Sensors

13 pages, 6438 KiB

Open AccessArticle

Enabling Modular Robotics with Secure Transducer Identification Based on Extended IEEE 21450 Transducer Electronic Datasheets

by Tobias Mitterer, Christian Lederer and Hubert Zangl

Sensors 2023, 23(5), 2873; https://doi.org/10.3390/s23052873 - 6 Mar 2023

Cited by 1 | Viewed by 2005

Abstract

In robotics, there are many different sensors and actuators mounted onto a robot which may also, in the case of modular robotics, be interchanged during operation. During development of new sensors or actuators, prototypes may also be mounted onto a robot to test [...] Read more.

In robotics, there are many different sensors and actuators mounted onto a robot which may also, in the case of modular robotics, be interchanged during operation. During development of new sensors or actuators, prototypes may also be mounted onto a robot to test functionality, where the new prototypes often have to be integrated manually into the robot environment. Proper, fast and secure identification of new sensor or actuator modules for the robot thus becomes important. In this work, a workflow to add new sensors or actuators to an existing robot environment while establishing trust in an automated manner using electronic datasheets has been developed. The new sensors or actuators are identified via near field communication (NFC) to the system and exchange security information via the same channel. By using electronic datasheets stored on the sensor or actuator, the device can be easily identified and trust can be established by using additional security information contained in the datasheet. In addition, the NFC hardware can simultaneously be used for wireless charging (WLC), thus allowing for wireless sensor and actuator modules. The developed workflow has been tested with prototype tactile sensors mounted onto a robotic gripper. Full article

(This article belongs to the Special Issue Intelligent Sensing and Decision-Making in Advanced Manufacturing)

► Show Figures

Figure 1

15 pages, 4375 KiB

Open AccessArticle

Advanced Pressure Compensation in High Accuracy NDIR Sensors for Environmental Studies

by Bakhram Gaynullin, Christine Hummelgård, Claes Mattsson, Göran Thungström and Henrik Rödjegård

Sensors 2023, 23(5), 2872; https://doi.org/10.3390/s23052872 - 6 Mar 2023

Cited by 5 | Viewed by 2834

Abstract

Measurements of atmospheric gas concentrations using of NDIR gas sensors requires compensation of ambient pressure variations to achieve reliable result. The extensively used general correction method is based on collecting data for varying pressures for a single reference concentration. This one-dimensional compensation approach [...] Read more.

Measurements of atmospheric gas concentrations using of NDIR gas sensors requires compensation of ambient pressure variations to achieve reliable result. The extensively used general correction method is based on collecting data for varying pressures for a single reference concentration. This one-dimensional compensation approach is valid for measurements carried out in gas concentrations close to reference concentration but will introduce significant errors for concentrations further away from the calibration point. For applications, requiring high accuracy, collecting, and storing calibration data at several reference concentrations can reduce the error. However, this method will cause higher demands on memory capacity and computational power, which is problematic for cost sensitive applications. We present here an advanced, but practical, algorithm for compensation of environmental pressure variations for relatively low-cost/high resolution NDIR systems. The algorithm consists of a two-dimensional compensation procedure, which widens the valid pressure and concentrations range but with a minimal need to store calibration data, compared to the general one-dimensional compensation method based on a single reference concentration. The implementation of the presented two-dimensional algorithm was verified at two independent concentrations. The results show a reduction in the compensation error from 5.1% and 7.3%, for the one-dimensional method, to −0.02% and 0.83% for the two-dimensional algorithm. In addition, the presented two-dimensional algorithm only requires calibration in four reference gases and the storing of four sets of polynomial coefficients used for calculations. Full article

(This article belongs to the Section Optical Sensors)

► Show Figures

Figure 1

17 pages, 9894 KiB

Open AccessArticle

An Online 3D Modeling Method for Pose Measurement under Uncertain Dynamic Occlusion Based on Binocular Camera

by Xuanchang Gao, Junzhi Yu and Min Tan

Sensors 2023, 23(5), 2871; https://doi.org/10.3390/s23052871 - 6 Mar 2023

Cited by 2 | Viewed by 2139

Abstract

3D modeling plays a significant role in many industrial applications that require geometry information for pose measurements, such as grasping, spraying, etc. Due to random pose changes in the workpieces on the production line, demand for online 3D modeling has increased and many [...] Read more.

3D modeling plays a significant role in many industrial applications that require geometry information for pose measurements, such as grasping, spraying, etc. Due to random pose changes in the workpieces on the production line, demand for online 3D modeling has increased and many researchers have focused on it. However, online 3D modeling has not been entirely determined due to the occlusion of uncertain dynamic objects that disturb the modeling process. In this study, we propose an online 3D modeling method under uncertain dynamic occlusion based on a binocular camera. Firstly, focusing on uncertain dynamic objects, a novel dynamic object segmentation method based on motion consistency constraints is proposed, which achieves segmentation by random sampling and poses hypotheses clustering without any prior knowledge about objects. Then, in order to better register the incomplete point cloud of each frame, an optimization method based on local constraints of overlapping view regions and a global loop closure is introduced. It establishes constraints in covisibility regions between adjacent frames to optimize the registration of each frame, and it also establishes them between the global closed-loop frames to jointly optimize the entire 3D model. Finally, a confirmatory experimental workspace is designed and built to verify and evaluate our method. Our method achieves online 3D modeling under uncertain dynamic occlusion and acquires an entire 3D model. The pose measurement results further reflect the effectiveness. Full article

(This article belongs to the Special Issue Recent Advances in Robotics and Intelligent Mechatronics Systems)

► Show Figures

Figure 1

Figure 1
An illustration of our pipeline. (1) Vision system captures left and right images. Stereo rectification is first performed to align the epipolar lines of the left and right images by rotating two images. The depth of the points can be calculated by disparity. (2) When two consecutive frames arrive, data association is established by Grid-based Motion Statistics (GMS) and image block correlation. Then, object segmentation is carried out by random sampling and pose hypotheses clustering. (3) The segmentation results, which include initial poses and 2D-3D point pairs, are sent to the 3D modeling part. It includes the target object tracking module and the dynamic object tracking module. The target object poses (namely, camera pose) and the dynamic object motions will be estimated and optimized. (4) Finally, local BA and global BA are carried out to optimize and update these variables. The outputs of our pipeline are poses of target and dynamic objects, a 3D model, and 3D point cloud of dynamic objects. Full article ">Figure 2
Notation and coordinate system. The big black and green circle represent the 3D point of <math display="inline"><semantics> <mrow> <msub> <mrow> <mi>O</mi> </mrow> <mrow> <mn>1</mn> </mrow> </msub> </mrow> </semantics></math> and 3D point of <math display="inline"><semantics> <mrow> <msub> <mrow> <mi>O</mi> </mrow> <mrow> <mi>l</mi> </mrow> </msub> </mrow> </semantics></math> in the <math display="inline"><semantics> <mrow> <msub> <mrow> <mi>O</mi> </mrow> <mrow> <mn>1</mn> </mrow> </msub> </mrow> </semantics></math> coordinate. Small black and green circles denote the corresponding observations in images of adjacent frames. We establish the camera coordinate system in the camera optical center and the object coordinate system in the centroid of the object. Full article ">Figure 3
The structure of the workspace. The vision system mainly included two symmetrically arranged high-resolution cameras and an auxiliary texture device in the middle. The workpiece was placed in the white slot on the turntable. The turntable and rotating shaft could change the pose of the object. Tools could be clamped by the end effector and be moved by the robotic arm. Full article ">Figure 4
The results of segmentation and clustering. In the grayscale images, the last and current frame represent two adjacent frames and the circles indicate the matched points. The circles with different colors indicate that the feature points are located on different objects. In the three-dimensional pose parameter space, the translation vectors with the real scale are demonstrated. The shade of the color means the density is different. These two cluster centers correspond to two objects with different motions. Full article ">Figure 5
The output of our pipeline. (1) The trajectory of the camera is represented by the blue line. The magenta pyramid denotes the pose of each frame. For a clearer presentation, the images show the pictures captured by the camera at frames 5, 15, 21, and 32. In each image, the green and blue marks mean the tracked feature points on the target object and moving tool, respectively. (2) 3D model. The blue point cloud in the center of the figure is the 3D model of <math display="inline"><semantics> <mrow> <msub> <mrow> <mi>W</mi> </mrow> <mrow> <mi>a</mi> </mrow> </msub> </mrow> </semantics></math>, surrounded by the moving tool. Here, different colors show the tool observed in each frame. With the rotation of the camera and the movement of the tool, the occluded parts will be observed and modeled in other frames, as shown in the middle of the 3D model with different colors. Here, different colors represent that the occluded parts of other frames are observed in the current frame. Full article ">Figure 6
Pose errors between measured poses and their corresponding ground truths consisting of rotation errors (along the x, y, z-axes in degrees) and translation errors (along the x, y, z-axes in millimeters). We chose the pose measurement results of <math display="inline"><semantics> <mrow> <msub> <mrow> <mi>T</mi> </mrow> <mrow> <mn>1</mn> </mrow> </msub> </mrow> </semantics></math> and <math display="inline"><semantics> <mrow> <msub> <mrow> <mi>T</mi> </mrow> <mrow> <mn>3</mn> </mrow> </msub> </mrow> </semantics></math> to show the error variations. (a,b) are the rotation and translation errors of measurement results of <math display="inline"><semantics> <mrow> <msub> <mrow> <mi>T</mi> </mrow> <mrow> <mn>1</mn> </mrow> </msub> </mrow> </semantics></math>, respectively. (c,d) are the rotation and translation errors of measurement results of <math display="inline"><semantics> <mrow> <msub> <mrow> <mi>T</mi> </mrow> <mrow> <mn>3</mn> </mrow> </msub> </mrow> </semantics></math>. Full article ">Figure 7
3D modeling results obtained by three methods. (a) is the result of the incremental open-loop registering method. The object structure cannot be recovered in (b) because of the disadvantages of ICP. Our method performs well in (c). Full article ">

