Search

Recently, AI systems such as autonomous driving and smart homes have become integral to daily life. Intelligent multi-sensors, once limited to single data types, now process complex text and image data, demanding faster and more accurate processing. While integrating NPUs and sensors has improved processing speed and accuracy, challenges like low resource utilization and long memory latency remain. This study proposes a method to reduce processing time and improve resource utilization by virtualizing NPUs to simultaneously handle multiple deep-learning models, leveraging a hardware scheduler and data prefetching techniques. Experiments with 30,000 SA resources showed that the hardware scheduler reduced memory cycles by over 10% across all models, with reductions of 30% for NCF and 70% for DLRM. The hardware scheduler effectively minimized memory latency and idle NPU resources in resource-constrained environments with frequent context switching. This approach is particularly valuable for real-time applications like autonomous driving, enabling smooth transitions between tasks such as object detection and route planning. It also enhances multitasking in smart homes by reducing latency when managing diverse data streams. The proposed system is well suited for resource-constrained environments that demand efficient multitasking and low-latency processing.

Sleep apnea (SA) is a prevalent disorder characterized by recurrent events of nocturnal apnea. Polysomnography (PSG) represents the gold standard for SA diagnosis. This laboratory-based procedure is complex and costly, and less cumbersome wearable devices have been proposed for SA detection and monitoring. A novel textile multi-sensor monitoring belt recording electrocardiogram (ECG) and breathing frequency (BF) measured by thorax excursion was developed and tested in a sleep laboratory for validation purposes. The aim of the current study was to evaluate the diagnostic performance of ECG-derived heart rate variability and BF-derived breathing rate variability and their combination for the detection of sleep apnea in a population of patients with a suspicion of SA. Fifty-one patients with a suspicion of SA were recruited in the sleep laboratory of the Cantonal Hospital St. Gallen. Patients were equipped with the monitoring belt and underwent a single overnight laboratory-based PSG. In addition, some patients further tested the monitoring belt at home. The ECG and BF signals from the belt were compared to PSG signals using the Bland-Altman methodology. Heart rate and breathing rate variability analyses were performed. Features derived from these analyses were used to build a support vector machine (SVM) classifier for the prediction of SA severity. Model performance was assessed using receiver operating characteristics (ROC) curves. Patients included 35 males and 16 females with a median age of 49 years (range: 21 to 65) and a median apnea-hypopnea index (AHI) of 33 (IQR: 16 to 58). Belt-derived data provided ECG and BF signals with a low bias and in good agreement with PSG-derived signals. The combined ECG and BF signals improved the classification accuracy for SA (area under the ROC curve: 0.98; sensitivity and specificity greater than 90%) compared to single parameter classification based on either ECG or BF alone. This novel wearable device combining ECG and BF provided accurate signals in good agreement with the gold standard PSG. Due to its unobtrusive nature, it is potentially interesting for multi-night assessments and home-based patient follow-up.

Large components are crucial in modern industrial applications, especially for internal gap monitoring and specific assembly methods. This paper examines the assembly of hydroelectric generator rotors and stators, introducing a spatial relative position monitoring system using multiple sensors. A dedicated position monitoring program is designed, and error sources within the system are thoroughly explored. Detailed error analysis and modeling reveal that verticality and angular errors significantly impact monitoring accuracy. To address this, two error control methods are proposed to effectively mitigate these issues, ensuring precise assembly of large components.

In recent years, offshore wind farms have frequently encountered engineering geological disasters such as seabed liquefaction and scouring. Consequently, in situ monitoring has become essential for the safe siting, construction, and operation of these installations. Current technologies are hampered by limitations in single-parameter monitoring and insufficient probe-penetration depth, hindering comprehensive multi-parameter dynamic monitoring of seabed sediments. To address these challenges, we propose a foldable multi-sensor probe and establish an underwater adaptive continuous penetration system capable of concurrently measuring seabed elevation changes and sediment pore water pressure profiles. The reliability of the equipment design is confirmed through static analysis of the frame structure and sealed cabin. Furthermore, laboratory tests validate the stability and accuracy of the electrical and mechanical sensor measurements. Preliminary tests conducted in a harbor environment demonstrate the system's effectiveness.

With the continuous development of new sensor features and tracking algorithms for object tracking, researchers have opportunities to experiment using different combinations. However, there is no standard or agreed method for selecting an appropriate architecture for autonomous vehicle (AV) crash reconstruction using multi-sensor-based sensor fusion. This study proposes a novel simulation method for tracking performance evaluation (SMTPE) to solve this problem. The SMTPE helps select the best tracking architecture for AV crash reconstruction. This study reveals that a radar-camera-based centralized tracking architecture of multi-sensor fusion performed the best among three different architectures tested with varying sensor setups, sampling rates, and vehicle crash scenarios. We provide a brief guideline for the best practices in selecting appropriate sensor fusion and tracking architecture arrangements, which can be helpful for future vehicle crash reconstruction and other AV improvement research.

For mobile robots, the high-precision integrated calibration and structural robustness of multi-sensor systems are important prerequisites for ensuring healthy operations in the later stage. Currently, there is no well-established validation method for the calibration accuracy and structural robustness of multi-sensor systems, especially for dynamic traveling situations. This paper presents a novel validation method for the calibration accuracy and structural robustness of a multi-sensor mobile robot. The method employs a ground-object-air cooperation mechanism, termed the "ground surface simulation field (GSSF)-mobile robot -photoelectric transmitter station (PTS)". Firstly, a static high-precision GSSF is established with the true north datum as a unified reference. Secondly, a rotatable synchronous tracking system (PTS) is assembled to conduct real-time pose measurements for a mobile vehicle. The relationship between each sensor and the vehicle body is utilized to measure the dynamic pose of each sensor. Finally, the calibration accuracy and structural robustness of the sensors are dynamically evaluated. In this context, epipolar line alignment is employed to assess the accuracy of the evaluation of relative orientation calibration of binocular cameras. Point cloud projection and superposition are utilized to realize the evaluation of absolute calibration accuracy and structural robustness of individual sensors, including the navigation camera (Navcam), hazard avoidance camera (Hazcam), multispectral camera, time-of-flight depth camera (TOF), and light detection and ranging (LiDAR), with respect to the vehicle body. The experimental results demonstrate that the proposed method offers a reliable means of dynamic validation for the testing phase of a mobile robot.

The dynamic of marine environmental parameters affects the distribution of commercial tuna in the sea of the maritime continent. Hence, the objectives of this study are to develop spatial distribution models for the four main tuna species in the Maritime continent's sea with reasonable accuracy, identify their correlation with marine environmental parameters, and investigate areas of interaction between those tuna species. The study develops the distribution models for albacore (Thunnus alalunga), bigeye (Thunnus obesus), yellowfin (Thunnus albacares), and skipjack (Katsuwonus pelamis) tuna species, utilizing multi-sensor satellite remote sensing and maximum entropy. The results show models have good performance, focusing on environmental factors such as sea surface temperature (SST), chlorophyll-a (CHL), and sea surface height anomalies (SSHA), combined with eddy kinetic energy (EKE). Seasonal variations in potential tuna habitats are revealed, emphasizing the influence of those marine environmental conditions. From December to May, the four commercial tuna species were distributed in conditions characterized by SST of 26-31.5 °C, CHL levels of 0-3 mg/l, SSHA of -0.3 to 0.2 m, and EKE of 0-1 m 2 /s 2 , while from June to November, they experienced SST of 23-31 °C, CHL levels of 0-4 mg/l, SSHA of -0.5 to 0.3 m, and EKE of 0-1.1 m 2 /s 2 . The spatial persistence of the four tuna species emerged mainly around the south sea of Java, with skipjack being the most common species found in the sea of the maritime continent. With sufficient and evenly distributed tuna presence records, the results indicate the potential for extrapolation beyond the training data to estimate habitat suitability for the four commercial tuna distributions. The results also suggest potential competition between tuna species sharing ecological niches and highlight possible overlapping areas where different tuna species interact with the same fishing gear., Competing Interests: Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper., (Copyright © 2024 Elsevier Ltd. All rights reserved.)

