimization of linear output filters for smoothening measurements and for their numeric differentiation has not yet been considered. This is challenging due to the fact that the joint consideration of both types of uncertainties, as well as the combined control and filter optimization lead to a problem that is constrained by nonlinear matrix inequalities. In the current paper, a novel iterative LMI-based procedure is presented for the solution of this optimization task. Finally, an illustrating example is presented to compare the new parameterization scheme for the output feedback controller—which was jointly optimized with a linear derivative estimator—with a heuristically tuned D-type control law of previous work that was implemented with the help of an optimized full-order state observer. Full article
(This article belongs to the Special Issue Modern Control in Theory and Practice)
Show Figures

Figure 1

Figure 1
<p>Observer-based state and output feedback control structure according to Case 1 with the gain matrix <math display="inline"><semantics> <mi mathvariant="bold">K</mi> </semantics></math> and Case 2 with the structured gain <math display="inline"><semantics> <mrow> <msub> <mi mathvariant="bold">K</mi> <mi mathvariant="normal">o</mi> </msub> <mi mathvariant="bold">C</mi> </mrow> </semantics></math>.</p>
Full article ">Figure 2
<p>Filter-based output feedback control structure according to Case 3.</p>
Full article ">Figure 3
<p>Domain of the eigenvalues compatible with the constraints (<a href="#FD23-sensors-21-03285" class="html-disp-formula">23</a>) and (<a href="#FD24-sensors-21-03285" class="html-disp-formula">24</a>), where <math display="inline"><semantics> <mrow> <mi>θ</mi> <mo>=</mo> <mn>0</mn> </mrow> </semantics></math> is desired to guarantee non-oscillatory dynamics.</p>
Full article ">Figure 4
<p>Structure diagram of the iteration procedure for the proposed filter-based control parameterization.</p>
Full article ">Figure 5
<p>Control performance of the proposed iterative LMI-based optimization technique. (<b>a</b>) Position <math display="inline"><semantics> <msub> <mi>x</mi> <mn>1</mn> </msub> </semantics></math> for the spring-mass-damper system. (<b>b</b>) Control signal <span class="html-italic">u</span> for the spring-mass-damper system.</p>
Full article ">Figure 6
<p>Reconstruction of the states <math display="inline"><semantics> <msub> <mi>x</mi> <mn>1</mn> </msub> </semantics></math> and <math display="inline"><semantics> <msub> <mi>x</mi> <mn>2</mn> </msub> </semantics></math> for the proposed iterative LMI-based filter and control optimization. (<b>a</b>) Reconstruction of the position <math display="inline"><semantics> <msub> <mi>x</mi> <mn>1</mn> </msub> </semantics></math> in comparison with the noisy measurement and the true state evolution (setting from <a href="#sec3dot1-sensors-21-03285" class="html-sec">Section 3.1</a>). (<b>b</b>) Reconstruction of the position <math display="inline"><semantics> <msub> <mi>x</mi> <mn>1</mn> </msub> </semantics></math> in comparison with the noisy measurement and the true state evolution (setting from <a href="#sec3dot2-sensors-21-03285" class="html-sec">Section 3.2</a>). (<b>c</b>) Reconstruction of the velocity <math display="inline"><semantics> <msub> <mi>x</mi> <mn>2</mn> </msub> </semantics></math> in comparison with the true state evolution (setting from <a href="#sec3dot1-sensors-21-03285" class="html-sec">Section 3.1</a>). (<b>d</b>) Reconstruction of the velocity <math display="inline"><semantics> <msub> <mi>x</mi> <mn>2</mn> </msub> </semantics></math> in comparison with the true state evolution (setting from <a href="#sec3dot2-sensors-21-03285" class="html-sec">Section 3.2</a>).</p>
Full article ">Figure 7
<p>Comparison of a heuristic D-type control approach with an optimized observer from the previous work [<a href="#B26-sensors-21-03285" class="html-bibr">26</a>].</p>
Full article ">Figure 8
<p>Comparison of a heuristic D-type control approach with an optimized observer from the previous work [<a href="#B26-sensors-21-03285" class="html-bibr">26</a>] (cont’d).</p>
Full article ">
16 pages, 3695 KiB  
Article
Estimating the Product of the X-ray Spectrum and Quantum Detection Efficiency of a CT System and Its Application to Beam Hardening Correction
by Joseph J. Lifton and Andrew A. Malcolm
Sensors 2021, 21(9), 3284; https://doi.org/10.3390/s21093284 - 10 May 2021
Cited by 3 | Viewed by 3280
Abstract
Lab-based X-ray computed tomography (XCT) systems use X-ray sources that emit a polychromatic X-ray spectrum and detectors that do not detect all X-ray photons with the same efficiency. A consequence of using a polychromatic X-ray source is that beam hardening artefacts may be [...] Read more.
Lab-based X-ray computed tomography (XCT) systems use X-ray sources that emit a polychromatic X-ray spectrum and detectors that do not detect all X-ray photons with the same efficiency. A consequence of using a polychromatic X-ray source is that beam hardening artefacts may be present in the reconstructed data, and the presence of such artefacts can degrade XCT image quality and affect quantitative analysis. If the product of the X-ray spectrum and the quantum detection efficiency (QDE) of the detector are known, alongside the material of the scanned object, then beam hardening artefacts can be corrected algorithmically. In this work, a method for estimating the product of the X-ray spectrum and the detector’s QDE is offered. The method approximates the product of the X-ray spectrum and the QDE as a Bézier curve, which requires only eight fitting parameters to be estimated. It is shown experimentally and through simulation that Bézier curves can be used to accurately simulate polychromatic attenuation and hence be used to correct beam hardening artefacts. The proposed method is tested using measured attenuation data and then used to calculate a beam hardening correction for an aluminium workpiece; the beam hardening correction leads to an increase in the contrast-to-noise ratio of the XCT data by 41% and the removal of cupping artefacts. Deriving beam hardening corrections in this manner is more versatile than using conventional material-specific step wedges. Full article
(This article belongs to the Special Issue Sensors and X-ray Detectors)
Show Figures