13 pages, 5150 KiB

Open AccessArticle

Highly Sensitive and Selective Dopamine Determination in Real Samples Using Au Nanoparticles Decorated Marimo-like Graphene Microbead-Based Electrochemical Sensors

by Qichen Tian, Yuanbin She, Yangguang Zhu, Dan Dai, Mingjiao Shi, Wubo Chu, Tao Cai, Hsu-Sheng Tsai, He Li, Nan Jiang, Li Fu, Hongyan Xia, Cheng-Te Lin and Chen Ye

Sensors 2023, 23(5), 2870; https://doi.org/10.3390/s23052870 - 6 Mar 2023

Cited by 8 | Viewed by 2686

Abstract

A sensitive and selective electrochemical dopamine (DA) sensor has been developed using gold nanoparticles decorated marimo-like graphene (Au NP/MG) as a modifier of the glassy carbon electrode (GCE). Marimo-like graphene (MG) was prepared by partial exfoliation on the mesocarbon microbeads (MCMB) through molten [...] Read more.

A sensitive and selective electrochemical dopamine (DA) sensor has been developed using gold nanoparticles decorated marimo-like graphene (Au NP/MG) as a modifier of the glassy carbon electrode (GCE). Marimo-like graphene (MG) was prepared by partial exfoliation on the mesocarbon microbeads (MCMB) through molten KOH intercalation. Characterization via transmission electron microscopy confirmed that the surface of MG is composed of multi-layer graphene nanowalls. The graphene nanowalls structure of MG provided abundant surface area and electroactive sites. Electrochemical properties of Au NP/MG/GCE electrode were investigated by cyclic voltammetry and differential pulse voltammetry techniques. The electrode exhibited high electrochemical activity towards DA oxidation. The oxidation peak current increased linearly in proportion to the DA concentration in a range from 0.02 to 10 μM with a detection limit of 0.016 μM. The detection selectivity was carried out with the presence of 20 μM uric acid in goat serum real samples. This study demonstrated a promising method to fabricate DA sensor-based on MCMB derivatives as electrochemical modifiers. Full article

(This article belongs to the Special Issue State-of-the-Art Electrochemical Biosensors)

► Show Figures

Figure 1

Figure 1
(a) Real photo of Marimo. (b) Schematic illustration of the fabrication of Marimo-like graphene (MG) and Au NP-decorated Marimo-like graphene (Au NP/MG). SEM images of (c) MCMB and (d) MG. (e) Raman spectra and (f) C1s XPS spectra of MCMB and MG. Full article ">Figure 2
(a) Scheme of stripping and collecting graphene nanowalls shell from MG. (b) Raman spectrum and (c) C1s XPS spectrum of graphene nanowalls. (d) TEM image and (e) HRTEM image of MG (inset: SAED pattern; scale bar: 5 nm–1). (f) TEM image of Au NP/MG (inset: SAED pattern; scale bar: 5 nm–1). Full article ">Figure 3
(a) DPV of Au NP/MG/GCE electrodes with and without 10 μM DA in PBS. (b) DPV curves of various modified electrodes with 10 μM DA in PBS. (c) DPV of Au NP/MG/GCE electrodes with various concentrations of MG to the same of AuNP with 10 μM DA in PBS. (d) Au NP/MG/GCE electrodes in 10 mM [Fe(CN)6]3−/4− and 0.1 M KCl electrolyte solution at scan rates from 20 to 250 mV s−1. (e) Linear plots of Iox/Ired versus scan rates (f) DPV of 10 μM DA on Au NP/MG/GCE electrodes with pH. Full article ">Figure 4
(a,b) DPV curves of MG/GCE and Au NP/MG/GCE electrodes with various concentrations of DA, respectively. (c) The corresponding peak current versus DA concentration. (d–f) The repeatability and good anti-interference of Au NP/MG/GCE electrodes. Full article ">

16 pages, 6480 KiB

Open AccessArticle

Cognitive Video Surveillance Management in Hierarchical Edge Computing System with Long Short-Term Memory Model

by Dilshod Bazarov Ravshan Ugli, Jingyeom Kim, Alaelddin F. Y. Mohammed and Joohyung Lee

Sensors 2023, 23(5), 2869; https://doi.org/10.3390/s23052869 - 6 Mar 2023

Cited by 6 | Viewed by 2715

Abstract

Nowadays, deep learning (DL)-based video surveillance services are widely used in smart cities because of their ability to accurately identify and track objects, such as vehicles and pedestrians, in real time. This allows a more efficient traffic management and improved public safety. However, [...] Read more.

Nowadays, deep learning (DL)-based video surveillance services are widely used in smart cities because of their ability to accurately identify and track objects, such as vehicles and pedestrians, in real time. This allows a more efficient traffic management and improved public safety. However, DL-based video surveillance services that require object movement and motion tracking (e.g., for detecting abnormal object behaviors) can consume a substantial amount of computing and memory capacity, such as (i) GPU computing resources for model inference and (ii) GPU memory resources for model loading. This paper presents a novel cognitive video surveillance management with long short-term memory (LSTM) model, denoted as the CogVSM framework. We consider DL-based video surveillance services in a hierarchical edge computing system. The proposed CogVSM forecasts object appearance patterns and smooths out the forecast results needed for an adaptive model release. Here, we aim to reduce standby GPU memory by model release while avoiding unnecessary model reloads for a sudden object appearance. CogVSM hinges on an LSTM-based deep learning architecture explicitly designed for future object appearance pattern prediction by training previous time-series patterns to achieve these objectives. By referring to the result of the LSTM-based prediction, the proposed framework controls the threshold time value in a dynamic manner by using an exponential weighted moving average (EWMA) technique. Comparative evaluations on both simulated and real-world measurement data on the commercial edge devices prove that the LSTM-based model in the CogVSM can achieve a high predictive accuracy, i.e., a root-mean-square error metric of 0.795. In addition, the suggested framework utilizes up to 32.1% less GPU memory than the baseline and 8.9% less than previous work. Full article

(This article belongs to the Special Issue Applications of Video Processing and Computer Vision Sensor II)

► Show Figures

Figure 1

21 pages, 8083 KiB

Open AccessArticle

PointPainting: 3D Object Detection Aided by Semantic Image Information

by Zhentong Gao, Qiantong Wang, Zongxu Pan, Zhenyu Zhai and Hui Long

Sensors 2023, 23(5), 2868; https://doi.org/10.3390/s23052868 - 6 Mar 2023

Cited by 1 | Viewed by 2837

Abstract

A multi-modal 3D object-detection method, based on data from cameras and LiDAR, has become a subject of research interest. PointPainting proposes a method for improving point-cloud-based 3D object detectors using semantic information from RGB images. However, this method still needs to improve on [...] Read more.

A multi-modal 3D object-detection method, based on data from cameras and LiDAR, has become a subject of research interest. PointPainting proposes a method for improving point-cloud-based 3D object detectors using semantic information from RGB images. However, this method still needs to improve on the following two complications: first, there are faulty parts in the image semantic segmentation results, leading to false detections. Second, the commonly used anchor assigner only considers the intersection over union (IoU) between the anchors and ground truth boxes, meaning that some anchors contain few target LiDAR points assigned as positive anchors. In this paper, three improvements are suggested to address these complications. Specifically, a novel weighting strategy is proposed for each anchor in the classification loss. This enables the detector to pay more attention to anchors containing inaccurate semantic information. Then, SegIoU, which incorporates semantic information, instead of IoU, is proposed for the anchor assignment. SegIoU measures the similarity of the semantic information between each anchor and ground truth box, avoiding the defective anchor assignments mentioned above. In addition, a dual-attention module is introduced to enhance the voxelized point cloud. The experiments demonstrate that the proposed modules obtained significant improvements in various methods, consisting of single-stage PointPillars, two-stage SECOND-IoU, anchor-base SECOND, and an anchor-free CenterPoint on the KITTI dataset. Full article

(This article belongs to the Section Sensing and Imaging)

► Show Figures

Figure 1

21 pages, 3245 KiB

Open AccessArticle

Real-Time Evaluation of Perception Uncertainty and Validity Verification of Autonomous Driving

by Mingliang Yang, Kun Jiang, Junze Wen, Liang Peng, Yanding Yang, Hong Wang, Mengmeng Yang, Xinyu Jiao and Diange Yang

Sensors 2023, 23(5), 2867; https://doi.org/10.3390/s23052867 - 6 Mar 2023

Cited by 3 | Viewed by 3175

Abstract

Deep neural network algorithms have achieved impressive performance in object detection. Real-time evaluation of perception uncertainty from deep neural network algorithms is indispensable for safe driving in autonomous vehicles. More research is required to determine how to assess the effectiveness and uncertainty of [...] Read more.

