Fine-grained visual classification (FGVC) in sustainable manufacturing faces challenges due to the diverse, complex, and highly similar objects in manufacturing environments. Traditional convolutional neural networks often require extensive annotations and high computational costs, limiting their effectiveness. This study introduces a fuzzy optimization feature fusion model (FOFFM) based on vision transformer, designed to enhance FGVC accuracy and efficiency. FOFFM addresses challenges, such as information loss during image-to-token mapping and high category similarity, by optimizing classification token capabilities and leveraging contrastive loss. By enhancing resource efficiency and reducing redundant computations, FOFFM contributes to lower energy consumption and operational costs, directly supporting sustainable manufacturing practices. Experimental results on the NABirds dataset demonstrate FOFFM's competitive performance with a streamlined, resource-efficient end-to-end training process. Unlike other methods, such as TransFG, FOFFM reduces computational complexity while maintaining robust accuracy, making it highly suitable for practical applications in sustainable manufacturing, particularly in optimizing resource utilization and minimizing environmental impact. This work provides valuable insights for manufacturing data analysis and contributes to advancing FGVC in sustainable manufacturing contexts.

Many countries worldwide aim to proactively conduct digital transformation in conventional industries as these businesses require a massive workforce to classify, inspect, and package, causing high costs. Moreover, manual approaches might lead to defective products due to human errors and productivity limits. This article develops an Internet of Things (IoT) system with an edge computing function to detect agricultural products' quality and classify them. The proposed system in this research detects the luster, weight, and appearance of agricultural products to judge the quality; hence, the equipment needs sensors and a camera that connects to the Artificial Intelligence (AI) module. The features of our research are as follows: 1. The system analyzes agricultural products' appearance, luster, and weight; when the equipment detects an object, the AI module will be activated to analyze the item; 2. The edge computing function in the system allows it to pass the task to the next step directly after the IoT development board completes the process without centralized operations on a server; 3. The 5G network used in this study ensures data transmission and reduces unnecessary circuits between devices; 4. The network safety mechanism built into the system also protects data transmission security. Additionally, this study creates a small-scale conveyor to experiment with the design and test the system's practical feasibility.

As today’s transportation systems have gradually developed into intelligent transportation systems, the fundamental elements of the system consist of vessels, harbors, and ship-shore information-communication technology applications. Safety is the priority for intelligent transportation, securing the transportation environment for vessels. Because there are more dangerous situations when sailing in the sea, our research deploys artificial intelligence of thing with the 5G network to ensure ship safety. To achieve the goal of intelligent ships for vessels to share information between groups; moreover, the geofencing technology can protect vessels from sailing into risky zones. The security mechanism of our system can detect malicious attacks from Dynamic Domain Name System and radio jamming attacks. Additionally, the network security mechanism proposed in this article can safeguard data reliability and safety, enabling vessels to detect collisions in the front and improve safety through the 5G network and the sensor. The performance analysis has proven that the network security approach of this article surpasses other studies; regarding the geofencing part, this article has also conducted a practical experiment to introduce it into the automatic-identification-system (AIS) and 5G system. The experimental results prove that the suggested approach can ensure the AIS network security in vessels; moreover, the system can precisely judge whether there are obstacles in front of the ship, making sure the vessel safety.

Long-term depression and negative emotional cycles affect life quality and work productivity. However, depression is not easy to detect, with current methods mostly relying on scales that make it impossible to quickly and directly measure the severity of depression. This study seeks to empirically identify brainwave stimulation feedback electrode points and brain regions related to potential depression. Using brainwave data collected by mood-induction procedures, the front and occipital lobes have the greatest role in the operation of depressive emotions, especially the Fp1 and Fp2 positions and the O1 and O2 positions. The Fourier brainwave bands are mainly affected in the α and θ band, while the wavelet brainwave bands have a significant impact on the minimum value of approximated signals. This study uses two signal processing methods, combined with deep neural network techniques (Multilayer perceptron, Deep neural network, Deep belief network, and Long Short-Term Memory) to develop 8 potential depression assessment models, with models constructed using deep neural networks providing the best and most stable performance. Therefore, this model can be developed as an auxiliary system for rapid and objective assessment of underlying depression, thereby assisting in the autonomous management of emotions and early detection and treatment of depression. In addition, the individual abnormality is found in the low mood stage and appropriate relief methods are provided, potentially reducing the occurrence of depression.