Figure 1

Figure 1
<p>A simulated polychromatic X-ray spectrum emitted by a tungsten reflection target X-ray source, obtained using the model in [<a href="#B7-sensors-21-03284" class="html-bibr">7</a>].</p>
Full article ">Figure 2
<p>The mass attenuation coefficient for CsI plotted for the same energy range as the X-ray spectrum shown in <a href="#sensors-21-03284-f001" class="html-fig">Figure 1</a>. Data downloaded from [<a href="#B8-sensors-21-03284" class="html-bibr">8</a>].</p>
Full article ">Figure 3
<p>Quantum detection efficiency for a 600-µm-thick CsI scintillator calculated using Equation (1).</p>
Full article ">Figure 4
<p>Plot of the initial estimate of <math display="inline"><semantics> <mrow> <mi>W</mi> <mrow> <mo>(</mo> <mi>E</mi> <mo>)</mo> </mrow> </mrow> </semantics></math> using a Bézier curve and its control points.</p>
Full article ">Figure 5
<p>Polychromatic attenuation measurements for steel, titanium and aluminium.</p>
Full article ">Figure 6
<p>Photograph of the aluminium part used as a case study for the beam hardening correction; the nominal outer diameter is 25 mm, and the nominal inner diameter is 6 mm.</p>
Full article ">Figure 7
<p>Polychromatic and monochromatic attenuation values for increasing material thicknesses. The polychromatic attenuation values are simulated using the estimated <math display="inline"><semantics> <mrow> <mi>W</mi> <mrow> <mo>(</mo> <mi>E</mi> <mo>)</mo> </mrow> </mrow> </semantics></math>. A polynomial function is fitted to the polychromatic data, and the first-order coefficient is the gradient of the monochromatic line.</p>
Full article ">Figure 8
<p>Beam hardening correction curve derived from the polychromatic and monochromatic attenuation curves in <a href="#sensors-21-03284-f007" class="html-fig">Figure 7</a>. The correction curve is approximated using a third-order polynomial, which is a good fit, as shown by the coefficient of determination.</p>
Full article ">Figure 9
<p>Comparison between the actual <math display="inline"><semantics> <mrow> <mi>W</mi> <mrow> <mo>(</mo> <mi>E</mi> <mo>)</mo> </mrow> </mrow> </semantics></math> used to simulate a set of attenuation values and the <math display="inline"><semantics> <mrow> <mi>W</mi> <mrow> <mo>(</mo> <mi>E</mi> <mo>)</mo> </mrow> </mrow> </semantics></math> estimated from the simulated attenuation values.</p>
Full article ">Figure 10
<p>Comparison between simulated and estimated X-ray attenuation for aluminium, titanium and steel.</p>
Full article ">Figure 11
<p>Comparison between initial and final estimated X-ray spectra based on measured attenuation values.</p>
Full article ">Figure 12
<p>Comparison between measured and estimated X-ray attenuation for aluminium, titanium and steel.</p>
Full article ">Figure 13
<p>Transverse plane CT images of the aluminium workpiece. <b>Left</b>: uncorrected. <b>Right</b>: beam hardening corrected.</p>
Full article ">Figure 14
<p>Comparison of line profiles drawn across the central row of the CT images in <a href="#sensors-21-03284-f013" class="html-fig">Figure 13</a>.</p>
Full article ">Figure 15
<p>Grey value histograms of the CT images in <a href="#sensors-21-03284-f013" class="html-fig">Figure 13</a>. The leftmost peak represents background grey values, and the rightmost peak represents the aluminium material grey values. <b>Left</b>: histogram of the uncorrected data. <b>Right</b>: histogram of the beam hardening corrected data. Grey dotted lines are the mode of each phase, whilst solid grey lines are the ±34% dispersion of each phase.</p>
Full article ">
30 pages, 12707 KiB  
Article
Towards Characterizing and Developing Formation and Migration Cues in Seafloor Sand Waves on Topology, Morphology, Evolution from High-Resolution Mapping via Side-Scan Sonar in Autonomous Underwater Vehicles
by Rui Nian, Lina Zang, Xue Geng, Fei Yu, Shidong Ren, Bo He and Xishuang Li
Sensors 2021, 21(9), 3283; https://doi.org/10.3390/s21093283 - 10 May 2021
Cited by 6 | Viewed by 3025
Abstract
Sand waves constitute ubiquitous geomorphology distribution in the ocean. In this paper, we quantitatively investigate the sand wave variation of topology, morphology, and evolution from the high-resolution mapping of a side scan sonar (SSS) in an Autonomous Underwater Vehicle (AUV), in favor of [...] Read more.
Sand waves constitute ubiquitous geomorphology distribution in the ocean. In this paper, we quantitatively investigate the sand wave variation of topology, morphology, and evolution from the high-resolution mapping of a side scan sonar (SSS) in an Autonomous Underwater Vehicle (AUV), in favor of online sequential Extreme Learning Machine (OS-ELM). We utilize echo intensity directly derived from SSS to help accelerate detection and localization, denote a collection of Gaussian-type morphological templates, with one integrated matching criterion for similarity assessment, discuss the envelope demodulation, zero-crossing rate (ZCR), cross-correlation statistically, and estimate the specific morphological parameters. It is demonstrated that the sand wave detection rate could reach up to 95.61% averagely, comparable to deep learning such as MobileNet, but at a much higher speed, with the average test time of 0.0018 s, which is particularly superior for sand waves at smaller scales. The calculation of morphological parameters primarily infer a wave length range and composition ratio in all types of sand waves, implying the possible dominant direction of hydrodynamics. The proposed scheme permits to delicately and adaptively explore the submarine geomorphology of sand waves with online computation strategies and symmetrically integrate evidence of its spatio-temporal responses during formation and migration. Full article
(This article belongs to the Section Remote Sensors)
Show Figures