Deep neural network algorithms have achieved impressive performance in object detection. Real-time evaluation of perception uncertainty from deep neural network algorithms is indispensable for safe driving in autonomous vehicles. More research is required to determine how to assess the effectiveness and uncertainty of perception findings in real-time.This paper proposes a novel real-time evaluation method combining multi-source perception fusion and deep ensemble. The effectiveness of single-frame perception results is evaluated in real-time. Then, the spatial uncertainty of the detected objects and influencing factors are analyzed. Finally, the accuracy of spatial uncertainty is validated with the ground truth in the KITTI dataset. The research results show that the evaluation of perception effectiveness can reach 92% accuracy, and a positive correlation with the ground truth is found for both the uncertainty and the error. The spatial uncertainty is related to the distance and occlusion degree of detected objects. Full article

(This article belongs to the Section Vehicular Sensing)

► Show Figures

Figure 1

16 pages, 3012 KiB

Open AccessArticle

A Novel Catheter Distal Contact Force Sensing for Cardiac Ablation Based on Fiber Bragg Grating with Temperature Compensation

by Yuyang Lou, Tianyu Yang, Dong Luo, Jianwei Wu and Yuming Dong

Sensors 2023, 23(5), 2866; https://doi.org/10.3390/s23052866 - 6 Mar 2023

Cited by 3 | Viewed by 3243

Abstract

Objective: To accurately achieve distal contact force, a novel temperature-compensated sensor is developed and integrated into an atrial fibrillation (AF) ablation catheter. Methods: A dual elastomer-based dual FBGs structure is used to differentiate the strain on the two FBGs to achieve temperature compensation, [...] Read more.

Objective: To accurately achieve distal contact force, a novel temperature-compensated sensor is developed and integrated into an atrial fibrillation (AF) ablation catheter. Methods: A dual elastomer-based dual FBGs structure is used to differentiate the strain on the two FBGs to achieve temperature compensation, and the design is optimized and validated by finite element simulation. Results: The designed sensor has a sensitivity of 90.5 pm/N, resolution of 0.01 N, and root–mean–square error (RMSE) of 0.02 N and 0.04 N for dynamic force loading and temperature compensation, respectively, and can stably measure distal contact forces with temperature disturbances. Conclusion: Due to the advantages, i.e., simple structure, easy assembly, low cost, and good robustness, the proposed sensor is suitable for industrial mass production. Full article

(This article belongs to the Special Issue Fiber Optic Sensing and Applications)

► Show Figures

Figure 1

12 pages, 2336 KiB

Open AccessArticle

Bioluminescent-Triple-Enzyme-Based Biosensor with Lactate Dehydrogenase for Non-Invasive Training Load Monitoring

by Galina V. Zhukova, Oleg S. Sutormin, Irina E. Sukovataya, Natalya V. Maznyak and Valentina A. Kratasyuk

Sensors 2023, 23(5), 2865; https://doi.org/10.3390/s23052865 - 6 Mar 2023

Cited by 3 | Viewed by 2287

Abstract

Saliva is one of the most significant biological liquids for the development of a simple, rapid, and non-invasive biosensor for training load diagnostics. There is an opinion that enzymatic bioassays are more relevant in terms of biology. The present paper is aimed at [...] Read more.

Saliva is one of the most significant biological liquids for the development of a simple, rapid, and non-invasive biosensor for training load diagnostics. There is an opinion that enzymatic bioassays are more relevant in terms of biology. The present paper is aimed at investigating the effects of saliva samples, upon altering the lactate content, on the activity of a multi-enzyme, namely lactate dehydrogenase + NAD(P)H:FMN-oxidoreductase + luciferase (LDH + Red + Luc). Optimal enzymes and their substrate composition of the proposed multi-enzyme system were chosen. During the tests of the lactate dependence, the enzymatic bioassay showed good linearity to lactate in the range from 0.05 mM to 0.25 mM. The activity of the LDH + Red + Luc enzyme system was tested in the presence of 20 saliva samples taken from students whose lactate levels were compared by the Barker and Summerson colorimetric method. The results showed a good correlation. The proposed LDH + Red + Luc enzyme system could be a useful, competitive, and non-invasive tool for correct and rapid monitoring of lactate in saliva. This enzyme-based bioassay is easy to use, rapid, and has the potential to deliver point-of-care diagnostics in a cost-effective manner. Full article

(This article belongs to the Special Issue Biosensors for Surveillance and Diagnosis)

► Show Figures

Figure 1

Figure 1
Values of light intensity of the LDH + Red + Luc enzyme system upon the LDH concentrations variation. Full article ">Figure 2
Values of light intensity of the LDH + Red + Luc enzyme system upon the lactate concentration variation. Full article ">Figure 3
Values of light intensity of the LDH + Red + Luc enzyme system upon the NAD+ concentration variation. Full article ">Figure 4
Values of light intensity of the LDH + Red + Luc enzyme system upon the FMN concentration variation. Full article ">Figure 5
Values of light intensity of the LDH + Red + Luc enzyme system upon the C14 concentration variation. Full article ">Figure 6
Activity of the LDH + Red + Luc enzyme system at different pH values. Full article ">Figure 7
Light intensity values of the multi-enzyme system in the presence of the salivary lactate concentrations mentioned in the published articles. Full article ">Figure 8
Effect of the saliva samples on the activity of the multi-enzyme system. Numbers above the mean calculated salivary lactate concentrations. Full article ">

19 pages, 7524 KiB

Open AccessArticle

A Study on the Effectiveness of Deep Learning-Based Anomaly Detection Methods for Breast Ultrasonography

by Changhee Yun, Bomi Eom, Sungjun Park, Chanho Kim, Dohwan Kim, Farah Jabeen, Won Hwa Kim, Hye Jung Kim and Jaeil Kim

Sensors 2023, 23(5), 2864; https://doi.org/10.3390/s23052864 - 6 Mar 2023

Cited by 3 | Viewed by 2452

Abstract

In the medical field, it is delicate to anticipate good performance in using deep learning due to the lack of large-scale training data and class imbalance. In particular, ultrasound, which is a key breast cancer diagnosis method, is delicate to diagnose accurately as [...] Read more.

In the medical field, it is delicate to anticipate good performance in using deep learning due to the lack of large-scale training data and class imbalance. In particular, ultrasound, which is a key breast cancer diagnosis method, is delicate to diagnose accurately as the quality and interpretation of images can vary depending on the operator’s experience and proficiency. Therefore, computer-aided diagnosis technology can facilitate diagnosis by visualizing abnormal information such as tumors and masses in ultrasound images. In this study, we implemented deep learning-based anomaly detection methods for breast ultrasound images and validated their effectiveness in detecting abnormal regions. Herein, we specifically compared the sliced-Wasserstein autoencoder with two representative unsupervised learning models autoencoder and variational autoencoder. The anomalous region detection performance is estimated with the normal region labels. Our experimental results showed that the sliced-Wasserstein autoencoder model outperformed the anomaly detection performance of others. However, anomaly detection using the reconstruction-based approach may not be effective because of the occurrence of numerous false-positive values. In the following studies, reducing these false positives becomes an important challenge. Full article

(This article belongs to the Special Issue Image and Video Processing and Recognition Based on Artificial Intelligence-2nd Edition)

► Show Figures

Figure 1

17 pages, 14109 KiB

Open AccessArticle

A Multi-Channel Ensemble Method for Error-Related Potential Classification Using 2D EEG Images

by Tangfei Tao, Yuxiang Gao, Yaguang Jia, Ruiquan Chen, Ping Li and Guanghua Xu

Sensors 2023, 23(5), 2863; https://doi.org/10.3390/s23052863 - 6 Mar 2023

Cited by 3 | Viewed by 2068

Abstract

An error-related potential (ErrP) occurs when people’s expectations are not consistent with the actual outcome. Accurately detecting ErrP when a human interacts with a BCI is the key to improving these BCI systems. In this paper, we propose a multi-channel method for error-related [...] Read more.