The Indoor Environmental Quality (IEQ) combines thermal, visual, acoustic, and air-quality conditions in indoor environments and affects occupants' health, well-being, and comfort. Performing continuous monitoring to assess IEQ is increasingly proving to be important, also due to the large amount of time that people spend in closed spaces. In the present study, the design, development, and metrological characterization of a low-cost multi-sensor device is presented. The device is part of a wider system, hereafter referred to as PROMET&O (PROactive Monitoring for indoor EnvironmenTal quality & cOmfort), that also includes a questionnaire for the collection of occupants' feedback on comfort perception and a dashboard to show end users all monitored data. The PROMET&O multi-sensor monitors the quality conditions of indoor environments thanks to a set of low-cost sensors that measure air temperature, relative humidity, illuminance, sound pressure level, carbon monoxide, carbon dioxide, nitrogen dioxide, particulate matter, volatile organic compounds, and formaldehyde. The device architecture is described, and the design criteria related to measurement requirements are highlighted. Particular attention is paid to the calibration of the device to ensure the metrological traceability of the measurements. Calibration procedures, based on the comparison to reference standards and following commonly employed or ad hoc developed technical procedures, were defined and applied to the bare sensors of air temperature and relative humidity, carbon dioxide, illuminance, sound pressure level, particulate matter, and formaldehyde. The next calibration phase in the laboratory will be aimed at analyzing the mutual influences of the assembled multi-sensor hardware components and refining the calibration functions.

Aiming at the traditional single sensor vibration signal cannot fully express the bearing running state, and in the high noise background, the traditional algorithm is insufficient for fault feature extraction. This paper proposes a fault diagnosis algorithm based on multi-sensor and hybrid multimodal feature fusion to achieve high-precision fault diagnosis by leveraging the operating state information of bearings in a high-noise environment to the fullest extent possible. First, the horizontal and vertical vibration signals from two sensors are fused using principal component analysis, aiming to provide a more comprehensive description of the bearing's operating condition, followed by data set segmentation. Following fusion, time-frequency feature maps are generated using a continuous wavelet transform for global time-frequency feature extraction. A first diagnostic model is then developed utilizing a residual neural network. Meanwhile, the feature data is normalized, and 28 time-frequency feature indexes are extracted. Subsequently, a second diagnostic model is constructed using a support vector machine. Lastly, the two diagnosis models are integrated to derive the final model through an ensemble learning algorithm fused at the decision level and complemented by a genetic algorithm solution to improve the diagnosis accuracy. Experimental results demonstrate the effectiveness of the proposed algorithm in achieving superior diagnostic performance with a 97.54% accuracy rate.

The machines of WF Maschinenbau process metal blanks into various workpieces using so-called flow-forming processes. The quality of these workpieces depends largely on the quality of the blanks and the condition of the machine. This creates an urgent need for automated monitoring of the forming processes and the condition of the machine. Since the complexity of the flow-forming processes makes physical modeling impossible, the present work deals with data-driven modeling using machine learning algorithms. The main contributions of this work lie in showcasing the feasibility of utilizing machine learning and sensor data to monitor flow-forming processes, along with developing a practical approach for this purpose. The approach includes an experimental design capable of providing the necessary data, as well as a procedure for preprocessing the data and extracting features that capture the information needed by the machine learning models to detect defects in the blank and the machine. To make efficient use of the small number of experiments available, the experimental design is generated using Design of Experiments methods. They consist of two parts. In the first part, a pre-selection of influencing variables relevant to the forming process is performed. In the second part of the design, the selected variables are investigated in more detail. The preprocessing procedure consists of feature engineering, feature extraction and feature selection. In the feature engineering step, the data set is augmented with time series variables that are meaningful in the domain. For feature extraction, an algorithm was developed based on the mechanisms of the r-STSF, a state-of-the-art algorithm for time series classification, extending them for multivariate time series and metric target variables. This feature extraction algorithm itself can be seen as an additional contribution of this work, because it is not tied to the application domain of monitoring flow-forming processes, but can be used as a feature extraction algorithm for multivariate time series classification in general. For feature selection, a Recursive Feature Elimination is employed. With the resulting features, random forests are trained to detect several quality features of the blank and defects of the machine. The trained models achieve good prediction accuracy for most of the target variables. This shows that the application of machine learning is a promising approach for the monitoring of flow-forming processes, which requires further investigation for confirmation.

To meet the complex and diverse demands for low-stress mechanical measurements of fabrics and other flexible materials, two integrated multidimensional force sensors with the same structure but different ranges were explored. They can support both rapid and precise low-noise, high-precision, low-cost, easy-to-use, reliable, and intelligent solutions for the complex measurement of fabric mechanics. Having analysed the mechanical relationship of the parallel beam theory, and considering the specific requirements of fabric measurement, a novel multi-dimensional force sensor is designed, capable of measuring tensile, shear, and buckling properties. Finite element analysis is used to simulate the mechanical performance of this sensor for fabric-loading/unloading measurement, and the sensitivity of the mechanical quantity transfer, the amount of sensor deformation, the stress distribution, and the degree of inter-dimensional coupling have been investigated and verified. The basis for subsequent digital processing is achieved by a low-offset, low-temperature-drift, low-power-consumption analogue front end, 24-bit ADC circuit, and signal conditioning electronics, suitable for the measurement of fabric mechanics under low stress, which is like the end-user requirements. The sensor information channel is supported by a host microcontroller with a DSP and a floating-point processing instruction set. Information processing is performed in time-sharing with the support of a multitasking real-time operating system. The purpose of designing this sensor is to facilitate the development of a new testing instrument, which will adopt the advances of current instruments whilst eliminating their shortcomings.

Global aging leads to a surge in neurological diseases. Quantitative gait analysis for the early detection of neurological diseases can effectively reduce the impact of the diseases. Recently, extensive research has focused on gait-abnormality-recognition algorithms using a single type of portable sensor. However, these studies are limited by the sensor's type and the task specificity, constraining the widespread application of quantitative gait recognition. In this study, we propose a multimodal gait-abnormality-recognition framework based on a Convolutional Neural Network-Bidirectional Long Short-Term Memory (CNN-BiLSTM) network. The as-established framework effectively addresses the challenges arising from smooth data interference and lengthy time series by employing an adaptive sliding window technique. Then, we convert the time series into time-frequency plots to capture the characteristic variations in different abnormality gaits and achieve a unified representation of the multiple data types. This makes our signal processing method adaptable to several types of sensors. Additionally, we use a pre-trained Deep Convolutional Neural Network (DCNN) for feature extraction, and the consequently established CNN-BiLSTM network can achieve high-accuracy recognition by fusing and classifying the multi-sensor input data. To validate the proposed method, we conducted diversified experiments to recognize the gait abnormalities caused by different neuropathic diseases, such as amyotrophic lateral sclerosis (ALS), Parkinson's disease (PD), and Huntington's disease (HD). In the PDgait dataset, the framework achieved an accuracy of 98.89% in the classification of Parkinson's disease severity, surpassing DCLSTM's 96.71%. Moreover, the recognition accuracy of ALS, PD, and HD on the PDgait dataset was 100%, 96.97%, and 95.43% respectively, surpassing the majority of previously reported methods. These experimental results strongly demonstrate the potential of the proposed multimodal framework for gait abnormality identification. Due to the advantages of the framework, such as its suitability for different types of sensors and fewer training parameters, it is more suitable for gait monitoring in daily life and the customization of medical rehabilitation schedules, which will help more patients alleviate the harm caused by their diseases.