The proliferation of mobile networked devices has made it easier and faster than ever for people to obtain and share information. However, this occasionally results in the propagation of erroneous information, which may be difficult to distinguish from the truth. The widespread diffusion of such information can result in irrational and poor decision making on potentially important issues. In 2020, this coincided with the global outbreak of Coronavirus Disease (COVID-19) , a highly contagious and deadly virus. The proliferation of misinformation about COVID-19 on social media has already been identified as an “infodemic” by the World Health Organization (WHO) , posing significant challenges for global governments seeking to manage the pandemic. This has driven an urgent need for methods to automatically detect and identify such misinformation. The research uses multiple deep learning model frameworks to detect misinformation in Chinese and English, and compare them based on different text feature selection s. The model learns the textual characteristics of each type of true and misinformation for subsequent true/false prediction. The long and short-term memory (LSTM) model, the gated recurrent unit (GRU) model, and the bidirectional long and short-term memory (BiLSTM) model were selected for fake news detection. BiLSTM produces the best detection result, with detection accuracy reaching 94% for short-sentence English texts, and 99% for long-sentence English texts, while the accuracy for Chinese texts was 82% .

It is worth noting that this 21st century has experienced so many economic, social, cultural and political turbulences throughout the world. The 2019 novel coronavirus (COVID-19) outbreak has been regarded by the World Health Organization (WHO) as a public health crisis of global concern. Nowadays, the chest X-ray (CXR) and chest computed tomography (CT) are a more effective imaging technique for diagnosing lung related problems. Deep learning has been more mature in the field of supervised learning, but other areas of machine learning have just started, especially for the areas of unsupervised learning and reinforcement learning. Deep learning has very good performance in speech recognition and image recognition. Using deep learning approaches to diagnose COVID-19 can achieve better cures and treatments. This research presents the data augmentation and L2 regularization approach for transfer learning in several state-of-the-art deep learning models such as VGG16, VGG19, ResNet, and AlexNet, with Convolutional Block Attention Module (CBAM) to perform binary classification ( such as normal and COVID-19/pneumonia cases or COVID-19 and pneumonia cases) and also multi-class classification (such as COVID-19, pneumonia, and normal cases) of covid-chestxray-dataset and NIH datasets. In addition, the performance evaluation adopted the confused matrix to evaluate the results of these models. To sum up, the CBAM can improve the accuracy of all deep learning models to achieve better performance in contrast to that without this architecture. The findings can be a reference for the related COVID-19 diagnosis researches especially during the post-pandemic era.

introduction Share on Introduction to the Special Section on Internet of Behavior for Emerging Technologies Authors: Mu-Yen Chen National Cheng Kung University, Taiwan National Cheng Kung University, Taiwan 0000-0002-3945-4363View Profile , Vincenzo Piuri University of Milan, Italy University of Milan, Italy 0000-0003-3178-8198View Profile , Alireza Souri Haliç University, Turkey Haliç University, Turkey 0000-0001-8314-9051View Profile , Mohammad Shojafar University of Surrey, UK University of Surrey, UK 0000-0003-3284-5086View Profile Authors Info & Claims ACM Transactions on Sensor NetworksVolume 19Issue 2Article No.: 23pp 1–3https://doi.org/10.1145/3589021Published:16 May 2023Publication History 0citation21DownloadsMetricsTotal Citations0Total Downloads21Last 12 Months21Last 6 weeks21 Get Citation AlertsNew Citation Alert added!This alert has been successfully added and will be sent to:You will be notified whenever a record that you have chosen has been cited.To manage your alert preferences, click on the button below.Manage my AlertsNew Citation Alert!Please log in to your account Save to BinderSave to BinderCreate a New BinderNameCancelCreateExport CitationPublisher SiteGet Access

Many countries have increasingly aging populations. Many of these elderly people live alone and independently, but those suffering from chronic diseases or disabilities are at risk of accidents that will require assistance. Many commercially available homecare systems provide remote monitoring functionality, but these systems require someone on the other end of the remote connection to be paying attention. This raises the need for intelligent home monitoring systems to ensure continuous awareness of user safety. This article proposes a system using beacon technology to assess subject well being based on lack of movement, and automatically activates prepositioned cameras and send an alert to family members or caregivers in response to potentially high-risk activity, such as activating particular kitchen appliances. However, the use of such cameras raises privacy concerns. To improve object readability and privacy, our research uses federated learning to enhance the readability of fast regions with convolutional neural networks features (RCNN). The presented approach stores the personal information of the elderly in the local server, which avoids revealing their home information and secures the data transmission safety by privacy protection. The article has proved the feasibility and practicality by running the experiment; the system can genuinely operate in home health care.