An error-related potential (ErrP) occurs when people’s expectations are not consistent with the actual outcome. Accurately detecting ErrP when a human interacts with a BCI is the key to improving these BCI systems. In this paper, we propose a multi-channel method for error-related potential detection using a 2D convolutional neural network. Multiple channel classifiers are integrated to make final decisions. Specifically, every 1D EEG signal from the anterior cingulate cortex (ACC) is transformed into a 2D waveform image; then, a model named attention-based convolutional neural network (AT-CNN) is proposed to classify it. In addition, we propose a multi-channel ensemble approach to effectively integrate the decisions of each channel classifier. Our proposed ensemble approach can learn the nonlinear relationship between each channel and the label, which obtains 5.27% higher accuracy than the majority voting ensemble approach. We conduct a new experiment and validate our proposed method on a Monitoring Error-Related Potential dataset and our dataset. With the method proposed in this paper, the accuracy, sensitivity and specificity were 86.46%, 72.46% and 90.17%, respectively. The result shows that the AT-CNNs-2D proposed in this paper can effectively improve the accuracy of ErrP classification, and provides new ideas for the study of classification of ErrP brain–computer interfaces. Full article

(This article belongs to the Section Biomedical Sensors)

► Show Figures

Figure 1

Figure 1
Grand average ErrP at channel FCz for correct, erroneous and different (error minus correct) conditions. t = 0 corresponds to the stimulus presentation onset. Full article ">Figure 2
The overview of the proposed multi-channel ensemble method for ErrP detection using 2D EEG images. Full article ">Figure 3
ErrP elicited at channel FCZ for a signal trail. (The x and y axes are shown for the sake of illustration, but actually, the x and y axes are hidden). Full article ">Figure 4
The schematic diagram of AT-CNN identifying a 2D EEG Image. Inside the dotted line is the proposed attention-based convolutional neural network. The convolutional block attention module (CMAM) is added before the fully connected layer and it contains a channel attention module (CAM) and a spatial attention module (SAM). Full article ">Figure 5
The training process of AT-CNN of each channel and the ensemble model. Full article ">Figure 6
Experimental paradigm. Two squares are displayed on the horizontal line in the center of the screen; one is the cursor (green square), the other is the target (red square), and the dotted square is the previous position of the cursor. The interval time of cursor movement is 2 s, and the cursor moves away from the target with the probability of 20%. Full article ">Figure 7
Experimental paradigm of dataset 2. A green square and a white box are displayed on the horizontal line in the center of the screen. With the same probability, the green square appears either on the left side or on the right side of the white box. After a 4 s interval, the green square moves, and there is a probability of 70% that the green square will move to the white box, and a probability of 30% that it will move away from the white box. Full article ">Figure 8
Visualization of the five channel groups chosen in the regions of anterior cingulate cortex. The five channel groups (A–E) illustrate the position of the 64 EEG electrodes on the scalp (small circles, each reporting the standard designation). The channels forming the channel group are highlighted in red. Full article ">

21 pages, 5853 KiB

Open AccessArticle

Abnormal Brain Circuits Characterize Borderline Personality and Mediate the Relationship between Childhood Traumas and Symptoms: A mCCA+jICA and Random Forest Approach

by Alessandro Grecucci, Harold Dadomo, Gerardo Salvato, Gaia Lapomarda, Sara Sorella and Irene Messina

Sensors 2023, 23(5), 2862; https://doi.org/10.3390/s23052862 - 6 Mar 2023

Cited by 8 | Viewed by 3698

Abstract

Borderline personality disorder (BPD) is a severe personality disorder whose neural bases are still unclear. Indeed, previous studies reported inconsistent findings concerning alterations in cortical and subcortical areas. In the present study, we applied for the first time a combination of an unsupervised [...] Read more.

Borderline personality disorder (BPD) is a severe personality disorder whose neural bases are still unclear. Indeed, previous studies reported inconsistent findings concerning alterations in cortical and subcortical areas. In the present study, we applied for the first time a combination of an unsupervised machine learning approach known as multimodal canonical correlation analysis plus joint independent component analysis (mCCA+jICA), in combination with a supervised machine learning approach known as random forest, to possibly find covarying gray matter and white matter (GM-WM) circuits that separate BPD from controls and that are also predictive of this diagnosis. The first analysis was used to decompose the brain into independent circuits of covarying grey and white matter concentrations. The second method was used to develop a predictive model able to correctly classify new unobserved BPD cases based on one or more circuits derived from the first analysis. To this aim, we analyzed the structural images of patients with BPD and matched healthy controls (HCs). The results showed that two GM-WM covarying circuits, including basal ganglia, amygdala, and portions of the temporal lobes and of the orbitofrontal cortex, correctly classified BPD against HC. Notably, these circuits are affected by specific child traumatic experiences (emotional and physical neglect, and physical abuse) and predict symptoms severity in the interpersonal and impulsivity domains. These results support that BPD is characterized by anomalies in both GM and WM circuits related to early traumatic experiences and specific symptoms. Full article

(This article belongs to the Special Issue Brain Activity Monitoring and Measurement)

► Show Figures

Figure 1

Figure 1
Schematic workflow of the analyses. After fusing the two modalities (GM and WM), the brain was decomposed into independent networks of covarying GM-WM (mCCA+jICA). Then Bonferroni corrected t-test was used to assess the networks that differed between groups. Full article ">Figure 2
Predictive model generation. The loading coefficients of the GM-WM networks derived from mCCA+jICA were entered in a random forest classification model to predict BPD diagnosis. Several trees were generated to classify the labels BPD and HC. Each tree voted for that class. The forest then chose the classification having most of the votes. Full article ">Figure 3
A covarying GM-WM network that differs from BPD and HC. Top: violin plots of the loading coefficients for GM and WM of the IC2. Central: network plot showing in green the strength of correlations between components. Bottom: brain plot of positive (increased GM-WM concentration) and negative (decreased GM-WM concentration) of IC2. Full article ">Figure 4
Prediction of new cases. Random forest classification performance metrics. Full article ">Figure 5
Brain plots from random forest analysis. Top: violin plots of the loading coefficients for GM and WM of the IC6. Bottom: brain plot of positive (increased GM - WM concentration) and negative (decreased GM-WM concentration) of IC6. Full article ">Figure 6
Mediation analysis results. Emotional neglect and physical abuse predicted the IC2 (WM) network and that this in turn predicted symptoms in the impulsivity domain. Physical neglect and abuse predicted IC6 (GM) and this in turn predicted interpersonal symptoms. The colors indicate the same indirect effect linking a given child trauma (IV) to a specific symptom (DV) mediated by a specific IC (MV). Full article ">

19 pages, 7913 KiB

Open AccessArticle

Low-Cost Dual-Frequency GNSS Receivers and Antennas for Surveying in Urban Areas

by Veton Hamza, Bojan Stopar, Oskar Sterle and Polona Pavlovčič-Prešeren

Sensors 2023, 23(5), 2861; https://doi.org/10.3390/s23052861 - 6 Mar 2023

Cited by 13 | Viewed by 4437

Abstract

Low-cost dual-frequency global navigation satellite system (GNSS) receivers have recently been tested in various positioning applications. Considering that these sensors can now provide high positioning accuracy at a lower cost, they can be considered an alternative to high-quality geodetic GNSS devices. The main [...] Read more.

Low-cost dual-frequency global navigation satellite system (GNSS) receivers have recently been tested in various positioning applications. Considering that these sensors can now provide high positioning accuracy at a lower cost, they can be considered an alternative to high-quality geodetic GNSS devices. The main objectives of this work were to analyze the differences between geodetic and low-cost calibrated antennas on the quality of observations from low-cost GNSS receivers and to evaluate the performance of low-cost GNSS devices in urban areas. In this study, a simple RTK2B V1 board u-blox ZED-F9P (Thalwil, Switzerland) was tested in combination with a low-cost calibrated and geodetic antenna in open-sky and adverse conditions in urban areas, while a high-quality geodetic GNSS device was used as a reference for comparison. The results of the observation quality check show that low-cost GNSS instruments have a lower carrier-to-noise ratio (C/N₀) than geodetic instruments, especially in the urban areas where the difference is larger and in favor of the geodetic GNSS instruments. The root-mean-square error (RMSE) of the multipath error in the open sky is twice as high for low-cost as for geodetic instruments, while this difference is up to four times greater in urban areas. The use of a geodetic GNSS antenna does not show a significant improvement in the C/N₀ and multipath of low-cost GNSS receivers. However, the ambiguity fix ratio is larger when geodetic antennas are used, with a difference of 1.5% and 18.4% for the open-sky and urban conditions, respectively. It should be noted that float solutions may become more evident when low-cost equipment is used, especially for short sessions and in urban areas with more multipath. In relative positioning mode, low-cost GNSS devices were able to provide horizontal accuracy lower than 10 mm in urban areas in 85% of sessions, while the vertical and spatial accuracy was lower than 15 mm in 82.5% and 77.5% of the sessions, respectively. In the open sky, low-cost GNSS receivers achieve a horizontal, vertical, and spatial accuracy of 5 mm for all sessions considered. In RTK mode, positioning accuracy varies between 10–30 mm in the open-sky and urban areas, while better performance is demonstrated for the former. Full article

(This article belongs to the Special Issue Advances in GNSS Positioning and GNSS Remote Sensing)

► Show Figures

Figure 1

26 pages, 8373 KiB

Open AccessArticle

Swarm Intelligence Internet of Vehicles Approaches for Opportunistic Data Collection and Traffic Engineering in Smart City Waste Management

by Gerald K. Ijemaru, Li-Minn Ang and Kah Phooi Seng

Sensors 2023, 23(5), 2860; https://doi.org/10.3390/s23052860 - 6 Mar 2023

Cited by 15 | Viewed by 2996

Abstract

Recent studies have shown the efficacy of mobile elements in optimizing the energy consumption of sensor nodes. Current data collection approaches for waste management applications focus on exploiting IoT-enabled technologies. However, these techniques are no longer sustainable in the context of smart city [...] Read more.