Figure 1

Figure 1
<p>SSS characterization in sand wave inspections.</p>
Full article ">Figure 2
<p>The flow chart of the proposed scheme.</p>
Full article ">Figure 3
<p>The flowchart of OS-ELM learning.</p>
Full article ">Figure 4
<p>Example Gaussian-type templates.</p>
Full article ">Figure 5
<p>AUV trajectories and the manually labeled sand wave region. (<b>a</b>) AUV trajectories. (<b>b</b>) The manually labeled sand wave region.</p>
Full article ">Figure 6
<p>AUV deployment.</p>
Full article ">Figure 7
<p>AUV path and SSS mosaicking. (<b>a</b>) AUV path. (<b>b</b>) SSS mosaicking.</p>
Full article ">Figure 8
<p>Example SSS annotation. (<b>a</b>) Example SSS imaging. (<b>b</b>) Labeling.</p>
Full article ">Figure 9
<p>Example echo intensity sub-sequence extraction from raw SSS imaging.</p>
Full article ">Figure 10
<p>TVG correction for example SSS imaging. (<b>a</b>) The corrected example SSS imaging 1. (<b>b</b>) The corrected example SSS imaging 2.</p>
Full article ">Figure 11
<p>TVG correction performance comparison. (<b>a</b>) Raw data. (<b>b</b>) <math display="inline"><semantics> <mrow> <mi>A</mi> <mrow> <mo>=</mo> <mn>30</mn> </mrow> </mrow> </semantics></math>, <math display="inline"><semantics> <mrow> <mi>B</mi> <mo>=</mo> <mn>20</mn> </mrow> </semantics></math>. (<b>c</b>) <math display="inline"><semantics> <mrow> <mi>A</mi> <mrow> <mo>=</mo> <mn>15</mn> </mrow> </mrow> </semantics></math>, <math display="inline"><semantics> <mrow> <mi>B</mi> <mo>=</mo> <mn>10</mn> </mrow> </semantics></math>. (<b>d</b>) <math display="inline"><semantics> <mrow> <mi>A</mi> <mrow> <mo>=</mo> <mn>5</mn> </mrow> </mrow> </semantics></math>, <math display="inline"><semantics> <mrow> <mi>B</mi> <mo>=</mo> <mn>3.1</mn> </mrow> </semantics></math>. (<b>e</b>) <math display="inline"><semantics> <mrow> <mi>A</mi> <mrow> <mo>=</mo> <mn>4</mn> </mrow> </mrow> </semantics></math>, <math display="inline"><semantics> <mrow> <mi>B</mi> <mo>=</mo> <mn>3</mn> </mrow> </semantics></math>. (<b>f</b>) <math display="inline"><semantics> <mrow> <mi>A</mi> <mrow> <mo>=</mo> <mn>2</mn> </mrow> </mrow> </semantics></math>, <math display="inline"><semantics> <mrow> <mi>B</mi> <mo>=</mo> <mn>1</mn> </mrow> </semantics></math>.</p>
Full article ">Figure 12
<p>Example sand wave profile waveform and denoising. (<b>a</b>) Raw echo intensity sub-sequence. (<b>b</b>) Median filtering. (<b>c</b>) Gaussian filtering. (<b>d</b>) EEMD denoising.</p>
Full article ">Figure 13
<p>Sand wave detection rate and time regarding to the number of hidden layer nodes in OS-ELM. (<b>a</b>) Test recognition rate. (<b>b</b>) Test time.</p>
Full article ">Figure 14
<p>Sand wave detection for example SSS imaging. (<b>a</b>) Basic MobileNet with SSS imaging. (<b>b</b>) OS-ELM with echo intensity.</p>
Full article ">Figure 15
<p>Example sand wave profile waveform matching.</p>
Full article ">Figure 15 Cont.
<p>Example sand wave profile waveform matching.</p>
Full article ">Figure 16
<p>Geometric correction example. (<b>a</b>) Raw SSS imaging. (<b>b</b>) Raw echo intensity values. (<b>c</b>) SSS imaging after geometric correction. (<b>d</b>) Echo intensity values after geometric correction.</p>
Full article ">Figure 17
<p>Zero-crossing spectrum. (<b>a</b>) Without sand wave. (<b>b</b>–<b>d</b>) With sand waves.</p>
Full article ">Figure 18
<p>ZCR statistics. (<b>a</b>) With sand waves. (<b>b</b>) Without sand waves.</p>
Full article ">Figure 19
<p>Cross-correlation statistics. (<b>a</b>) With sand wave. (<b>b</b>)Without sand wave.</p>
Full article ">Figure 20
<p>Wave length estimation distribution for sand waves.</p>
Full article ">Figure 21
<p>Morphological relationship of wave length estimation versus local maxima echo intensity for sand waves.</p>
Full article ">Figure 22
<p>Asymmetric index definition.</p>
Full article ">Figure 23
<p>Asymmetrical index estimation distribution for sand waves.</p>
Full article ">Figure 24
<p>Example echo intensity in SSS imaging at the intersection of two AUV trajectories within the duration of more than one year. (<b>a</b>) SSS imaging in December 2019. (<b>b</b>) SSS imaging in January 2021. (<b>c</b>) Echo intensity curve regarding the ping in red on SSS imaging (<b>a</b>). (<b>d</b>) Echo intensity curve regarding the ping in red on SSS imaging (<b>b</b>).</p>
Full article ">
25 pages, 2013 KiB  
Article
Asynchronous Chirp Slope Keying for Underwater Acoustic Communication
by Dominik Jan Schott, Andrea Gabbrielli, Wenxin Xiong, Georg Fischer, Fabian Höflinger, Johannes Wendeberg, Christian Schindelhauer and Stefan Johann Rupitsch
Sensors 2021, 21(9), 3282; https://doi.org/10.3390/s21093282 - 10 May 2021
Cited by 14 | Viewed by 3230
Abstract
We propose an asynchronous acoustic chirp slope keying to map short bit sequences on single or multiple bands without preamble or error correction coding on the physical layer. We introduce a symbol detection scheme in the demodulator that uses the superposed matched filter [...] Read more.
We propose an asynchronous acoustic chirp slope keying to map short bit sequences on single or multiple bands without preamble or error correction coding on the physical layer. We introduce a symbol detection scheme in the demodulator that uses the superposed matched filter results of up and down chirp references to estimate the symbol timing, which removes the requirement of a preamble for symbol synchronization. Details of the implementation are disclosed and discussed, and the performance is verified in a pool measurement on laboratory scale, as well as the simulation for a channel containing Rayleigh fading and Additive White Gaussian Noise. For time-bandwidth products (TB) of 50 in single band mode, a raw data rate of 100 bit/s is simulated to achieve bit error rates (BER) below 0.001 for signal-to-noise ratios above −6 dB. In dual-band mode, for TB of 25 and a data rate of 200 bit/s, the same bit error level was achieved for signal-to-noise ratios above 0 dB. The simulated packet error rates (PER) follow the general behavior of the BER, but with a higher error probability, which increases with the length of bits in each packet. Full article
(This article belongs to the Special Issue Applications of Ultrasonic Sensors)
Show Figures