Vehicle detection using data fusion techniques from overhead platforms (RGB/MSI imagery and LiDAR point clouds) with vector and shape data can be a powerful tool in a variety of fields, including, but not limited to, national security, disaster relief efforts, and traffic monitoring. Knowing the location and number of vehicles in a given area can provide insight into the surrounding activities and patterns of life, as well as support decision-making processes. While researchers have developed many approaches to tackling this problem, few have exploited the multi-data approach with a classical technique. In this paper, a primarily LiDAR-based method supported by RGB/MSI imagery and road network shapefiles has been developed to detect stationary vehicles. The addition of imagery and road networks, when available, offers an improved classification of points from LiDAR data and helps to reduce false positives. Furthermore, detected vehicles can be assigned various 3D, relational, and spectral attributes, as well as height profiles. This method was evaluated on the Houston, TX dataset provided by the IEEE 2018 GRSS Data Fusion Contest, which includes 1476 ground truth vehicles from LiDAR data. On this dataset, the algorithm achieved a 92% precision and 92% recall. It was also evaluated on the Vaihingen, Germany dataset provided by ISPRS, as well as data simulated using an image generation model called DIRSIG. Some known limitations of the algorithm include false positives caused by low vegetation and the inability to detect vehicles (1) in extremely close proximity with high precision and (2) from low-density point clouds.

Under the influence of multiple types of noises, missing measurement, one-step measurement delay and packet loss, the robust Kalman estimation problem is studied for the multi-sensor descriptor system (MSDS) in this paper. Moreover, the established MSDS model describes uncertain-variance noises, multiplicative noises, time delay and packet loss phenomena. Different types of noises and packet loss make it more difficult to build the estimators of MSDS. Firstly, MSDS is transformed to the new system model by applying the singular value decomposition (SVD) method, augmented state and fictitious noise approach. Furthermore, the robust Kalman estimator is constructed for the newly deduced augmented system based on the min-max robust estimation principle and Kalman filter theory. In addition, the given estimator consists of four parts, which are the usual Kalman filter, predictor, smoother and white noise deconvolution estimator. Then, the robust fusion Kalman estimator is obtained for MSDS according to the relation of augmented state and the original system state. Simultaneously, the robustness is demonstrated for the actual Kalman estimator of MSDS by using the mathematical induction method and Lyapunov's equation. Furthermore, the error variance of the obtained Kalman estimator is guaranteed to the upper bound for all admissible uncertain noise variance. Finally, the simulation example of a circuit system is examined to illustrate the performance and effectiveness of the robust estimators.

Tungsten heavy alloys (WHAs) are an extremely hard-to-machine material extensively used in demanding applications such as missile liners, aerospace, and optical molds. However, the machining of WHAs remains a challenging task as a result of their high density and elastic stiffness which lead to the deterioration of the machined surface roughness. This paper proposes a brand-new multi-objective dung beetle algorithm. It does not take the cutting parameters (i.e., cutting speed, feed rate, and depth of cut) as the optimization objects but directly optimizes cutting forces and vibration signals monitored using a multi-sensor (i.e., dynamometer and accelerometer). The cutting parameters in the WHA turning process are analyzed through the use of the response surface method (RSM) and the improved dung beetle optimization algorithm. Experimental verification shows that the algorithm has better convergence speed and optimization ability compared with similar algorithms. The optimized forces and vibration are reduced by 9.7% and 46.47%, respectively, and the surface roughness R a of the machined surface is reduced by 18.2%. The proposed modeling and optimization algorithms are anticipated to be powerful to provide the basis for the parameter optimization in the cutting of WHAs.

With the development of industrial automation, articulated robots have gradually replaced labor in the field of bolt installation. Although the installation efficiency has been improved, installation defects may still occur. Bolt installation defects can considerably affect the mechanical properties of structures and even lead to safety accidents. Therefore, in order to ensure the success rate of bolt assembly, an efficient and timely detection method of incorrect or missing assembly is needed. At present, the automatic detection of bolt installation defects mainly depends on a single type of sensor, which is prone to mis-inspection. Visual sensors can identify the incorrect or missing installation of bolts, but it cannot detect torque defects. Torque sensors can only be judged according to the torque and angel information, but cannot accurately identify the incorrect or missing installation of bolts. To solve this problem, a detection method of bolt installation defects based on multiple sensors is proposed. The trained YOLO (You Only Look Once) v3 network is used to judge the images collected by the visual sensor, and the recognition rate of visual detection is up to 99.75%, and the average confidence of the output is 0.947. The detection speed is 48 FPS, which meets the real-time requirement. At the same time, torque and angle sensors are used to judge the torque defects and whether bolts have slipped. Combined with the multi-sensor judgment results, this method can effectively identify defects such as missing bolts and sliding teeth. Finally, this paper carried out experiments to identify bolt installation defects such as incorrect, missing torque defects, and bolt slips. At this time, the traditional detection method based on a single type of sensor cannot be effectively identified, and the detection method based on multiple sensors can be accurately identified.

Robotic systems are evolving to include a large number of sensors and diverse sensor modalities. In order to operate a system with multiple sensors, the geometric transformations between those sensors must be accurately estimated. The process by which these transformations are estimated is known as sensor calibration. Behind every sensor calibration approach is a formulation and a framework. The formulation is the method by which the transformations are estimated. The framework is the set of operations required to carry out the calibration procedure. This paper proposes a novel calibration framework that gives more flexibility, control and information to the user, enhancing the user interface and the user experience of calibrating a robotic system. The framework consists of several visualization and interaction functionalities useful for a calibration procedure, such as the estimation of the initial pose of the sensors, the data collection and labeling, the data review and correction and the visualization of the estimation of the extrinsic and intrinsic parameters. This framework is supported by the Atomic Transformations Optimization Method formulation, referred to as ATOM. Results show that this framework is applicable to various robotic systems with different configurations, number of sensors and sensor modalities. In addition to this, a survey comparing the frameworks of different calibration approaches shows that ATOM provides a very good user experience.

This article investigates a security problem in cyber-physical systems under multisensory framework. Data packets from each sensor are transmitted to a remote estimator through wireless channels which could be attacked by a attacker and injected false data. Based on the data of reliable sensors and the relativity between reliable and unreliable sensors, a revised multi-sensor Kalman filter fusion algorithm is proposed. Under the proposed algorithm, an optimal linear false data injection attack strategy which is more general with an arbitrary mean of Gaussian distribution is designed. To further improve the detection performance, an Expected SARSA-based attack detection algorithm is proposed. Finally, simulation results based on an unmanned aerial vehicle are provided to illustrate the feasibility and efficiency of the obtained results., Competing Interests: Declaration of Competing Interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper., (Copyright © 2022 ISA. Published by Elsevier Ltd. All rights reserved.)