Recent studies have shown the efficacy of mobile elements in optimizing the energy consumption of sensor nodes. Current data collection approaches for waste management applications focus on exploiting IoT-enabled technologies. However, these techniques are no longer sustainable in the context of smart city (SC) waste management applications due to the emergence of large-scale wireless sensor networks (LS-WSNs) in smart cities with sensor-based big data architectures. This paper proposes an energy-efficient swarm intelligence (SI) Internet of Vehicles (IoV)-based technique for opportunistic data collection and traffic engineering for SC waste management strategies. This is a novel IoV-based architecture exploiting the potential of vehicular networks for SC waste management strategies. The proposed technique involves deploying multiple data collector vehicles (DCVs) traversing the entire network for data gathering via a single-hop transmission. However, employing multiple DCVs comes with additional challenges including costs and network complexity. Thus, this paper proposes analytical-based methods to investigate critical tradeoffs in optimizing energy consumption for big data collection and transmission in an LS-WSN such as (1) finding the optimal number of data collector vehicles (DCVs) required in the network and (2) determining the optimal number of data collection points (DCPs) for the DCVs. These critical issues affect efficient SC waste management and have been overlooked by previous studies exploring waste management strategies. Simulation-based experiments using SI-based routing protocols validate the efficacy of the proposed method in terms of the evaluation metrics. Full article

(This article belongs to the Special Issue Novel Sensing Technologies for Environmental Systems and Sensing in the Wild)

► Show Figures

Figure 1

29 pages, 6191 KiB

Open AccessReview

Natural Intelligence as the Brain of Intelligent Systems

by Mahdi Naghshvarianjahromi, Shiva Kumar and Mohammed Jamal Deen

Sensors 2023, 23(5), 2859; https://doi.org/10.3390/s23052859 - 6 Mar 2023

Cited by 3 | Viewed by 2287

Abstract

This article discusses the concept and applications of cognitive dynamic systems (CDS), which are a type of intelligent system inspired by the brain. There are two branches of CDS, one for linear and Gaussian environments (LGEs), such as cognitive radio and cognitive radar, [...] Read more.

This article discusses the concept and applications of cognitive dynamic systems (CDS), which are a type of intelligent system inspired by the brain. There are two branches of CDS, one for linear and Gaussian environments (LGEs), such as cognitive radio and cognitive radar, and another one for non-Gaussian and nonlinear environments (NGNLEs), such as cyber processing in smart systems. Both branches use the same principle, called the perception action cycle (PAC), to make decisions. The focus of this review is on the applications of CDS, including cognitive radios, cognitive radar, cognitive control, cyber security, self-driving cars, and smart grids for LGEs. For NGNLEs, the article reviews the use of CDS in smart e-healthcare applications and software-defined optical communication systems (SDOCS), such as smart fiber optic links. The results of implementing CDS in these systems are very promising, with improved accuracy, performance, and lower computational costs. For example, CDS implementation in cognitive radars achieved a range estimation error that is as good as 0.47 (m) and a velocity estimation error of 3.30 (m/s), outperforming traditional active radars. Similarly, CDS implementation in smart fiber optic links improved the quality factor by 7 dB and the maximum achievable data rate by 43% compared to those of other mitigation techniques. Full article

(This article belongs to the Section Intelligent Sensors)

► Show Figures

Figure 1

26 pages, 27778 KiB

Open AccessArticle

Home Chimney Pinwheels (HCP) as Steh and Remote Monitoring for Smart Building IoT and WSN Applications

by Ajibike Eunice Akin-Ponnle, Paulo Capitão, Ricardo Torres and Nuno Borges Carvalho

Sensors 2023, 23(5), 2858; https://doi.org/10.3390/s23052858 - 6 Mar 2023

Cited by 5 | Viewed by 2814

Abstract

Smart, and ultra-low energy consuming Internet of Things (IoTs), wireless sensor networks (WSN), and autonomous devices are being deployed to smart buildings and cities, which require continuous power supply, whereas battery usage has accompanying environmental problems, coupled with additional maintenance cost. We present [...] Read more.

Smart, and ultra-low energy consuming Internet of Things (IoTs), wireless sensor networks (WSN), and autonomous devices are being deployed to smart buildings and cities, which require continuous power supply, whereas battery usage has accompanying environmental problems, coupled with additional maintenance cost. We present Home Chimney Pinwheels (HCP) as the Smart Turbine Energy Harvester (STEH) for wind; and Cloud-based remote monitoring of its output data. The HCP commonly serves as an external cap to home chimney exhaust outlets; they have very low inertia to wind; and are available on the rooftops of some buildings. Here, an electromagnetic converter adapted from a brushless DC motor was mechanically fastened to the circular base of an 18-blade HCP. In simulated wind, and rooftop experiments, an output voltage of 0.3 V to 16 V was realised for a wind speed between 0.6 to 16 km/h. This is sufficient to operate low-power IoT devices deployed around a smart city. The harvester was connected to a power management unit and its output data was remotely monitored via the IoT analytic Cloud platform “ThingSpeak” by means of LoRa transceivers, serving as sensors; while also obtaining supply from the harvester. The HCP can be a battery-less “stand-alone” low-cost STEH, with no grid connection, and can be installed as attachments to IoT or wireless sensors nodes in smart buildings and cities. Full article

(This article belongs to the Collection Wireless Sensor Networks towards the Internet of Things)

► Show Figures

Figure 1

11 pages, 2951 KiB

Open AccessCommunication

A Dew-Condensation Sensor Exploiting Local Variations in the Relative Refractive Index on the Dew-Friendly Surface of a Waveguide

by Subin Hwa, Eun-Seon Sim, Jun-Hee Na, Ik-Hoon Jang, Jin-Hyuk Kwon and Min-Hoi Kim

Sensors 2023, 23(5), 2857; https://doi.org/10.3390/s23052857 - 6 Mar 2023

Viewed by 2209

Abstract

We propose a sensor technology for detecting dew condensation, which exploits a variation in the relative refractive index on the dew-friendly surface of an optical waveguide. The dew-condensation sensor is composed of a laser, waveguide, medium (i.e., filling material for the waveguide), and [...] Read more.

We propose a sensor technology for detecting dew condensation, which exploits a variation in the relative refractive index on the dew-friendly surface of an optical waveguide. The dew-condensation sensor is composed of a laser, waveguide, medium (i.e., filling material for the waveguide), and photodiode. The formation of dewdrops on the waveguide surface causes local increases in the relative refractive index accompanied by the transmission of the incident light rays, hence reducing the light intensity inside the waveguide. In particular, the dew-friendly surface of the waveguide is obtained by filling the interior of the waveguide with liquid H₂O, i.e., water. A geometric design for the sensor was first carried out considering the curvature of the waveguide and the incident angles of the light rays. Moreover, the optical suitability of waveguide media with various absolute refractive indices, i.e., water, air, oil, and glass, were evaluated through simulation tests. In actual experiments, the sensor with the water-filled waveguide displayed a wider gap between the measured photocurrent levels under conditions with and without dew, than those with the air- and glass-filled waveguides, as a result of the relatively high specific heat of the water. The sensor with the water-filled waveguide exhibited excellent accuracy and repeatability as well. Full article

(This article belongs to the Section Physical Sensors)

► Show Figures

Figure 1

Figure 1
(a) Schematic diagram of the sensor, (b) key denotations for the constituent media and surfaces of the sensor as well as those for the absolute refractive indices of the media, and (c) the path difference between the incident light rays, with and without dew. Full article ">Figure 2
The geometry of the waveguide with curvature and the design strategy in which the condition of the total reflection at the external interface is considered. Full article ">Figure 3
Optical simulation results (the light wavelength: 650 nm). Optical paths obtained with various internal media for the waveguide, i.e., (a) glass, (b) oil, (c) water, and (d) air. Full article ">Figure 4
In the optical simulation results: (a) variations in the intensity according to the simulated dew condensation for the various internal media, and (b) the regions on the waveguide surface where the incident light rays are transmitted under conditions with dew for the various internal media. Full article ">Figure 5
Responses of the actual sensors with the water-, glass-, and air-filled waveguides to dew condensation (the laser light wavelength: 650 nm). Full article ">Figure 6
The photos of the (a) air-filled waveguide, (b) glass-filled waveguide, and (c) water-filled waveguide before the dew formation (‘initial’), after the dew formation (‘with dew’), and after the dew evaporation (‘without dew’) in sequence. Full article ">Figure 7
Responses of the five actual sensors with the water-filled waveguides to dew condensation. Full article ">

14 pages, 4036 KiB

Open AccessArticle

Identification and Classification of Small Sample Desert Grassland Vegetation Communities Based on Dynamic Graph Convolution and UAV Hyperspectral Imagery

by Tao Zhang, Yuge Bi, Xiangbing Zhu and Xinchao Gao

Sensors 2023, 23(5), 2856; https://doi.org/10.3390/s23052856 - 6 Mar 2023

Cited by 6 | Viewed by 1868

Abstract

Desert steppes are the last barrier to protecting the steppe ecosystem. However, existing grassland monitoring methods still mainly use traditional monitoring methods, which have certain limitations in the monitoring process. Additionally, the existing deep learning classification models of desert and grassland still use [...] Read more.