Figure 1

Figure 1
<p>Flow diagram of the basic communication chain: The data <span class="html-italic">d</span> is modulated, amplified before transmission, filtered and amplified at reception, and demodulated as <math display="inline"><semantics> <msub> <mi>d</mi> <mi>est</mi> </msub> </semantics></math>. The entire analog domain is regarded as part of the communication channel.</p>
Full article ">Figure 2
<p>Modulator block in detail: The data input <span class="html-italic">d</span> is mapped onto <math display="inline"><semantics> <msub> <mi>N</mi> <mi>lo</mi> </msub> </semantics></math> sub-bands through a multiplexer (MUX) and modulated by the up-converted chirped symbols from the DUC. The transmission sequence <math display="inline"><semantics> <msub> <mi>y</mi> <mi>tx</mi> </msub> </semantics></math> is assembled by the CSK block, already in the transmission band. Simple arrow lines indicate vectors, double lines arrays.</p>
Full article ">Figure 3
<p>Resampling example for a linear chirp with <math display="inline"><semantics> <mrow> <msub> <mi>f</mi> <mi mathvariant="normal">c</mi> </msub> <mo>=</mo> <mn>3</mn> <mtext> </mtext> <mi>kHz</mi> </mrow> </semantics></math>, <math display="inline"><semantics> <mrow> <msub> <mi>B</mi> <mi mathvariant="normal">f</mi> </msub> <mo>=</mo> <mn>5</mn> <mtext> </mtext> <mi>kHz</mi> </mrow> </semantics></math>, and <math display="inline"><semantics> <mrow> <mi>T</mi> <mo>=</mo> <mn>10</mn> <mtext> </mtext> <mi mathvariant="normal">ms</mi> </mrow> </semantics></math>. <b>leftmost:</b> Base band signal <math display="inline"><semantics> <msub> <mi>y</mi> <mi>tb</mi> </msub> </semantics></math> at the transmitter, <b>center left:</b> Transmission band <math display="inline"><semantics> <msub> <mi>y</mi> <mi>ib</mi> </msub> </semantics></math>, <b>center right:</b> Undersampled signal on reception, <b>rightmost:</b> Down-converted base band signal <math display="inline"><semantics> <msub> <mi>y</mi> <mi>bb</mi> </msub> </semantics></math> at the receiver where the originally transmitted signal is overlayed in gray. The transmission band’s center frequency is at <math display="inline"><semantics> <mrow> <mn>67.5</mn> </mrow> </semantics></math> <math display="inline"><semantics> <mi>kHz</mi> </semantics></math>. In the experiments, the intermediate band on reception is around <math display="inline"><semantics> <mrow> <mn>20.5</mn> </mrow> </semantics></math> <math display="inline"><semantics> <mi>kHz</mi> </semantics></math> due to undersampling with only <math display="inline"><semantics> <mrow> <msub> <mi>f</mi> <mi mathvariant="normal">s</mi> </msub> <mo>=</mo> <mn>88</mn> <mtext> </mtext> <mi>kHz</mi> </mrow> </semantics></math>. Note the changed frequency scale for the Bode plots in the two columns on the right.</p>
Full article ">Figure 4
<p>Autocorrelation magnitude comparison of a selection of shaping window functions. All magnitudes are normalized by the Dirichlet shaped chirp power for comparison. Gaussian noise was added to a signal-to-noise ratio <math display="inline"><semantics> <mrow> <mi>S</mi> <mi>N</mi> <mi>R</mi> <mo>=</mo> <mn>0</mn> <mrow> <mtext> </mtext> <mi mathvariant="normal">dB</mi> </mrow> </mrow> </semantics></math>, as well as two echoes at <math display="inline"><semantics> <mrow> <mi>n</mi> <mo>=</mo> <mn>9</mn> </mrow> </semantics></math> and <math display="inline"><semantics> <mrow> <mi>n</mi> <mo>=</mo> <mn>19</mn> </mrow> </semantics></math>. To the left the spatial resolution and peak power is higher, to the right the inter-signal interference and spectral leakage are reduced.</p>
Full article ">Figure 5
<p>Simulated spectrograms of the autocorrelations of a selection of shaping window functions. All magnitudes are normalized by the Dirichlet shaped chirp power for comparison. Gaussian noise was added to a signal-to-noise ratio of 0 <math display="inline"><semantics> <mi mathvariant="normal">d</mi> </semantics></math><math display="inline"><semantics> <mi mathvariant="normal">B</mi> </semantics></math>, as well as two echoes at <math display="inline"><semantics> <mrow> <mi>n</mi> <mo>=</mo> <mn>9</mn> </mrow> </semantics></math> and <math display="inline"><semantics> <mrow> <mi>n</mi> <mo>=</mo> <mn>19</mn> </mrow> </semantics></math>.</p>
Full article ">Figure 6
<p>Comparison of the conventional and proposed demodulation as a block diagram in detail. (<b>a</b>) In the former case, the received sequence <span class="html-italic">y</span><sub>rx</sub> is processed by Digital Down-Coverter (DDC), compressed through a Fast Hilbert Cross-Correlator (FHX), converted into symbol space through Frame Detect &amp; Downsample (FDDS), which is interpreted by a binary decision (Decide) block, and finally assembled into the estimated data output <span class="html-italic">d</span><sub>est</sub> through a reverting multiplexer (De-MUX). (<b>b</b>) We propose the insertion of a superposition in the compressed sample space through the Join &amp; Downsample (JDS) block that creates a sum signal for symbol timing and a difference signal for symbol extraction.</p>
Full article ">Figure 7
<p>Schematic experimental set-up in for the acoustic transmission inside a water tank. The tank is filled with fresh water and located inside a closed building. A comparable scenario would be two divers working on a ship’s hull or an UAV inspecting a lake harbor’s foundations.</p>
Full article ">Figure 8
<p>Spectrograms showing the intermediate frequency over time of parts of the signal to illustrate the effects of the channel and undersampling. (<b>a</b>) Single band signal after up-conversion before transmission; (<b>b</b>) Multi-band signal after up-conversion before transmission; (<b>c</b>) Single band after reception before down-conversion; (<b>d</b>) Multi-band after reception before down-conversion. Each package is transmitted three times in the experiment.</p>
Full article ">Figure 8 Cont.
<p>Spectrograms showing the intermediate frequency over time of parts of the signal to illustrate the effects of the channel and undersampling. (<b>a</b>) Single band signal after up-conversion before transmission; (<b>b</b>) Multi-band signal after up-conversion before transmission; (<b>c</b>) Single band after reception before down-conversion; (<b>d</b>) Multi-band after reception before down-conversion. Each package is transmitted three times in the experiment.</p>
Full article ">Figure 9
<p>Averaged spectral power plots of the raw received signals. (<b>a</b>) Single band communication; (<b>b</b>) Dual-band communication. The colored area marks the ± 1 <span class="html-italic">σ</span> region of each frequency bin.</p>
Full article ">Figure 10
<p>Time domain plot of the detected frames (red) and estimated symbols (blue dots). (<b>a</b>) Single band communication; (<b>b</b>) Dual-band communication. The symbol difference is not optimally detected, as the amplitude of the signal exceeds the amplitude of the estimated symbols. The histograms to the right of each time plot are normalized by the total number of samples (red) and symbols (blue) in each frame.</p>
Full article ">Figure 11
<p>Plots of the simulated bit error rate (BER) and packet error rate (PER) for both single band (<b>a</b>) and dual-band transmission (<b>b</b>). The markers indicate each simulated SNR condition, the lines are manually fitted curves.</p>
Full article ">
18 pages, 4220 KiB  
Article
Non-Local and Multi-Scale Mechanisms for Image Inpainting
by Xu He and Yong Yin
Sensors 2021, 21(9), 3281; https://doi.org/10.3390/s21093281 - 10 May 2021
Cited by 5 | Viewed by 2531
Abstract
Recently, deep learning-based techniques have shown great power in image inpainting especially dealing with squared holes. However, they fail to generate plausible results inside the missing regions for irregular and large holes as there is a lack of understanding between missing regions and [...] Read more.
Recently, deep learning-based techniques have shown great power in image inpainting especially dealing with squared holes. However, they fail to generate plausible results inside the missing regions for irregular and large holes as there is a lack of understanding between missing regions and existing counterparts. To overcome this limitation, we combine two non-local mechanisms including a contextual attention module (CAM) and an implicit diversified Markov random fields (ID-MRF) loss with a multi-scale architecture which uses several dense fusion blocks (DFB) based on the dense combination of dilated convolution to guide the generative network to restore discontinuous and continuous large masked areas. To prevent color discrepancies and grid-like artifacts, we apply the ID-MRF loss to improve the visual appearance by comparing similarities of long-distance feature patches. To further capture the long-term relationship of different regions in large missing regions, we introduce the CAM. Although CAM has the ability to create plausible results via reconstructing refined features, it depends on initial predicted results. Hence, we employ the DFB to obtain larger and more effective receptive fields, which benefits to predict more precise and fine-grained information for CAM. Extensive experiments on two widely-used datasets demonstrate that our proposed framework significantly outperforms the state-of-the-art approaches both in quantity and quality. Full article
(This article belongs to the Section Sensing and Imaging)
Show Figures