The damage caused by natural disasters in rural areas differs in nature extent, landscape, and structure, from the damage caused in urban environments. Previous and current studies have focused mainly on mapping damaged structures in urban areas after catastrophic events such as earthquakes or tsunamis. However, research focusing on the level of damage or its distribution in rural areas is lacking. This study presents a methodology for mapping, characterizing, and assessing the damage in rural environments following natural disasters, both in built-up and vegetation areas, by combining synthetic-aperture radar (SAR) and optical remote sensing data. As a case study, we applied the methodology to characterize the rural areas affected by the Sulawesi earthquake and the subsequent tsunami event in Indonesia that occurred on 28 September 2018. High-resolution COSMO-SkyMed images obtained pre- and post-event, alongside Sentinel-2 images, were used as inputs. This study's results emphasize that remote sensing data from rural areas must be treated differently from that of urban areas following a disaster. Additionally, the analysis must include the surrounding features, not only the damaged structures. Furthermore, the results highlight the applicability of the methodology for a variety of disaster events, as well as multiple hazards, and can be adapted using a combination of different optical and SAR sensors.

Intelligent mechanical systems are a focused area nowadays. One of the requirements of intelligent mechanical systems is to achieve intelligent fault diagnosis through the real-time acquisition and analysis of data from various sensors installed on mechanical components. In this paper, a new fault diagnosis method is proposed to solve the problems of difficulty in integrating the fault diagnosis algorithm and locating fault parts due to the complexity of modern mechanical systems. The complexity of modern industrial intelligent systems is due to the fact that the systems are composed of multiple components and there are various connections between them. Common fault diagnosis is to design specialized fault identification algorithms for the physical characteristics of each component, and the integration of different algorithms is a major challenge for system performance. Therefore, this paper investigates a general algorithm for the fault diagnosis of complex systems using the timing characteristics of sensors and transfer entropy. The fault diagnosis algorithm is based on the prediction of multi-dimensional long time series using Autoformer, and fault identification is performed based on the deviation of the predicted value from the actual value. After fault identification, a root cause analysis method of faults based on transfer entropy is proposed. The method can locate the component where the fault occurs more accurately based on the analysis of the cause-effect relationship of each component and help maintenance personnel to troubleshoot the fault.

Personalised healthcare has seen significant improvements due to the introduction of health monitoring technologies that allow wearable devices to unintrusively monitor physiological parameters such as heart health, blood pressure, sleep patterns, and blood glucose levels, among others. Additionally, utilising advanced sensing technologies based on flexible and innovative biocompatible materials in wearable devices allows high accuracy and precision measurement of biological signals. Furthermore, applying real-time Machine Learning algorithms to highly accurate physiological parameters allows precise identification of unusual patterns in the data to provide health event predictions and warnings for timely intervention. However, in the predominantly adopted architectures, health event predictions based on Machine Learning are typically obtained by leveraging Cloud infrastructures characterised by shortcomings such as delayed response times and privacy issues. Fortunately, recent works highlight that a new paradigm based on Edge Computing technologies and on-device Artificial Intelligence significantly improve the latency and privacy issues. Applying this new paradigm to personalised healthcare architectures can significantly improve their efficiency and efficacy. Therefore, this paper reviews existing IoT healthcare architectures that utilise wearable devices and subsequently presents a scalable and modular system architecture to leverage emerging technologies to solve identified shortcomings. The defined architecture includes ultrathin, skin-compatible, flexible, high precision piezoelectric sensors, low-cost communication technologies, on-device intelligence, Edge Intelligence, and Edge Computing technologies. To provide development guidelines and define a consistent reference architecture for improved scalable wearable IoT-based critical healthcare architectures, this manuscript outlines the essential functional and non-functional requirements based on deductions from existing architectures and emerging technology trends. The presented system architecture can be applied to many scenarios, including ambient assisted living, where continuous surveillance and issuance of timely warnings can afford independence to the elderly and chronically ill. We conclude that the distribution and modularity of architecture layers, local AI-based elaboration, and data packaging consistency are the more essential functional requirements for critical healthcare application use cases. We also identify fast response time, utility, comfort, and low cost as the essential non-functional requirements for the defined system architecture.

Previous works indicate that data fusion, compared to single data modelling can improve the assessment of soil attributes using spectroscopy. In this work, two different kinds of proximal soil sensing techniques i.e., mid-infrared (MIR) and X-ray fluorescence (XRF) spectroscopy were evaluated, for assessment of seven fertility attributes. These soil attributes include pH, organic carbon (OC), phosphorous (P), potassium (K), magnesium (Mg), calcium (Ca) and moisture contents (MC). Three kinds of spectra fusion (SF) (spectra concatenation) approaches of MIR and XRF spectra were compared, namely, spectra fusion-Partial least square (SF-PLS), spectra fusion-Sequential Orthogonalized Partial least square (SF-SOPLS) and spectra fusion-Variable Importance Projection-Sequential Orthogonalized Partial least square (SF-VIP-SOPLS). Furthermore, the performance of SF models was compared with the developed single sensor model (based on individual spectra of MIR and XRF). Compared with the results obtained from single sensor model, SF models showed improvement in the prediction performance for all studied attributes, except for OC, Mg, and K prediction. More specifically, the highest improvement was observed with SF-SOPLS model for pH [ R 2 p = 0.90, root mean square error prediction (RMSEP) = 0.15, residual prediction deviation (RPD) = 3.30, and ratio of performance inter-quantile (RPIQ) = 3.59], successively followed by P ( R = 0.80, RMSEP = 1.91%, RPD = 2.31, RPIQ = 2.62). Overall the study concluded that SF approach with SOPLS attained better performance over the traditional model developed with the single sensor spectra, hence, SF is recommended as the best SF method for improving the prediction accuracy of studied soil attributes. Moreover, the multi-sensor spectra fusion approach is not limited for only MIR and XRF data but in general can be extended for complementary information fusion in order to improve the model performance in precision agriculture (PA) applications.2 p = 0.91, RMSEP = 4.45 mg/100 g, RPD = 3.53, and RPIQ = 4.90), Ca ( R 2 p = 0.92, RMSEP = 177.11 mg/100 g, RPD = 3.66, and RPIQ = 3.22) and MC ( R 2 p = 0.80, RMSEP = 1.91%, RPD = 2.31, RPIQ = 2.62). Overall the study concluded that SF approach with SOPLS attained better performance over the traditional model developed with the single sensor spectra, hence, SF is recommended as the best SF method for improving the prediction accuracy of studied soil attributes. Moreover, the multi-sensor spectra fusion approach is not limited for only MIR and XRF data but in general can be extended for complementary information fusion in order to improve the model performance in precision agriculture (PA) applications.