Desert steppes are the last barrier to protecting the steppe ecosystem. However, existing grassland monitoring methods still mainly use traditional monitoring methods, which have certain limitations in the monitoring process. Additionally, the existing deep learning classification models of desert and grassland still use traditional convolutional neural networks for classification, which cannot adapt to the classification task of irregular ground objects, which limits the classification performance of the model. To address the above problems, this paper uses a UAV hyperspectral remote sensing platform for data acquisition and proposes a spatial neighborhood dynamic graph convolution network (SN_DGCN) for degraded grassland vegetation community classification. The results show that the proposed classification model had the highest classification accuracy compared to the seven classification models of MLP, 1DCNN, 2DCNN, 3DCNN, Resnet18, Densenet121, and SN_GCN; its OA, AA, and kappa were 97.13%, 96.50%, and 96.05% in the case of only 10 samples per class of features, respectively; The classification performance was stable under different numbers of training samples, had better generalization ability in the classification task of small samples, and was more effective for the classification task of irregular features. Meanwhile, the latest desert grassland classification models were also compared, which fully demonstrated the superior classification performance of the proposed model in this paper. The proposed model provides a new method for the classification of vegetation communities in desert grasslands, which is helpful for the management and restoration of desert steppes. Full article

(This article belongs to the Section Remote Sensors)

► Show Figures

Figure 1

19 pages, 3939 KiB

Open AccessArticle

Multiple Dipole Source Position and Orientation Estimation Using Non-Invasive EEG-like Signals

by Saina Namazifard and Kamesh Subbarao

Sensors 2023, 23(5), 2855; https://doi.org/10.3390/s23052855 - 6 Mar 2023

Cited by 8 | Viewed by 2082

Abstract

The problem of precisely estimating the position and orientation of multiple dipoles using synthetic EEG signals is considered in this paper. After determining a proper forward model, a nonlinear constrained optimization problem with regularization is solved, and the results are compared with a [...] Read more.

The problem of precisely estimating the position and orientation of multiple dipoles using synthetic EEG signals is considered in this paper. After determining a proper forward model, a nonlinear constrained optimization problem with regularization is solved, and the results are compared with a widely used research code, namely EEGLAB. A thorough sensitivity analysis of the estimation algorithm to the parameters (such as the number of samples and sensors) in the assumed signal measurement model is conducted. To confirm the efficacy of the proposed source identification algorithm on any category of data sets, three different kinds of data-synthetic model data, visually evoked clinical EEG data, and seizure clinical EEG data are used. Furthermore, the algorithm is tested on both the spherical head model and the realistic head model based on the MNI coordinates. The numerical results and comparisons with the EEGLAB show very good agreement, with little pre-processing required for the acquired data. Full article

(This article belongs to the Section Biomedical Sensors)

► Show Figures

Figure 1

16 pages, 1559 KiB

Open AccessArticle

Morphological Autoencoders for Beat-by-Beat Atrial Fibrillation Detection Using Single-Lead ECG

by Rafael Silva, Ana Fred and Hugo Plácido da Silva

Sensors 2023, 23(5), 2854; https://doi.org/10.3390/s23052854 - 6 Mar 2023

Cited by 3 | Viewed by 2505

Abstract

Engineered feature extraction can compromise the ability of Atrial Fibrillation (AFib) detection algorithms to deliver near real-time results. Autoencoders (AEs) can be used as an automatic feature extraction tool, tailoring the resulting features to a specific classification task. By coupling an encoder to [...] Read more.

Engineered feature extraction can compromise the ability of Atrial Fibrillation (AFib) detection algorithms to deliver near real-time results. Autoencoders (AEs) can be used as an automatic feature extraction tool, tailoring the resulting features to a specific classification task. By coupling an encoder to a classifier, it is possible to reduce the dimension of the Electrocardiogram (ECG) heartbeat waveforms and classify them. In this work we show that morphological features extracted using a Sparse AE are sufficient to distinguish AFib from Normal Sinus Rhythm (NSR) beats. In addition to the morphological features, rhythm information was included in the model using a proposed short-term feature called Local Change of Successive Differences (LCSD). Using single-lead ECG recordings from two referenced public databases, and with features from the AE, the model was able to achieve an F1-score of 88.8%. These results show that morphological features appear to be a distinct and sufficient factor for detecting AFib in ECG recordings, especially when designed for patient-specific applications. This is an advantage over state-of-the-art algorithms that need longer acquisition times to extract engineered rhythm features, which also requires careful preprocessing steps. To the best of our knowledge, this is the first work that presents a near real-time morphological approach for AFib detection under naturalistic ECG acquisition with a mobile device. Full article

(This article belongs to the Special Issue Advanced Machine Intelligence for Biomedical Signal Processing)

► Show Figures

Figure 1

23 pages, 20417 KiB

Open AccessArticle

Sign2Pose: A Pose-Based Approach for Gloss Prediction Using a Transformer Model

by Jennifer Eunice, Andrew J, Yuichi Sei and D. Jude Hemanth

Sensors 2023, 23(5), 2853; https://doi.org/10.3390/s23052853 - 6 Mar 2023

Cited by 12 | Viewed by 3591

Abstract

Word-level sign language recognition (WSLR) is the backbone for continuous sign language recognition (CSLR) that infers glosses from sign videos. Finding the relevant gloss from the sign sequence and detecting explicit boundaries of the glosses from sign videos is a persistent challenge. In [...] Read more.

Word-level sign language recognition (WSLR) is the backbone for continuous sign language recognition (CSLR) that infers glosses from sign videos. Finding the relevant gloss from the sign sequence and detecting explicit boundaries of the glosses from sign videos is a persistent challenge. In this paper, we propose a systematic approach for gloss prediction in WLSR using the Sign2Pose Gloss prediction transformer model. The primary goal of this work is to enhance WLSR’s gloss prediction accuracy with reduced time and computational overhead. The proposed approach uses hand-crafted features rather than automated feature extraction, which is computationally expensive and less accurate. A modified key frame extraction technique is proposed that uses histogram difference and Euclidean distance metrics to select and drop redundant frames. To enhance the model’s generalization ability, pose vector augmentation using perspective transformation along with joint angle rotation is performed. Further, for normalization, we employed YOLOv3 (You Only Look Once) to detect the signing space and track the hand gestures of the signers in the frames. The proposed model experiments on WLASL datasets achieved the top 1% recognition accuracy of 80.9% in WLASL100 and 64.21% in WLASL300. The performance of the proposed model surpasses state-of-the-art approaches. The integration of key frame extraction, augmentation, and pose estimation improved the performance of the proposed gloss prediction model by increasing the model’s precision in locating minor variations in their body posture. We observed that introducing YOLOv3 improved gloss prediction accuracy and helped prevent model overfitting. Overall, the proposed model showed 17% improved performance in the WLASL 100 dataset. Full article

(This article belongs to the Special Issue Sensor Systems for Gesture Recognition II)

► Show Figures

Figure 1

23 pages, 14661 KiB

Open AccessArticle

A Non-Equal Time Interval Incremental Motion Prediction Method for Maritime Autonomous Surface Ships

by Zhijie Zhou, Haixiang Xu, Hui Feng and Wenjuan Li

Sensors 2023, 23(5), 2852; https://doi.org/10.3390/s23052852 - 6 Mar 2023

Cited by 1 | Viewed by 1755

Abstract

Recent technological advancements facilitate the autonomous navigation of maritime surface ships. The accurate data given by a range of various sensors serve as the primary assurance of a voyage’s safety. Nevertheless, as sensors have various sample rates, they cannot obtain information at the [...] Read more.

Recent technological advancements facilitate the autonomous navigation of maritime surface ships. The accurate data given by a range of various sensors serve as the primary assurance of a voyage’s safety. Nevertheless, as sensors have various sample rates, they cannot obtain information at the same time. Fusion decreases the accuracy and reliability of perceptual data if different sensor sample rates are not taken into account. Hence, it is helpful to increase the quality of the fusion information to precisely anticipate the motion status of ships at the sampling time of each sensor. This paper proposes a non-equal time interval incremental prediction method. In this method, the high dimensionality of the estimated state and nonlinearity of the kinematic equation are taken into consideration. First, the cubature Kalman filter is employed to estimate a ship’s motion at equal intervals based on the ship’s kinematic equation. Next, a ship motion state predictor based on a long short-term memory network structure is created, using the increment and time interval of the historical estimation sequence as the network input and the increment of the motion state at the projected time as the network output. The suggested technique can lessen the effect of the speed difference between the test set and the training set on the prediction accuracy compared with the traditional long short-term memory prediction method. Finally, comparison experiments are carried out to validate the precision and effectiveness of the proposed approach. The experimental results show that the root-mean-square error coefficient of the prediction error is decreased on average by roughly 78% for various modes and speeds when compared with the conventional non-incremental long short-term memory prediction approach. Additionally, the proposed prediction technology and the traditional approach have virtually the same algorithm times, which may fulfill the real engineering requirements. Full article

(This article belongs to the Special Issue Algorithms, Systems and Applications of Smart Sensor Networks)

► Show Figures

Figure 1

17 pages, 3557 KiB

Open AccessArticle

Detecting Grapevine Virus Infections in Red and White Winegrape Canopies Using Proximal Hyperspectral Sensing

by Yeniu Mickey Wang, Bertram Ostendorf and Vinay Pagay

Sensors 2023, 23(5), 2851; https://doi.org/10.3390/s23052851 - 6 Mar 2023

Cited by 7 | Viewed by 2900

Abstract

Grapevine virus-associated disease such as grapevine leafroll disease (GLD) affects grapevine health worldwide. Current diagnostic methods are either highly costly (laboratory-based diagnostics) or can be unreliable (visual assessments). Hyperspectral sensing technology is capable of measuring leaf reflectance spectra that can be used for [...] Read more.