Figure 1

Figure 1
<p>The overall architecture of our method. Region-wise convolution indicates using different convolution filters for different regions, more details can be found in paper [<a href="#B15-sensors-21-03281" class="html-bibr">15</a>]. In this architecture, 256 by 256 and 32 denote the size and channels of the feature map respectively.</p>
Full article ">Figure 2
<p>The framework of the dense fusion block. “Conv-3-2” indicates the 3 by 3 convolution layer and the dilated rate is 2. <math display="inline"><semantics> <mo>⊕</mo> </semantics></math> is element-wise summation. The output channels of all convolutional layers are 64, except for the last layer which is 256.</p>
Full article ">Figure 3
<p>Qualitative comparisons of different methods on discontinuous missing areas.</p>
Full article ">Figure 4
<p>Qualitative comparisons of different methods on continuous missing areas.</p>
Full article ">Figure 5
<p>Qualitative results of ablation studies on discontinuous missing regions. (Best viewed with zoom-in.)</p>
Full article ">Figure 6
<p>Qualitative results of ablation studies on continuous missing regions. (Best viewed with zoom-in.)</p>
Full article ">
10 pages, 2075 KiB  
Communication
Utilization of Inertial Measurement Units for Determining the Sequential Chain of Baseball Strike Posture
by Yun-Ju Lee, Po-Chieh Lin, Ling-Ying Chen, Yu-Jung Chen and Jing Nong Liang
Sensors 2021, 21(9), 3280; https://doi.org/10.3390/s21093280 - 10 May 2021
Cited by 1 | Viewed by 2525
Abstract
The purpose of this study was to employ inertial measurement units (IMU) with an eye-tracking device to investigate different swing strategies between two levels of batters. The participants were 20 healthy males aged 20 to 30 years old, with ten professional and ten [...] Read more.
The purpose of this study was to employ inertial measurement units (IMU) with an eye-tracking device to investigate different swing strategies between two levels of batters. The participants were 20 healthy males aged 20 to 30 years old, with ten professional and ten amateur batters. Eye gaze position, head, shoulder, trunk, and pelvis angular velocity, and ground reaction forces were recorded. The results showed that professional batters rotated segments more rhythmically and efficiently than the amateur group. Firstly, the professional group spent less time in the preparation stages. Secondly, the maximum angular velocity timing of each segment of the professional group was centralized in the swing cycle. Thirdly, the amateur group had significantly earlier gaze timing of the maximum angular velocity than the professional group. Moreover, the maximum angular velocity timing of the gaze was the earliest parameter among the five segments, and significantly earlier (at least 16.32% of cycle time) than the maximum angular velocity of the head, shoulder, trunk, and pelvis within the amateur group. The visual-motor coordination strategies were different between the two groups, which could successfully be determined by wearable instruments of IMU. Full article
(This article belongs to the Collection Instrument and Measurement)
Show Figures