Fused deposition modelling (FDM)-based 3D printing is a trending technology in the era of Industry 4.0 that manufactures products in layer-by-layer form. It shows remarkable benefits such as rapid prototyping, cost-effectiveness, flexibility, and a sustainable manufacturing approach. Along with such advantages, a few defects occur in FDM products during the printing stage. Diagnosing defects occurring during 3D printing is a challenging task. Proper data acquisition and monitoring systems need to be developed for effective fault diagnosis. In this paper, the authors proposed a low-cost multi-sensor data acquisition system (DAQ) for detecting various faults in 3D printed products. The data acquisition system was developed using an Arduino micro-controller that collects real-time multi-sensor signals using vibration, current, and sound sensors. The different types of fault conditions are referred to introduce various defects in 3D products to analyze the effect of the fault conditions on the captured sensor data. Time and frequency domain analyses were performed on captured data to create feature vectors by selecting the chi-square method, and the most significant features were selected to train the CNN model. The K-means cluster algorithm was used for data clustering purposes, and the bell curve or normal distribution curve was used to define individual sensor threshold values under normal conditions. The CNN model was used to classify the normal and fault condition data, which gave an accuracy of around 94%, by evaluating the model performance based on recall, precision, and F1 score.

Use of a multi-sensor approach can provide citizens with holistic insights into the air quality of their immediate surroundings and their personal exposure to urban stressors. Our work, as part of the ICARUS H2020 project, which included over 600 participants from seven European cities, discusses the data fusion and harmonization of a diverse set of multi-sensor data streams to provide a comprehensive and understandable report for participants. Harmonizing the data streams identified issues with the sensor devices and protocols, such as non-uniform timestamps, data gaps, difficult data retrieval from commercial devices, and coarse activity data logging. Our process of data fusion and harmonization allowed us to automate visualizations and reports, and consequently provide each participant with a detailed individualized report. Results showed that a key solution was to streamline the code and speed up the process, which necessitated certain compromises in visualizing the data. A thought-out process of data fusion and harmonization of a diverse set of multi-sensor data streams considerably improved the quality and quantity of distilled data that a research participant received. Though automation considerably accelerated the production of the reports, manual and structured double checks are strongly recommended.

Aquatic invasive weeds affect hydrological, ecological, and socio-economic activities on freshwater ecosystems. On the Lower Volta River (LVR) of Ghana, invasive aquatic weeds have been known to be nuisance to fishing, navigation, aquaculture, hydropower production and other agricultural practices in the area. While information on the spatial and temporal distribution of aquatic weeds would be beneficial in improving weed management and control measures on the river, such information is very scanty. Also, these aquatic weeds are also biomass resources, that can be transformed to bioenergy. Thus, this study evaluated the spatial and temporal variations of aquatic weeds on the Lower Volta River, and assessed their potential biomass for bioenergy production. Random Forest (RF) algorithm and Landsat images were used to map the distribution of the weeds in 1975, 2003, and 2020, respectively. Accuracy assessment results showed mean Overall Accuracy (OA) of 83.44% and mean User Accuracy (UA) of 79.24%. The results indicated that as of 1975, aquatic weeds covered only 1495 ha and appeared in some specific locations such as Kpong and Ada. However, by 2003, the weeds had spread to most parts of the river covering 5600 ha, which was an increase of approximately 4-fold within a period of 28 years. The area covered by the weeds, however declined by 1505 ha between 2003 and 2020. Thus, in 2020, water hyacinth covered about 36% of the aquatic weeds relative to 28% in 2003. The results showed that, the quantity of the water hyacinth biomass per unit area was 21.5 kg/m 2 . This result can also be used as the basis for resource assessment as well as determination of its viability for bioenergy production and strategies for its modern utilisation. The conversion of water hyacinth into bioenergy remains one of the best aquatic weed management strategies that must be adopted in LVR., Competing Interests: The authors declare no conflict of interest., (© 2021 The Authors. Published by Elsevier Ltd.)

Many people live under stressful conditions which has an adverse effect on their health. Human stress, especially long-term one, can lead to a serious illness. Therefore, monitoring of human stress influence can be very useful. We can monitor stress in strictly controlled laboratory conditions, but it is time-consuming and does not capture reactions, on everyday stressors or in natural environment using wearable sensors, but with limited accuracy. Therefore, we began to analyze the current state of promising wearable stress-meters and the latest advances in the record of related physiological variables. Based on these results, we present the concept of an accurate, reliable and easier to use telemedicine device for long-term monitoring of people in a real life. In our concept, we ratify with two synchronized devices, one on the finger and the second on the chest. The results will be obtained from several physiological variables including electrodermal activity, heart rate and respiration, body temperature, blood pressure and others. All these variables will be measured using a coherent multi-sensors device. Our goal is to show possibilities and trends towards the production of new telemedicine equipment and thus, opening the door to a widespread application of human stress-meters.

Understanding mechanisms of materials deterioration during service life is fundamental for their confident use in the building sector. This work presents analysis of time series of data related to wood weathering acquired at three scales (molecular, microscopic, macroscopic) with different sensors. By using several complementary techniques, the material description is precise and complete; however, the data provided by multiple equipment are often not directly comparable due to different resolution, sensitivity and/or data format. This paper presents an alternative approach for multi-sensor data fusion and modelling of the deterioration processes by means of PARAFAC model. Time series data generated within this research were arranged in a data cube of dimensions samples × sensors × measuring time. The original protocol for data fusion as well as novel meta parameters, such as cumulative nested biplot, was proposed and tested. It was possible to successfully differentiate weathering trends of diverse materials on the basis of the NIR spectra and selected surface appearance indicators. A unique advantage for such visualization of the PARAFAC model output is the possibility of straightforward comparison of the degradation kinetics and deterioration trends simultaneously for all tested materials., (Copyright © 2020 The Author(s). Published by Elsevier B.V. All rights reserved.)

Human pose estimation and tracking in real-time from multi-sensor systems is essential for many applications. Combining multiple heterogeneous sensors increases opportunities to improve human motion tracking. Using only a single sensor type, e.g., inertial sensors, human pose estimation accuracy is affected by sensor drift over longer periods. This paper proposes a human motion tracking system using lidar and inertial sensors to estimate 3D human pose in real-time. Human motion tracking includes human detection and estimation of height, skeletal parameters, position, and orientation by fusing lidar and inertial sensor data. Finally, the estimated data are reconstructed on a virtual 3D avatar. The proposed human pose tracking system was developed using open-source platform APIs. Experimental results verified the proposed human position tracking accuracy in real-time and were in good agreement with current multi-sensor systems.

There are three state estimation fusion methods for a class of strong nonlinear measurement systems, based on the characteristic function filter, namely the centralized filter, parallel filter, and sequential filter. Under ideal communication conditions, the centralized filter can obtain the best state estimation accuracy, and the parallel filter can simplify centralized calculation complexity and improve feasibility; in addition, the performance of the sequential filter is very close to that of the centralized filter and far better than that of the parallel filter. However, the sequential filter can tolerate non-ideal conditions, such as delay and packet loss, and the first two filters cannot operate normally online for delay and will be invalid for packet loss. The performance of the three designed fusion filters is illustrated by three typical cases, which are all better than that of the most popular Extended Kalman Filter (EKF) performance.

Indoor occupancy prediction is a prerequisite for the management of energy consumption, security, health, and other systems in smart buildings. Previous studies have shown that buildings that automatize their heating, lighting, air conditioning, and ventilation systems through considering the occupancy and activity information might reduce energy consumption by more than 50%. However, it is difficult to use high-resolution sensors and cameras for occupancy prediction due to privacy concerns. In this paper, we propose a novel solution for predicting occupancy using multiple low-cost and low-resolution heat sensors. We suggest two different methods for fusing and processing the data captured from multiple heat sensors and we use a Convolutional Neural Network for predicting occupancy. We conduct experiments to assess both the performance of the proposed solutions and analyze the impact of sensor field view overlaps on the prediction results. In summary, our experimental results show that the implemented solutions show high occupancy prediction accuracy and real-time processing capabilities.