Grapevine virus-associated disease such as grapevine leafroll disease (GLD) affects grapevine health worldwide. Current diagnostic methods are either highly costly (laboratory-based diagnostics) or can be unreliable (visual assessments). Hyperspectral sensing technology is capable of measuring leaf reflectance spectra that can be used for the non-destructive and rapid detection of plant diseases. The present study used proximal hyperspectral sensing to detect virus infection in Pinot Noir (red-berried winegrape cultivar) and Chardonnay (white-berried winegrape cultivar) grapevines. Spectral data were collected throughout the grape growing season at six timepoints per cultivar. Partial least squares-discriminant analysis (PLS-DA) was used to build a predictive model of the presence or absence of GLD. The temporal change of canopy spectral reflectance showed that the harvest timepoint had the best prediction result. Prediction accuracies of 96% and 76% were achieved for Pinot Noir and Chardonnay, respectively. Our results provide valuable information on the optimal time for GLD detection. This hyperspectral method can also be deployed on mobile platforms including ground-based vehicles and unmanned aerial vehicles (UAV) for large-scale disease surveillance in vineyards. Full article

(This article belongs to the Special Issue Methodologies Used in Hyperspectral Remote Sensing in Agriculture)

► Show Figures

Figure 1

Figure 1
Spectral data measurement. The yellow shaded area shows the circular field of view of diameter of approx. 20 cm based on the approx. 0.5 m horizontal measurement distance. Full article ">Figure 2
Spectral data pre-processing steps for Pinot Noir at the February 2021 timepoint. The raw spectral data is smoothed using the SavGol algorithm with seven bandwidths, then normalised using the standard normal variate algorithm and, lastly, scaled using the Mean Centring algorithm for 750 bands. Full article ">Figure 3
PLS-DA threshold determination. (a) ROC curve for disease class; (b) plot of PLS predicted value for disease class for Chardonnay in March 2021. The threshold was 0.698 in this model. Full article ">Figure 4
The GLD-infected vines at different development stages. Chardonnay in (a) November—flowering stage (EL-17); (b) December— pea-size berries; (c) February—veraison; (d) April—post-harvest. Pinot Noir in (e) November—flowering; (f) December—pea-size berries; (g) February—veraison; (h) April—post-harvest. Full article ">Figure 5
The difference of normalised averaged spectral reflectance of diseased to healthy vines (value at 0) for each timepoint. Full article ">Figure 6
The combined violin plot and box plot of the predicted value of disease from the PLS-DA model for (a) Chardonnay, and (b) Pinot Noir at each timepoint. The larger the value, the higher the probability of a sample belonging to a diseased vine, and the smaller the value, the lower the probability of a diseased vine. In this binary model, a low value means the sample more likely belongs to a healthy vine. The greater the separation between actual disease and healthy samples, the better the performance of the classification model. Full article ">Figure 7
The matrix of MCC from each model prediction of other timepoints for (a) Chardonnay, and (b) Pinot Noir. The bold cells are the self-predicted results, which are equal to the calibration results. The colour score ranges from red (low value) to yellow (medium) to green (high value). NA means √((TP + FP)∙(TP + FN)∙(TN + FP)∙(TN + FN)) equal to zero that cannot be divided. Full article ">

6 pages, 1805 KiB

Open AccessCommunication

Epoxy-Coated Side-Polished Fiber-Optic Temperature Sensor for Cryogenic Conditions

by Umesh Sampath and Minho Song

Sensors 2023, 23(5), 2850; https://doi.org/10.3390/s23052850 - 6 Mar 2023

Viewed by 1754

Abstract

We propose coating side-polished optical fiber (SPF) with epoxy polymer to form a fiber-optic sensor for cryogenic temperature measuring applications. The thermo-optic effect of the epoxy polymer coating layer enhances the interaction between the SPF evanescent field and surrounding medium, considerably improving the [...] Read more.

We propose coating side-polished optical fiber (SPF) with epoxy polymer to form a fiber-optic sensor for cryogenic temperature measuring applications. The thermo-optic effect of the epoxy polymer coating layer enhances the interaction between the SPF evanescent field and surrounding medium, considerably improving the temperature sensitivity and robustness of the sensor head in a very low-temperature environment. In tests, due to the evanescent field–polymer coating interlinkage, transmitted optical intensity variation of 5 dB and an average sensitivity of

- 0.024

dB/K were obtained in the 90–298 K range. Full article

(This article belongs to the Special Issue Applications of Optical Fiber Sensors and Measurement Systems)

► Show Figures

Figure 1

16 pages, 770 KiB

Open AccessArticle

Self-Excited Microcantilever with Higher Mode Using Band-Pass Filter

by Yuji Hyodo and Hiroshi Yabuno

Sensors 2023, 23(5), 2849; https://doi.org/10.3390/s23052849 - 6 Mar 2023

Cited by 1 | Viewed by 1915

Abstract

Microresonators have a variety of scientific and industrial applications. The measurement methods based on the natural frequency shift of a resonator have been studied for a wide range of applications, including the detection of the microscopic mass and measurements of viscosity and stiffness. [...] Read more.

Microresonators have a variety of scientific and industrial applications. The measurement methods based on the natural frequency shift of a resonator have been studied for a wide range of applications, including the detection of the microscopic mass and measurements of viscosity and stiffness. A higher natural frequency of the resonator realizes an increase in the sensitivity and a higher-frequency response of the sensors. In the present study, by utilizing the resonance of a higher mode, we propose a method to produce the self-excited oscillation with a higher natural frequency without downsizing the resonator. We establish the feedback control signal for the self-excited oscillation using the band-pass filter so that the signal consists of only the frequency corresponding to the desired excitation mode. It results that careful position setting of the sensor for constructing a feedback signal, which is needed in the method based on the mode shape, is not necessary. By the theoretical analysis of the equations governing the dynamics of the resonator coupled with the band-pass filter, it is clarified that the self-excited oscillation is produced with the second mode. Furthermore, the validity of the proposed method is experimentally confirmed by an apparatus using a microcantilever. Full article

(This article belongs to the Special Issue Feature Papers in Physical Sensors 2022)

► Show Figures

Figure 1

12 pages, 1436 KiB

Open AccessArticle

Pre-Trained Joint Model for Intent Classification and Slot Filling with Semantic Feature Fusion

by Yan Chen and Zhenghang Luo

Sensors 2023, 23(5), 2848; https://doi.org/10.3390/s23052848 - 6 Mar 2023

Cited by 5 | Viewed by 3443

Abstract

The comprehension of spoken language is a crucial aspect of dialogue systems, encompassing two fundamental tasks: intent classification and slot filling. Currently, the joint modeling approach for these two tasks has emerged as the dominant method in spoken language understanding modeling. However, the [...] Read more.

The comprehension of spoken language is a crucial aspect of dialogue systems, encompassing two fundamental tasks: intent classification and slot filling. Currently, the joint modeling approach for these two tasks has emerged as the dominant method in spoken language understanding modeling. However, the existing joint models have limitations in terms of their relevancy and utilization of contextual semantic features between the multiple tasks. To address these limitations, a joint model based on BERT and semantic fusion (JMBSF) is proposed. The model employs pre-trained BERT to extract semantic features and utilizes semantic fusion to associate and integrate this information. The results of experiments on two benchmark datasets, ATIS and Snips, in spoken language comprehension demonstrate that the proposed JMBSF model attains 98.80% and 99.71% intent classification accuracy, 98.25% and 97.24% slot-filling F1-score, and 93.40% and 93.57% sentence accuracy, respectively. These results reveal a significant improvement compared to other joint models. Furthermore, comprehensive ablation studies affirm the effectiveness of each component in the design of JMBSF. Full article

(This article belongs to the Section Intelligent Sensors)

► Show Figures

Figure 1

4 pages, 195 KiB

Open AccessEditorial

Editorial–Special Issue on “Sensor Technology for Enhancing Training and Performance in Sport”

by Pui Wah Kong

Sensors 2023, 23(5), 2847; https://doi.org/10.3390/s23052847 - 6 Mar 2023

Cited by 4 | Viewed by 1761

Abstract

Sensor technology opens up exciting opportunities for sports [...] Full article

(This article belongs to the Special Issue Sensor Technology for Enhancing Training and Performance in Sport)

13 pages, 4017 KiB

Open AccessArticle

An Adaptive Pedaling Assistive Device for Asymmetric Torque Assistant in Cycling

by Jesse Lozinski, Seyed Hamidreza Heidary, Scott C. E. Brandon and Amin Komeili

Sensors 2023, 23(5), 2846; https://doi.org/10.3390/s23052846 - 6 Mar 2023

Cited by 4 | Viewed by 2698

Abstract

Dynamic loads have short and long-term effects in the rehabilitation of lower limb joints. However, an effective exercise program for lower limb rehabilitation has been debated for a long time. Cycling ergometers were instrumented and used as a tool to mechanically load the [...] Read more.