Figure 1

Figure 1
<p>Experimental setting. Five circles represent the placements of five inertial measurement units (IMU). 1—head; 2—vertebra prominens (C7); 3—thoracic vertebrae 10 (T10); 4—sacrum; and 5—the bottom of the bat.</p>
Full article ">Figure 2
<p>The swing cycle could be divided into four stages by five events. The five events are FF<sub>min</sub>, FF<sub>50</sub>, FF<sub>max</sub>, IMP, and BF<sub>50</sub> (FF stands for the ground reaction force (GRF) of the front foot, and IMP is the impact moment). The four stages are S<sub>st</sub> (start stage), S<sub>ff</sub> (front foot stage), S<sub>bf</sub> (back foot stage), and S<sub>fn</sub> (finish stage).</p>
Full article ">Figure 3
<p>This chart shows the time percentage and timing of the four swing stages of the two groups. In stage S<sub>fn</sub>, the percentage of the amateur batter’s time was significantly lower than professional batters. (*: <span class="html-italic">p</span> &lt; 0.05).</p>
Full article ">Figure 4
<p>The angular velocity of the amateur and professional groups of the five key events of the head, shoulders, trunk, pelvis, and gaze. (*: <span class="html-italic">p</span> &lt; 0.05).</p>
Full article ">Figure 5
<p>The maximum angular velocity timing of each segment in the swing cycle. The amateur group cost about 23.78% cycle time to complete the maximum angular velocity. However, the professional group cost only about 5.37% cycle time. The maximum gaze angular velocity timing of the amateur group was significantly earlier (16.32% cycle time) than the other segments’ intragroup. Further, the maximum gaze angular velocity timing was earlier, about 20.25% cycle time before IMP of the amateur group. However, only 2.25% cycle time before IMP of the professional group.</p>
Full article ">Figure 6
<p>For the amateur group, there was a simple main effect of the segment. The timing of gaze was significantly earlier than the head, shoulder, trunk, and pelvis of the amateur group (*: <span class="html-italic">p</span> &lt; 0.001). For the professional group, there was no significant single main effect.</p>
Full article ">
26 pages, 2009 KiB  
Article
Toward an Automatic Quality Assessment of Voice-Based Telemedicine Consultations: A Deep Learning Approach
by Maria Habib, Mohammad Faris, Raneem Qaddoura, Manal Alomari, Alaa Alomari and Hossam Faris
Sensors 2021, 21(9), 3279; https://doi.org/10.3390/s21093279 - 10 May 2021
Cited by 11 | Viewed by 3733
Abstract
Maintaining a high quality of conversation between doctors and patients is essential in telehealth services, where efficient and competent communication is important to promote patient health. Assessing the quality of medical conversations is often handled based on a human auditory-perceptual evaluation. Typically, trained [...] Read more.
Maintaining a high quality of conversation between doctors and patients is essential in telehealth services, where efficient and competent communication is important to promote patient health. Assessing the quality of medical conversations is often handled based on a human auditory-perceptual evaluation. Typically, trained experts are needed for such tasks, as they follow systematic evaluation criteria. However, the daily rapid increase of consultations makes the evaluation process inefficient and impractical. This paper investigates the automation of the quality assessment process of patient–doctor voice-based conversations in a telehealth service using a deep-learning-based classification model. For this, the data consist of audio recordings obtained from Altibbi. Altibbi is a digital health platform that provides telemedicine and telehealth services in the Middle East and North Africa (MENA). The objective is to assist Altibbi’s operations team in the evaluation of the provided consultations in an automated manner. The proposed model is developed using three sets of features: features extracted from the signal level, the transcript level, and the signal and transcript levels. At the signal level, various statistical and spectral information is calculated to characterize the spectral envelope of the speech recordings. At the transcript level, a pre-trained embedding model is utilized to encompass the semantic and contextual features of the textual information. Additionally, the hybrid of the signal and transcript levels is explored and analyzed. The designed classification model relies on stacked layers of deep neural networks and convolutional neural networks. Evaluation results show that the model achieved a higher level of precision when compared with the manual evaluation approach followed by Altibbi’s operations team. Full article
(This article belongs to the Section Intelligent Sensors)
Show Figures

Figure 1

Figure 1
<p>The process of extracting the Mel spectrogram from an acoustic signal, where the output of each Mel filter is summed then combined to create the Mel spectrogram, which is visualized in terms of the amplitude of the frequency components over time.</p>
Full article ">Figure 2
<p>The anatomy of a deep neural network model, <span class="html-italic">a</span> is the number of hidden layers.</p>
Full article ">Figure 3
<p>Description of the convolutional neural networks. In (<b>A</b>), the filter size is 3 and the stride is 1, (<b>B</b>) the filter size is 2 and the stride is 1, and (<b>C</b>) the filter size is 2 and the stride is 2.</p>
Full article ">Figure 4
<p>A schematic overview of the conducted methodology.</p>
Full article ">Figure 5
<p>An illustration of implementing the first approach. In (<b>A</b>), the dataset is preprocessed and created, while (<b>B</b>) shows the utilized DNN model.</p>
Full article ">Figure 6
<p>A representation of a consultation’s acoustic signal. The first plot is the signal in time domain, the second is the ZCR, the third shows the MFCCs (showing the first 8 coefficients), and the fourth is the spectrogram representation.</p>
Full article ">Figure 7
<p>The process of converting the acoustic recordings into text and then extracting text-based features using AraVec.</p>
Full article ">Figure 8
<p>The structural design of the second approach. MP is the max-pooling operation, and BN is the batch normalization layer.</p>
Full article ">Figure 9
<p>The structure of combining the two approaches of the signal-based submodel and the transcript-based submodel.</p>
Full article ">Figure 10
<p>The heatmap representation of the confusion matrix of the best models obtained from using the MFCCs alone (<b>a</b>) and using the combination of all spectral features (<b>b</b>).</p>
Full article ">Figure 11
<p>A heatmap representation of the confusion matrices of the best models from the transcript-based approach with different embedding models.</p>
Full article ">Figure 12
<p>The convergence curves in terms of the accuracy of the best models for the three approaches using the spectral features only, the transcript features only, and the hybrid of both.</p>
Full article ">Figure 12 Cont.
<p>The convergence curves in terms of the accuracy of the best models for the three approaches using the spectral features only, the transcript features only, and the hybrid of both.</p>
Full article ">Figure 13
<p>The convergence curves in terms of the loss of the best models for the three approaches of using the spectral features only, the transcript features only, and the hybrid of both.</p>
Full article ">Figure 13 Cont.
<p>The convergence curves in terms of the loss of the best models for the three approaches of using the spectral features only, the transcript features only, and the hybrid of both.</p>
Full article ">
10 pages, 5432 KiB  
Communication
Spectral Characteristics of Square-Wave-Modulated Type II Long-Period Fiber Gratings Inscribed by a Femtosecond Laser
by Xiaofan Zhao, Hongye Li, Binyu Rao, Meng Wang, Baiyi Wu and Zefeng Wang
Sensors 2021, 21(9), 3278; https://doi.org/10.3390/s21093278 - 10 May 2021
Cited by 4 | Viewed by 2442
Abstract
We study here the spectral characteristics of square-wave-modulated type II long-period fiber gratings (LPFGs) inscribed by a femtosecond laser. Both theoretical and experimental results indicate that higher-order harmonics refractive index (RI) modulation commonly exists together with the fundamental harmonic RI modulation in such [...] Read more.
We study here the spectral characteristics of square-wave-modulated type II long-period fiber gratings (LPFGs) inscribed by a femtosecond laser. Both theoretical and experimental results indicate that higher-order harmonics refractive index (RI) modulation commonly exists together with the fundamental harmonic RI modulation in such LPFGs, and the duty cycle of a square wave has a great influence on the number and amplitudes of higher-order harmonics. A linear increase in the duty cycle in a series of square wave pulses will induce another LPFG with a minor difference in periods, which is useful for expanding the bandwidth of LPFGs. We also propose a method to reduce insertion loss by fabricating type II LPFGs without higher-order harmonic resonances. This work intensifies our comprehension of type II fiber gratings with which novel optical fiber sensors can be fabricated. Full article
(This article belongs to the Special Issue Advanced Fiber Photonic Devices and Sensors)
Show Figures