Accurate short-term small-area meteorological forecasts are essential to ensure the safety of operations and equipment operations in the Antarctic interior. This study proposes a deep learning-based multi-input neural network model to address this problem. The newly proposed model is predicted by combining a stacked autoencoder and a long- and short-term memory network. The self-stacking autoencoder maximises the features and removes redundancy from the target weather station's sensor data and extracts temporal features from the sensor data using a long- and short-term memory network. The proposed new model evaluates the prediction performance and generalisation capability at four observation sites at different East Antarctic latitudes (including the Antarctic maximum and the coastal region). The performance of five deep learning networks is compared through five evaluation metrics, and the optimal form of input combination is discussed. The results show that the prediction capability of the model outperforms the other models. It provides a new method for short-term meteorological prediction in a small inland Antarctic region., Competing Interests: The authors declare no conflicts of interest.

Urban areas and their vertical characteristics have a manifold and far-reaching impact on our environment. However, openly accessible information at high spatial resolution is still missing at large for complete countries or regions. In this study, we combined Sentinel-1A/B and Sentinel-2A/B time series to map building heights for entire Germany on a 10 m grid resolving built-up structures in rural and urban contexts. We utilized information from the spectral/polarization, temporal and spatial dimensions by combining band-wise temporal aggregation statistics with morphological metrics. We trained machine learning regression models with highly accurate building height information from several 3D building models. The novelty of this method lies in the very fine resolution yet large spatial extent to which it can be applied, as well as in the use of building shadows in optical imagery. Results indicate that both radar-only and optical-only models can be used to predict building height, but the synergistic combination of both data sources leads to superior results. When testing the model against independent datasets, very consistent performance was achieved (frequency-weighted RMSE of 2.9 m to 3.5 m), which suggests that the prediction of the most frequently occurring buildings was robust. The average building height varies considerably across Germany with lower buildings in Eastern and South-Eastern Germany and taller ones along the highly urbanized areas in Western Germany. We emphasize the straightforward applicability of this approach on the national scale. It mostly relies on freely available satellite imagery and open source software, which potentially permit frequent update cycles and cost-effective mapping that may be relevant for a plethora of different applications, e.g. physical analysis of structural features or mapping society's resource usage., Competing Interests: None., (© 2020 The Author(s).)

Internet of things (IoT) cloud-based applications deliver advanced solutions for smart cities to decrease traffic accidents caused by driver fatigue while driving on the road. Environmental conditions or driver behavior can ultimately lead to serious roadside accidents. In recent years, the authors have developed many low-cost, computerized, driver fatigue detection systems (DFDs) to help drivers, by using multi-sensors, and mobile and cloud-based computing architecture. To promote safe driving, these are the most current emerging platforms that were introduced in the past. In this paper, we reviewed state-of-the-art approaches for predicting unsafe driving styles using three common IoT-based architectures. The novelty of this article is to show major differences among multi-sensors, smartphone-based, and cloud-based architectures in multimodal feature processing. We discussed all of the problems that machine learning techniques faced in recent years, particularly the deep learning (DL) model, to predict driver hypovigilance, especially in terms of these three IoT-based architectures. Moreover, we performed state-of-the-art comparisons by using driving simulators to incorporate multimodal features of the driver. We also mention online data sources in this article to test and train network architecture in the field of DFDs on public available multimodal datasets. These comparisons assist other authors to continue future research in this domain. To evaluate the performance, we mention the major problems in these three architectures to help researchers use the best IoT-based architecture for detecting DFDs in a real-time environment. Moreover, the important factors of Multi-Access Edge Computing (MEC) and 5th generation (5G) networks are analyzed in the context of deep learning architecture to improve the response time of DFD systems. Lastly, it is concluded that there is a research gap when it comes to implementing the DFD systems on MEC and 5G technologies by using multimodal features and DL architecture.

Over the past few decades, partitioning of the surface ocean into ecologically-meaningful spatial domains has been approached using a range of data types, with the aim of improving our understanding of open ocean processes, supporting marine management decisions and constraining coupled ocean-biogeochemical models. The simplest partitioning method, which could provide low-latency information for managers at low cost, remains a purely optical classification based on ocean colour remote sensing. The question is whether such a simple approach has value. Here, the efficacy of optical classifications in constraining physical variables that modulate the epipelagic environment is tested for the tropical Indian Ocean, with a focus on the Chagos marine protected area (MPA). Using remote sensing data, it was found that optical classes corresponded to distinctive ranges of wind speed, wind stress curl, sea surface temperature, sea surface slope, sea surface height anomaly and geostrophic currents (Kruskal-Wallis and post-hoc Tukey honestly significantly different tests, α = 0.01). Between-class differences were significant for a set of sub-domains that resolved zonal and meridional gradients across the MPA and Seychelles-Chagos Thermocline Ridge, whereas between-domain differences were only significant for the north-south gradient (PERMANOVA, α = 0.01). A preliminary test of between-class differences in surface CO 2 concentrations from the Orbiting Carbon Observatory-2 demonstrated a small decrease in mean pCO 2 with increasing chlorophyll (chl), from 418 to 398 ppm. Simple optical class maps therefore provide an overview of growth conditions, the spatial distribution of resources - from which habitat fragmentation metrics can be calculated, and carbon sequestration potential. Within the 17 year study period, biotic variables were found to have decreased at up to 0.025%a -1 for all optical classes, which is slower than reported elsewhere (Mann-Kendall-Sen regression, α = 0.01). Within the MPA, positive Indian Ocean Dipole conditions and negative Southern Oscillation Indices were weakly associated with decreasing chl, fluorescence line height (FLH), eddy kinetic energy, easterly wind stress and wind stress curl, and with increasing FLH/chl, sea surface temperature, SSH gradients and northerly wind stress, consistent with reduced surface mixing and increased stratification. The optical partitioning scheme described here can be applied in Google Earth Engine to support management decisions at daily or monthly scales, and potential applications are discussed., (Crown Copyright © 2020. Published by Elsevier Ltd. All rights reserved.)

Subway structure safety detection is an important method to ensure the safe operation of trains. Efficient, high-precision, and automatic tunnel clearance detection is the key to ensure safe operations. This study introduces a mobile tunnel scanning system that integrates a scanner, an inertial measurement unit (IMU), and a rail car. Global Navigation Satellite System (GNSS) time and system hardware calibration are used to synchronize time and space information of the system; the attitude and speed are corrected using the control points from the tunnel to improve the accuracy of absolute positioning. The section coordinate system is converted using the control points and system calibration parameters to complete the tunnel clearance inspection, and the distance between the nearest point of the section and the clear height of the vault is given. Taking Fengxi Road's Bashan tunnel section of Chongqing Metro Line 5 as an example, the overall system accuracy was tested. The accuracy of chord line measurements was within 1 mm, the internal coincidence accuracy of repeated measurements of the vault clear height was 1.1 mm, the internal coincidence accuracy of repeated measurements of the closest gauge point was 4.8 mm, and the system calibration accuracy was approximately 2 mm. Compared with the existing scheme, the system combines absolute measurement and relative measurement mode to judge the structural safety of tunnel section from multiple angles, high precision, and high efficiency.