Rapid breakthroughs in information technologies have driven substantial developments in artificial intelligence applications, particularly the widespread use of deep learning techniques in domains such as speech, image and text recognition. However, real world data distribution applications suffer from significant problems including data imbalance which can easily lead to machine learning biased towards the side with more data, resulting in inaccurate classification or prediction results. Therefore, effectively addressing data imbalance is a pressing research topic. Generative Adversarial Networks (GAN) addresses data imbalance, but is prone to vanishing gradients. Recent work has thus focused on improving the GAN architecture to resolve this problem. The present research extends these efforts, applying C4.5, Random Forest, Support Vector Machine, K-Nearest Neighbor and Naïve Bayes classification algorithms to a single imbalanced traffic collision dataset to identify methods for improving prediction results. Experimental results show that classification performance significantly improves after data augmentation using Synthetic Minority Oversampling Technique, GAN, Conditional GAN, and Gaussian Discriminant Analysis GAN as compared with the non-augmented dataset. In addition, the Gaussian Discriminant Analysis GAN with Naïve Bayes classifier produces a dataset that optimizes classification performance for traffic accident prediction at highway intersections.

Drivers and traffic system planners require accurate forecasting of future travel times. To alleviate traffic congestion on Taiwan’s Provincial Highway No. 61, this study considers temporal (weekdays, weekends or continuous holidays) and spatial characteristics (road types), and establishes short-term and long-term travel time prediction models. Data pre-processing is accomplished using the Google Maps API and floating car method to verify travel time comparisons, finding that post-processing travel times are reliable, credible and reasonable. This study develops a collaborative intelligent transportation system (CITS) based on 9 different algorithms for the prediction of current and future travel-time. The results show that$k$NN-R provides the most accurate short-term predictions, with average prediction error within 20 seconds per kilometer. For long-term forecasting, SARIMAX and fbProphet provide the most accurate results for weekday and continuous holiday modes. Travel time prediction can assist traffic management agencies in the timely implementation of appropriate traffic management measures. CITS also provides real-time traffic query and travel time prediction functions to help drivers avoid traffic congestion.

In this study, a high accident risk prediction model is developed to analyze traffic accident data, and identify priority intersections for improvement. A database of the traffic accidents was organized and analyzed, and an intersection accident risk prediction model based on different mechanical learning methods was created to estimate the possible high accident risk locations for traffic management departments to use in planning countermeasures to reduce accident risk. Using Bayes’ theorem to identify environmental variables at intersections that affect accident risk levels, this study found that road width, speed limit and roadside markings are the significant risk factors for traffic accidents. Meanwhile, Naïve Bayes, Decision tree C4.5, Bayesian Network, Multilayer perceptron (MLP), Deep Neural Networks (DNN), Deep Belief Network (DBN) and Convolutional Neural Network (CNN) were used to develop an accident risk prediction model. This model can also identify the key factors that affect the occurrence of high-risk intersections, and provide traffic management departments with a better basis for decision-making for intersection improvement. Using the same environmental characteristics as high-risk intersections for model inputs to estimate the degree of risk that may occur in the future, which can be used to prevent traffic accidents in the future. Moreover, it also can be used as a reference for future intersection design and environmental improvements.

The heart is one of the most important organs of the human body. It circulates blood throughout the human body and delivers oxygen and nutrients to all the organs for metabolism. Cardiac muscle contraction results in blood circulation, which maintains the body temperature at approximately 37 °C. If the cardiac function is abnormal, then the body temperature will be affected. The cardiac function degenerates as the human body ages, and the degeneration can occasionally result in cardiovascular diseases. When the Internet of Medical Things is integrated into heart disease screening systems to detect heart diseases, people can perform self-examinations to evaluate whether their hearts exhibit irregularities for early heart disease detection. STM32 is used in this study as the main Internet-of-Medical Things controller and is combined with the Internet of Things devices—a sphygmomanometer cuff, temperature sensor, and pulse sensor—for instrument control and data acquisition. This assembly is used to develop a valvular heart disease screening system, whose structure incorporates deep learning for the development of fitting models and analysis. An experiment is performed where blood flow is blocked temporarily and released to observe changes in the surface temperature of the fingertip skin, and the blood supply capability of the heart is assessed indirectly based on the temperature change curve. Eighteen subjects were recruited in the experiment, where one subject exhibited cardiac valve insufficiency and arrhythmia. In the experiment, temperature curve variation data are successfully obtained from the healthy subjects, whereas the temperature curve irregularities of the patient with cardiac valve insufficiency are identified. This subject’s temperature range throughout three test steps is smaller by within 0.52 °C compared with those of most of the other subjects. In addition, during blood blocking and release, the overall temperature curve decreases, whereas some curves escalate first before plummeting slowly. The data analysis results show that the temperature curve variation and values of Subject 2 are similar to those of Subject 10, suggesting the incidence of valvular heart disease. This valvular heart disease screening system can successfully analyze and assess the characteristic signal values of patients with valvular heart disease.