Figure 1

Figure 1
<p>Phase-matching condition of the first-order harmonic resonance.</p>
Full article ">Figure 2
<p>(<b>a</b>) Schematic of a square wave. (<b>b</b>) Spatial spectrum characteristics of a square wave with different duty cycles (color bar indicates the amplitude of each harmonic).</p>
Full article ">Figure 3
<p>Amplitude of each resonance versus the number of periods.</p>
Full article ">Figure 4
<p>(<b>a</b>) Waveform of a square wave with a linearly growing duty cycle. (<b>b</b>) Spatial spectrum characteristics of a square wave with a linearly growing duty cycle.</p>
Full article ">Figure 5
<p>(<b>a</b>) Schematic of LPFG fabrication. (<b>b</b>) RI modulation generated point-by-point (magnification: 100×). (<b>c</b>) Microscope image of a square-wave-modulated LPFG (magnification: 100×).</p>
Full article ">Figure 6
<p>Spectrum evolution during LPFG fabrication (period: 560 μm; duty cycle: 50%. Insertion loss from 1200 to 1350 nm increases with the grating length).</p>
Full article ">Figure 7
<p>Transmission spectra of LPFGs with different duty cycles (insertion loss of the LPFG with a duty cycle of 10% is the smallest, and that of the LPFG with a duty cycle of 25% is the highest).</p>
Full article ">Figure 8
<p>Transmission spectra of LPFGs with periods of 560, 1120, and 1680 μm (grating length: 40,320 μm. Resonant intensity decreases with the period, and when the duty cycle is 50%, the second-order harmonic resonance of the grating disappears).</p>
Full article ">Figure 9
<p>Spectrum evolution during LPFG fabrication (insertion loss increases with the grating length).</p>
Full article ">
11 pages, 1279 KiB  
Article
IMU-Based Effects Assessment of the Use of Foot Orthoses in the Stance Phase during Running and Asymmetry between Extremities
by Juan Luis Florenciano Restoy, Jordi Solé-Casals and Xantal Borràs-Boix
Sensors 2021, 21(9), 3277; https://doi.org/10.3390/s21093277 - 10 May 2021
Cited by 4 | Viewed by 3943
Abstract
The objectives of this study were to determine the amplitude of movement differences and asymmetries between feet during the stance phase and to evaluate the effects of foot orthoses (FOs) on foot kinematics in the stance phase during running. In total, 40 males [...] Read more.
The objectives of this study were to determine the amplitude of movement differences and asymmetries between feet during the stance phase and to evaluate the effects of foot orthoses (FOs) on foot kinematics in the stance phase during running. In total, 40 males were recruited (age: 43.0 ± 13.8 years, weight: 72.0 ± 5.5 kg, height: 175.5 ± 7.0 cm). Participants ran on a running treadmill at 2.5 m/s using their own footwear, with and without the FOs. Two inertial sensors fixed on the instep of each of the participant’s footwear were used. Amplitude of movement along each axis, contact time and number of steps were considered in the analysis. The results indicate that the movement in the sagittal plane is symmetric, but that it is not in the frontal and transverse planes. The right foot displayed more degrees of movement amplitude than the left foot although these differences are only significant in the abduction case. When FOs are used, a decrease in amplitude of movement in the three axes is observed, except for the dorsi-plantar flexion in the left foot and both feet combined. The contact time and the total step time show a significant increase when FOs are used, but the number of steps is not altered, suggesting that FOs do not interfere in running technique. The reduction in the amplitude of movement would indicate that FOs could be used as a preventive tool. The FOs do not influence the asymmetry of the amplitude of movement observed between feet, and this risk factor is maintained. IMU devices are useful tools to detect risk factors related to running injuries. With its use, even more personalized FOs could be manufactured. Full article
(This article belongs to the Special Issue Biomedical Sensors-Recent Advances and Future Challenges)
Show Figures