Inefficient remaining useful life (RUL) estimation may cause unpredictable failures and unscheduled maintenance of machining tools. Multi-sensor data fusion will improve the RUL prediction reliability by fusing more sensor information related to the machining process of tools. In this paper, a multi-sensor data fusion system for online RUL prediction of machining tools is proposed. The system integrates multi-sensor signal collection, signal preprocess by a complementary ensemble empirical mode decomposition, feature extraction in time domain, frequency domain and time-frequency domain by such methods as statistical analysis, power spectrum density analysis and Hilbert-Huang transform, feature selection by a Light Gradient Boosting Machine method, feature fusion by a tool wear prediction model based on back propagation neural network optimized by improved artificial bee colony (IABC-BPNN) algorithm, and the online RUL prediction model by a polynomial curve fitting method. An example is used to verify whether if the prediction performance of the proposed system is stable and reliable, and the results show that it is superior to its rivals.

In the late summer of 2016, our team deployed a network of low-cost air quality sensing systems in partnership with community-based organizations in a neighborhood in South Los Angeles, California. Residents of this community were concerned about possible emissions from local oil and gas activity, however in addition to these potential emissions, the neighborhood is also subject to a complex mixture of pollutants from other nearby sources including major highways. For this deployment, metal-oxide VOC sensors were quantified to provide methane (CH 4 ) and total non-methane hydrocarbon (TNMHCs) concentration estimates. This data along with other sensor signals, meteorological data, and community member observations was used to examine the composition and possible origins of observed emissions. The sensor network displayed expected environmental trends and highlighted short-term elevations in CH 4 and/or TNMHCs, which we were then able to investigate more closely. The results indicated that sources of both combusted and volatilized hydrocarbons were likely affecting air quality throughout the community, including near the site of the local oil and gas activity. This deployment may serve as a model for how multi-sensor systems deployed in networks can be leveraged to better understand sources in complex areas, potentially supporting future community-based air quality research.

Brain source imaging and time frequency mapping (TFM) are commonly used in magneto/electro encephalography (M/EEG) imaging. However, these methods suffer from important limitations. Source imaging is based on an ill-posed inverse problem leading to instability of source localization solutions, has a limited capacity to localize high frequency oscillations and loses its robustness for induced responses (ill-defined trigger). The drawback of TFM is that it involves independent analysis of signals from a number of frequency bands, and from co-localized sensors. In the present article, a regression-based multi-sensor space-time-frequency analysis (MSA) approach, which integrates co-localized sensors and/or multi-frequency information, is proposed. To estimate task-specific brain activations, MSA uses cross-validated, shifted, multiple Pearson correlation, calculated from the time-frequency transformed brain signal and the binary signal of stimuli. The results are projected from the sensor space onto the cortical surface. To assess MSA performance, the proposed method was compared to the weighted minimum norm estimate (wMNE) source imaging method, in terms of spatial selectivity and robustness against an ill-defined trigger. Magnetoencephalography (MEG) recordings were performed in fourteen subjects during two motor tasks: finger tapping and elbow flexion/extension. In particular, our results show that the MSA approach provides good localization performance when compared to wMNE and statistically significant improvement of robustness against ill-defined trigger.

Due to the increasing rates of chronic diseases and an aging population, the use of assistive devices for ambulation is expected to grow rapidly over the next several years. Instrumenting these devices has been proposed as a non-invasive way to proactively monitor changes in gait due to the presence of pain or a condition in outdoor and indoor environments. In this paper, we evaluated the effectiveness of a multi-sensor cane in detecting changes in gait due to the presence of simulated gait abnormalities, walking terrains, impaired vision, and incorrect cane lengths. The effectiveness of the instrumented cane was compared with the results obtained directly from a shank-mounted inertial measurement unit. Results from 30 healthy participants obtained while simulating gait abnormalities and walking over different terrains demonstrated the ability of the cane to reliably and effectively discriminate among these walking conditions. Moreover, the results obtained while walking with impaired vision and incorrect cane lengths indicate the ability of cane to detect changes in gait during these scenarios as well., Competing Interests: The authors declare no conflict of interest.

To solve the problem of passive sensor data association in multi-sensor multi-target tracking, a novel linear-time direct data assignment (DDA) algorithm is proposed in this paper. Different from existing methods which solve the data association problem in the measurement domain, the proposed algorithm solves the problem directly in the target state domain. The number and state of candidate targets are preset in the region of interest, which can avoid the problem of combinational explosion. The time complexity of the proposed algorithm is linear with the number of sensors and targets while that of the existing algorithms are exponential. Computer simulations show that the proposed algorithm can achieve almost the same association accuracy as the existing algorithms, but the time consumption can be significantly reduced.

Autonomous posture detection and self-localization of roadheaders is the key to automatic tunneling and roadheader robotization. In this paper, a multi-sensor based positioning method, involving an inertial system for altitude angles measurement, total station for coordinate measurement, and sensors for measuring the real-time length of the hydraulic cylinder is presented for roadheader position measurement and posture detection. Based on this method, a positioning model for roadheader and cutter positioning is developed. Additionally, flexible trajectory planning methods are provided for automatic cutting. Based on the positioning model and the trajectory planning methods, an automatic cutting procedure is proposed and applied in practical tunneling. The experimental results verify the high accuracy and efficiency of both the positioning method and the model. Furthermore, it is indicated that arbitrary shapes can be generated automatically and precisely according to the planned trajectory, employing the automatic cutting procedure. Therefore, unmanned tunneling can be realized by employing the proposed automatic cutting process.

Usage of Unmanned Aerial Vehicles (UAVs) is growing rapidly in a wide range of consumer applications, as they prove to be both autonomous and flexible in a variety of environments and tasks. However, this versatility and ease of use also brings a rapid evolution of threats by malicious actors that can use UAVs for criminal activities, converting them to passive or active threats. The need to protect critical infrastructures and important events from such threats has brought advances in counter UAV (c-UAV) applications. Nowadays, c-UAV applications offer systems that comprise a multi-sensory arsenal often including electro-optical, thermal, acoustic, radar and radio frequency sensors, whose information can be fused to increase the confidence of threat's identification. Nevertheless, real-time surveillance is a cumbersome process, but it is absolutely essential to detect promptly the occurrence of adverse events or conditions. To that end, many challenging tasks arise such as object detection, classification, multi-object tracking and multi-sensor information fusion. In recent years, researchers have utilized deep learning based methodologies to tackle these tasks for generic objects and made noteworthy progress, yet applying deep learning for UAV detection and classification is considered a novel concept. Therefore, the need to present a complete overview of deep learning technologies applied to c-UAV related tasks on multi-sensor data has emerged. The aim of this paper is to describe deep learning advances on c-UAV related tasks when applied to data originating from many different sensors as well as multi-sensor information fusion. This survey may help in making recommendations and improvements of c-UAV applications for the future.

To release the strong dependence of the conventional inertial navigation mechanization on the a priori low-cost inertial measurement unit (IMU) error model, this research applies an unconventional multi-sensor integration strategy to integrate multiple low-cost IMUs and a global positioning system (GPS) for mass-market automotive applications. The unconventional integration strategy utilizes a basic three-dimensional (3D) kinematic trajectory model as the system model to directly estimate navigational parameters, and it allows the measurements from all of the sensors independently participating in measurement updates. However, the less complex kinematic model cannot realize smooth transitions between different motion statuses for the road vehicle with acceleration maneuvers. In this manuscript, we establish a more practical 3D kinematic trajectory model based on a "current" statistical Singer acceleration model to realize smooth transitions for the maneuvering vehicle. In addition, taking advantage of the unconventional strategy, we individually model the systematic errors of each IMU and the measurements of all sensors, in contrast to most existing approaches that adopt the common-mode errors for different sensors of the same design. A real dataset involving a GPS and multiple IMUs is processed to validate the success of the proposed algorithm model under the unconventional integration strategy., Competing Interests: The authors declare no conflicts of interest.