Summary Attention‐deficit hyperactivity disorder (ADHD) is a symptom of behavioral or emotional problems as these problems affect children's learning and social integration. With the advancements in the Internet of Things (IoTs), emotions can be detected through image and physiological data. However, some critical ADHD children are often accompanied by the inability to control their body and even facial expressions, making emotion recognition technologies difficult to develop successfully. This study aims to predict the emotions of ADHD children and to address their emotional problems with related IoT robotic devices. Data fusion analysis technology for facial expressions was used to combine thermal images and recognition data, while deep reinforcement learning technology was used to periodically stream information for ADHD students, in alignment with intervention strategies that were designed to address behavioral problems.

Topological data analysis is a new theoretical trend using topological techniques to mine data. This approach helps determine topological data structures. It focuses on investigating the global shape of data rather than on local information of high-dimensional data. The Mapper algorithm is considered as a sound representative approach in this area. It is used to cluster and identify concise and meaningful global topological data structures that are out of reach for many other clustering methods. In this article, we propose a new method called the Shape Fuzzy C-Means (SFCM) algorithm, which is constructed based on the Fuzzy C-Means algorithm with particular features of the Mapper algorithm. The SFCM algorithm can not only exhibit the same clustering ability as the Fuzzy C-Means but also reveal some relationships through visualizing the global shape of data supplied by the Mapper. We present a formal proof and include experiments to confirm our claims. The performance of the enhanced algorithm is demonstrated through a comparative analysis involving the original algorithm, Mapper, and the other fuzzy set based improved algorithm, F-Mapper, for synthetic and real-world data. The comparison is conducted with respect to output visualization in the topological sense and clustering stability.

The popularity of 5G networks has made smart manufacturing not limited to high-tech industries such as semiconductors due to its high speed, ultra-high reliability, and low latency. With the advance of system on chip (SoC) design and manufacturing, 5G is also suitable for data transmission in harsh manufacturing environments such as high temperatures, dust, and extreme vibration. The defect of the screw head is caused by the wear and deformation of the die forming the head after mass production. Therefore, the screw quality inspection system based on the tactile network in this article monitors the production quality of the screw; the system will send a warning signal through the router to remind the technician to solve the production problem when the machine produces a defective product. Sensors are embedded into the traditional screw heading machine, and sensing data are transmitted through a gateway to the voluntary computing node for screw slot quality inspection. The anomaly detection data set collected by the screw heading machine has a ratio of anomaly to normal data of 0.006; thus, we propose a time-series deep AutoEncoder architecture for anomaly detection of screw slots. Our experimental results show that the proposed solution outperforms existing works in terms of efficiency and that the specificity and accuracy can reach 97% through the framework proposed in this article.

MOTIVATION: Ovarian cancer is the fifth leading cause of cancer deaths in women in the western world for 2013. In ovarian cancer, benign tumors turn malignant, but the point of transition is difficult to predict and diagnose. The 5-year survival rate of all types of ovarian cancer is 44%, but this can be improved to 92% if the cancer is found and treated before it spreads beyond the ovary. However, only 15% of all ovarian cancers are found at this early stage. Therefore, the ability to automatically identify and diagnose ovarian cancer precisely and efficiently as the tissue changes from benign to invasive is important for clinical treatment and for increasing the cure rate. This study proposes a new ovarian carcinoma classification model using two algorithms: a novel discretization of food sources for an artificial bee colony (DfABC), and a support vector machine (SVM). For the first time in the literature, oncogene detection using this method is also investigated. RESULTS: A novel bio-inspired computing model and hybrid algorithms combining DfABC and SVM was applied to ovarian carcinoma and oncogene classification. This study used the human ovarian cDNA expression database to collect 41 patient samples and 9600 genes in each pathological stage. Feature selection methods were used to detect and extract 15 notable oncogenes. We then used the DfABC-SVM model to examine these 15 oncogenes, dividing them into eight different classifications according to their gene expressions of various pathological stages. The average accuracyof the eight classification experiments was 94.76%. This research also found some oncogenes that had not been discovered or indicated in previous scientific studies. The main contribution of this research is the proof that these newly discovered oncogenes are highly related to ovarian or other cancers. AVAILABILITY AND IMPLEMENTATION: http://mht.mis.nchu.edu.tw/moodle/course/view.php?id=7.

This study presents a similarity-determining method for measuring regulatory relationships between pairs of genes from microarray time series data. The proposed similarity metrics are based on a new method to measure structural similarity to compare the quality of images. We make use of the fact that dual-tree wavelet transform (DTWT) can provide approximate shift invariance and maintain the structures between pairs of regulation-related time series expression data. Despite the simplicity of the presented method, experimental results demonstrate that it enhances the similarity index when tested on known transcriptional regulatory genes.

Radar image denoising with wavelet packets has been studied in this paper. We propose a recursive thresholding and reconstruction algorithm on wavelet packet coefficients to obtain a clear radar scan. Utilizing the same algorithm, we can estimate the noise variance as well. We illustrate the performance of the proposed approach with both synthesized and real radar images.