Figure 1

Figure 1
<p>(<b>Left</b>): position of the sensor attached to the instep of the running shoe. The triple orthogonal system represented by the arrows indicate the dorsi–planar flexion (red), abduction–adduction (blue) and eversion–inversion (green) movements. (<b>Right</b>): type of FOs used with the polypro-pylene and the EVA layers.</p>
Full article ">Figure 2
<p>(<b>Left</b>): angular displacement of the foot as a function of the percentage of the running cycle. (<b>Right</b>): stance phase segmentation graph where A indicates the start of contact, B corresponds to the mid-stance (stabilization), and C indicates the end of the stance. Legend: D–PF indicates dorsi–plantar flexion, ABD–ADD indicates abduction–abduction, EV–INV indicates eversion–inversion.</p>
Full article ">
16 pages, 381 KiB  
Article
Spiking Neural Network with Linear Computational Complexity for Waveform Analysis in Amperometry
by Szymon Szczęsny, Damian Huderek and Łukasz Przyborowski
Sensors 2021, 21(9), 3276; https://doi.org/10.3390/s21093276 - 10 May 2021
Cited by 4 | Viewed by 2223
Abstract
The paper describes the architecture of a Spiking Neural Network (SNN) for time waveform analyses using edge computing. The network model was based on the principles of preprocessing signals in the diencephalon and using tonic spiking and inhibition-induced spiking models typical for the [...] Read more.
The paper describes the architecture of a Spiking Neural Network (SNN) for time waveform analyses using edge computing. The network model was based on the principles of preprocessing signals in the diencephalon and using tonic spiking and inhibition-induced spiking models typical for the thalamus area. The research focused on a significant reduction of the complexity of the SNN algorithm by eliminating most synaptic connections and ensuring zero dispersion of weight values concerning connections between neuron layers. The paper describes a network mapping and learning algorithm, in which the number of variables in the learning process is linearly dependent on the size of the patterns. The works included testing the stability of the accuracy parameter for various network sizes. The described approach used the ability of spiking neurons to process currents of less than 100 pA, typical of amperometric techniques. An example of a practical application is an analysis of vesicle fusion signals using an amperometric system based on Carbon NanoTube (CNT) sensors. The paper concludes with a discussion of the costs of implementing the network as a semiconductor structure. Full article
(This article belongs to the Section Physical Sensors)
Show Figures

Figure 1

Figure 1
<p>Tonic spiking neuron response for the input current in the range of 0 ÷ 50 pA.</p>
Full article ">Figure 2
<p>Inhibition-induced spiking neuron response for an input current of −30 ÷ 70 pA.</p>
Full article ">Figure 3
<p>SNN architecture with linear computational complexity.</p>
Full article ">Figure 4
<p>The applied SNN response coding.</p>
Full article ">Figure 5
<p>Application of the described SNN in the task of analyzing data from CNT sensors.</p>
Full article ">Figure 6
<p>Amperometric waveform showing full vesicle fusion.</p>
Full article ">Figure 7
<p>Positive <span style="color:#0000FF">—</span> and negative <span style="color:#FF0000">—</span> patterns used in SNN training with the sampling parameter <math display="inline"><semantics> <mrow> <mi>N</mi> <mo>=</mo> <mn>20</mn> </mrow> </semantics></math>.</p>
Full article ">Figure 8
<p>Accuracy vs. weight mismatch.</p>
Full article ">
11 pages, 1484 KiB  
Communication
Enhancing the Accuracy of Non-Invasive Glucose Sensing in Aqueous Solutions Using Combined Millimeter Wave and Near Infrared Transmission
by Helena Cano-Garcia, Rohit Kshirsagar, Roberto Pricci, Ahmed Teyeb, Fergus O’Brien, Shimul Saha, Panagiotis Kosmas and Efthymios Kallos
Sensors 2021, 21(9), 3275; https://doi.org/10.3390/s21093275 - 10 May 2021
Cited by 7 | Viewed by 5400
Abstract
We reported measurement results relating to non-invasive glucose sensing using a novel multiwavelength approach that combines radio frequency and near infrared signals in transmission through aqueous glucose-loaded solutions. Data were collected simultaneously in the 37–39 GHz and 900–1800 nm electromagnetic bands. We successfully [...] Read more.
We reported measurement results relating to non-invasive glucose sensing using a novel multiwavelength approach that combines radio frequency and near infrared signals in transmission through aqueous glucose-loaded solutions. Data were collected simultaneously in the 37–39 GHz and 900–1800 nm electromagnetic bands. We successfully detected changes in the glucose solutions with varying glucose concentrations between 80 and 5000 mg/dl. The measurements showed for the first time that, compared to single modality systems, greater accuracy on glucose level prediction can be achieved when combining transmission data from these distinct electromagnetic bands, boosted by machine learning algorithms. Full article
(This article belongs to the Section Electronic Sensors)
Show Figures

Figure 1

Figure 1
<p>(<b>a</b>) Photography of the experimental setup consisting of 2-mm wave sensors enclosed on a custom-made 3D printed case, two optical fibers, and an acrylic tank containing the glucose samples; (<b>b</b>) schematic view of the experimental setup.</p>
Full article ">Figure 2
<p>(<b>a</b>) NIR transmittance for a water sample (0 mg/dl); (<b>b</b>) RF transmission amplitude for a water sample (0 mg/dl).</p>
Full article ">Figure 3
<p>(<b>a</b>) Difference in NIR transmittance for a measurement round using 0 mg/dl as a reference measurement (glucose concentrations between 0 to 5000 mg/dl); (<b>b</b>) difference in RF transmission for one measurement round using 0 mg/dl as a reference (glucose concentrations between 0 to 5000 mg/dl).</p>
Full article ">Figure 4
<p>(<b>a</b>) Mean of the NIR transmittance at 1390 nm for all the measurement rounds for different glucose concentration values; (<b>b</b>) mean of the RF transmission amplitude (S<sub>21</sub>) at 36.5 GHz for all the measurement rounds and different glucose concentration values. For (<b>a</b>,<b>b</b>), the blue dots represent the datapoints and the pink trace represents the mean of all these points. The pink shaded area represents the measurement error obtained by calculating the standard deviation of the measurements. Note the X axes have logarithmic scale.</p>
Full article ">Figure 5
<p>Predicted glucose concentration obtained after applying three ML models against the real glucose concentration. The circle datapoints represent the predicted values obtained when using the ML model tailored for the RF data only (MARD: 76%); the triangle datapoints represent the predictions when applying an ML model only to the NIR data (MARD: 127%); and the square datapoints represent the predicted values when using a ML model using RF and NIR data combined (MARD: 46%). The dashed line indicates the loci of the points where the reference and predicted values would match.</p>
Full article ">
Previous Issue
Next Issue
Back to TopTop