The fast progress in research and development of multifunctional, distributed sensor networks has brought challenges in processing data from a large number of sensors. Using deep learning methods such as convolutional neural networks (CNN), it is possible to build smarter systems to forecasting future situations as well as precisely classify large amounts of data from sensors. Multi-sensor data from atmospheric pollutants measurements that involves five criteria, with the underlying analytic model unknown, need to be categorized, so do the Diabetic Retinopathy (DR) fundus images dataset. In this work, we created automatic classifiers based on a deep convolutional neural network (CNN) with two models, a simpler feedforward model with dual modules and an Inception Resnet v2 model, and various structural tweaks for classifying the data from the two tasks. For segregating multi-sensor data, we trained a deep CNN-based classifier on an image dataset extracted from the data by a novel image generating method. We created two deepened and one reductive feedforward network for DR phase classification. The validation accuracies and visualization results show that increasing deep CNN structure depth or kernels number in convolutional layers will not indefinitely improve the classification quality and that a more sophisticated model does not necessarily achieve higher performance when training datasets are quantitatively limited, while increasing training image resolution can induce higher classification accuracies for trained CNNs. The methodology aims at providing support for devising classification networks powering intelligent sensors.

In the recent past, a few fire warning and alarm systems have been presented based on a combination of a smoke sensor and an alarm device to design a life-safety system. However, such fire alarm systems are sometimes error-prone and can react to non-actual indicators of fire presence classified as false warnings. There is a need for high-quality and intelligent fire alarm systems that use multiple sensor values (such as a signal from a flame detector, humidity, heat, and smoke sensors, etc.) to detect true incidents of fire. An Adaptive neuro-fuzzy Inference System (ANFIS) is used in this paper to calculate the maximum likelihood of the true presence of fire and generate fire alert. The novel idea proposed in this paper is to use ANFIS for the identification of a true fire incident by using change rate of smoke, the change rate of temperature, and humidity in the presence of fire. The model consists of sensors to collect vital data from sensor nodes where Fuzzy logic converts the raw data in a linguistic variable which is trained in ANFIS to get the probability of fire occurrence. The proposed idea also generates alerts with a message sent directly to the user's smartphone. Our system uses small size, cost-effective sensors and ensures that this solution is reproducible. MATLAB-based simulation is used for the experiments and the results show a satisfactory output., Competing Interests: The authors declare no conflict of interest.

Sericulture is traditionally a labor-intensive rural-based industry. In modern contexts, the development of process automation faces new challenges related to quality and efficiency. During the silkworm farming life cycle, a common issue is represented by the gender classification of the cocoons. Improper cocoon separation negatively affects quantity and quality of the yield resulting in disruptive bottlenecks for the productivity. To tackle this issue, this paper proposes a multi sensor system for silkworm cocoons gender classification and separation. Utilizing a load sensor and a digital camera, the system acquires weight and digital images from individual silkworm cocoons. An image processing procedure is then applied to extract significant shape-related features from each image instance, which, combined with the weight data, are provided as inputs to train a Support Vector Machine-based pattern classifier for gender classification. Subsequently, an air blower mechanism and a conveyor system sort the cocoons into their respective bins. The developed system was trained and tested on two different types of silkworm cocoons breeds, respectively CSR2 and Pure Mysore. The system performances are finally discussed in terms of accuracy, robustness and computation time.

The application of the Micro Electro-mechanical System (MEMS) inertial measurement unit has become a new research hotspot in the field of inertial navigation. In order to solve the problems of the poor accuracy and stability of MEMS sensors, the redundant design is an effective method under the restriction of current technology. The redundant data processing is the most important part in the MEMS redundant inertial navigation system, which includes the processing of abnormal data and the fusion estimation of redundant data. A developed quality index of the MEMS gyro measurement data is designed by the parity vector and the covariance matrix of the distributed Kalman filtering. The weight coefficients of gyros are calculated according to this index. The fault-tolerant fusion estimation of the redundant data is realized through the framework of the distributed Kalman filtering. Simulation experiments are conducted to test the performance of the new method with different types of anomalies.

It is necessary to measure accurately the rolling noise generated by the friction between wheels and rails in railway transport systems. Although many systems have recently been developed to measure the surface roughness of wheels and rails, there exist large deviations in measurements between each system whose measuring mechanism is based on a single sensor. To correct the structural problems in existing systems, we developed an automatic mobile measurement platform, named the Automatic Rail Checker (ARCer), which measures the acoustic roughness of a longitudinal railhead profile maintaining a constant speed. In addition, a new chord offset synchronization algorithm has been developed. This uses three displacement sensors to improve the measuring accuracy of the acoustic roughness of a longitudinal railhead profile, thereby minimizing the limitations of mobile platform measurement systems and measurement deviation. We then verified the accuracy of the measurement system and the algorithm through field tests on rails with different surface wear conditions.

Three-dimensional (3D) object detection has important applications in robotics, automatic loading, automatic driving and other scenarios. With the improvement of devices, people can collect multi-sensor/multimodal data from a variety of sensors such as Lidar and cameras. In order to make full use of various information advantages and improve the performance of object detection, we proposed a Complex-Retina network, a convolution neural network for 3D object detection based on multi-sensor data fusion. Firstly, a unified architecture with two feature extraction networks was designed, and the feature extraction of point clouds and images from different sensors realized synchronously. Then, we set a series of 3D anchors and projected them to the feature maps, which were cropped into 2D anchors with the same size and fused together. Finally, the object classification and 3D bounding box regression were carried out on the multipath of fully connected layers. The proposed network is a one-stage convolution neural network, which achieves the balance between the accuracy and speed of object detection. The experiments on KITTI datasets show that the proposed network is superior to the contrast algorithms in average precision (AP) and time consumption, which shows the effectiveness of the proposed network.

Objective : To explore the validity of energy expenditure estimates using the SenseWear Armband during a sequence of four daily living activities in patients post-stroke. Method : Patients with stroke who were able to walk during 6 min without human assistance were asked to wear the SenseWear Armband on the non-paretic arm while performing transfers, a manual task, walking, and walking up and down stairs. The energy expenditure estimated using the SenseWear Armband was compared to the energy expenditure calculated from oxygen consumption, measured by a portable indirect calorimeter (Metamax 3B). The mean of energy expenditure was pooled for each task. Accuracy was explored by mean bias (MB) of Bland-Altman analysis and root mean square error (RMSE), agreement by 95% of limits of agreement (95%LoA) and coefficient of correlation (r). Results : Thirty-eight participants (65.7 ± 13.5 years) were included. The SenseWear Armband globally underestimated energy expenditure, MB = 9.77 kcal for the whole sequence. RMSE were large, accounting for 15% to 41% of the measured energy expenditure. Agreement was low with r < 0.70 and 95%LoA from 42% to 93% of the measured energy expenditure. Conclusions : This study reported a global underestimation and a low level of agreement of the energy expenditure estimated by SenseWear Armband in four daily living activities in patients after stroke. Abbreviations: EE: Energy Expenditure; NIHSS: National Institute of Health Stroke Score.