Dynamic loads have short and long-term effects in the rehabilitation of lower limb joints. However, an effective exercise program for lower limb rehabilitation has been debated for a long time. Cycling ergometers were instrumented and used as a tool to mechanically load the lower limbs and track the joint mechano-physiological response in rehabilitation programs. Current cycling ergometers apply symmetrical loading to the limbs, which may not reflect the actual load-bearing capacity of each limb, as in Parkinson’s and Multiple Sclerosis diseases. Therefore, the present study aimed to develop a new cycling ergometer capable of applying asymmetric loads to the limbs and validate its function using human tests. The instrumented force sensor and crank position sensing system recorded the kinetics and kinematics of pedaling. This information was used to apply an asymmetric assistive torque only to the target leg using an electric motor. The performance of the proposed cycling ergometer was studied during a cycling task at three different intensities. It was shown that the proposed device reduced the pedaling force of the target leg by 19% to 40%, depending on the exercise intensity. This reduction in pedal force caused a significant reduction in the muscle activity of the target leg (p < 0.001), without affecting the muscle activity of the non-target leg. These results demonstrated that the proposed cycling ergometer device is capable of applying asymmetric loading to lower limbs, and thus has the potential to improve the outcome of exercise interventions in patients with asymmetric function in lower limbs. Full article

(This article belongs to the Special Issue Sensors and Actuators for Wearable and Implantable Devices)

► Show Figures

Figure 1

19 pages, 4136 KiB

Open AccessArticle

LiDAR-as-Camera for End-to-End Driving

by Ardi Tampuu, Romet Aidla, Jan Aare van Gent and Tambet Matiisen

Sensors 2023, 23(5), 2845; https://doi.org/10.3390/s23052845 - 6 Mar 2023

Cited by 9 | Viewed by 3824

Abstract

The core task of any autonomous driving system is to transform sensory inputs into driving commands. In end-to-end driving, this is achieved via a neural network, with one or multiple cameras as the most commonly used input and low-level driving commands, e.g., steering [...] Read more.

The core task of any autonomous driving system is to transform sensory inputs into driving commands. In end-to-end driving, this is achieved via a neural network, with one or multiple cameras as the most commonly used input and low-level driving commands, e.g., steering angle, as output. However, simulation studies have shown that depth-sensing can make the end-to-end driving task easier. On a real car, combining depth and visual information can be challenging due to the difficulty of obtaining good spatial and temporal alignment of the sensors. To alleviate alignment problems, Ouster LiDARs can output surround-view LiDAR images with depth, intensity, and ambient radiation channels. These measurements originate from the same sensor, rendering them perfectly aligned in time and space. The main goal of our study is to investigate how useful such images are as inputs to a self-driving neural network. We demonstrate that such LiDAR images are sufficient for the real-car road-following task. Models using these images as input perform at least as well as camera-based models in the tested conditions. Moreover, LiDAR images are less sensitive to weather conditions and lead to better generalization. In a secondary research direction, we reveal that the temporal smoothness of off-policy prediction sequences correlates with the actual on-policy driving ability equally well as the commonly used mean absolute error. Full article

(This article belongs to the Special Issue Advances in Sensor Related Technologies for Autonomous Driving)

► Show Figures

Figure 1

Figure 1
The location of sensors used in this work. There are other sensors on the vehicle not illustrated here. Full article ">Figure 2
Input modalities. The red box marks the area used as model input. Top: surround view LiDAR image, with red: intensity, blue: depth, and green: ambient. Bottom: 120-degree FOV camera. Full article ">Figure 3
The modified PilotNet architecture. Each box represents the output from a layer, with the first box corresponding to the input of size (264, 68, 3). The model consists of 5 convolutional layers and 4 fully connected layers. The flattening operation is not made visible here. See the filter sizes, usage of batch normalization, and activation functions in <a href="#sensors-23-02845-t001" class="html-table">Table 1</a>. Full article ">Figure 4
Safety-driver interventions in the experiments where the test track was not included in the training set. Interventions from 3 test runs with different versions of the same model and from both driving directions are overlaid on one map. Interventions due to traffic are not filtered out from these maps, unlike in <a href="#sensors-23-02845-t002" class="html-table">Table 2</a>. Left: camera models v1–v3 (first 3 rows of <a href="#sensors-23-02845-t002" class="html-table">Table 2</a>). Middle: LiDAR models v1–v3 (rows 4–6 of <a href="#sensors-23-02845-t002" class="html-table">Table 2</a>). Right: an example of a situation where the safety driver has to take over due to traffic. Such situations are not counted as interventions in <a href="#sensors-23-02845-t002" class="html-table">Table 2</a>. Full article ">Figure A1
LiDAR and camera images in summer, autumn, and winter (from top to down for LiDAR, left to right for the camera). The area used for model inputs is marked with a red rectangle. In LiDAR images, the red channel corresponds to intensity, green to depth, and blue to ambient radiation. Full article ">Figure A2
LiDAR channels at the same location across the three seasons, in order from top down: summer, autumn, winter. (a) In the intensity channel, we see a significant difference in how the road itself looks, while vegetation is surprisingly similar despite deciduous plants having no leaves in autumn and winter. (b) Depth image looks stable across seasons, but rather uninformative, as road and low vegetation areas are hard to discern. (c) Ambient radiation images vary strongly in brightness across the seasons, while also displaying strong noise. The noise looks akin to white noise or salt-and-pepper noise and authors do not know its cause. Full article ">Figure A2 Cont.
LiDAR channels at the same location across the three seasons, in order from top down: summer, autumn, winter. (a) In the intensity channel, we see a significant difference in how the road itself looks, while vegetation is surprisingly similar despite deciduous plants having no leaves in autumn and winter. (b) Depth image looks stable across seasons, but rather uninformative, as road and low vegetation areas are hard to discern. (c) Ambient radiation images vary strongly in brightness across the seasons, while also displaying strong noise. The noise looks akin to white noise or salt-and-pepper noise and authors do not know its cause. Full article ">Figure A3
Interventions of a LiDAR v1 model in the winter. The interventions are far more frequent in open fields, whereas the model can handle driving in the forest much better. Furthermore, the middle section of the route which contains bushes by the roadside is driven well. Full article ">

24 pages, 783 KiB

Open AccessEditor’s ChoiceReview

Unsupervised Anomaly Detection for IoT-Based Multivariate Time Series: Existing Solutions, Performance Analysis and Future Directions

by Mohammed Ayalew Belay, Sindre Stenen Blakseth, Adil Rasheed and Pierluigi Salvo Rossi

Sensors 2023, 23(5), 2844; https://doi.org/10.3390/s23052844 - 6 Mar 2023

Cited by 31 | Viewed by 16801

Abstract

The recent wave of digitalization is characterized by the widespread deployment of sensors in many different environments, e.g., multi-sensor systems represent a critical enabling technology towards full autonomy in industrial scenarios. Sensors usually produce vast amounts of unlabeled data in the form of [...] Read more.

The recent wave of digitalization is characterized by the widespread deployment of sensors in many different environments, e.g., multi-sensor systems represent a critical enabling technology towards full autonomy in industrial scenarios. Sensors usually produce vast amounts of unlabeled data in the form of multivariate time series that may capture normal conditions or anomalies. Multivariate Time Series Anomaly Detection (MTSAD), i.e., the ability to identify normal or irregular operative conditions of a system through the analysis of data from multiple sensors, is crucial in many fields. However, MTSAD is challenging due to the need for simultaneous analysis of temporal (intra-sensor) patterns and spatial (inter-sensor) dependencies. Unfortunately, labeling massive amounts of data is practically impossible in many real-world situations of interest (e.g., the reference ground truth may not be available or the amount of data may exceed labeling capabilities); therefore, robust unsupervised MTSAD is desirable. Recently, advanced techniques in machine learning and signal processing, including deep learning methods, have been developed for unsupervised MTSAD. In this article, we provide an extensive review of the current state of the art with a theoretical background about multivariate time-series anomaly detection. A detailed numerical evaluation of 13 promising algorithms on two publicly available multivariate time-series datasets is presented, with advantages and shortcomings highlighted. Full article

(This article belongs to the Special Issue Signal Processing and AI in Sensor Networks and IoT)

► Show Figures

Figure 1

Journal Menu

Journal Browser

Sensors, Volume 23, Issue 5 (March-1 2023) – 509 articles

Further Information

Guidelines

MDPI Initiatives

Follow MDPI