Search

We aimed to gather evidence on the feasibility of using simulator-based driving assessments for prescribing powered mobility devices (PMDs). Therefore, we compared the driving characteristics of real driving and two types of simulated driving. Thirty participants with difficulty walking more than 100 meters independently were enrolled. We developed a full-cabin and desktop simulator and created driving scenarios that closely resembled a real driving route in a park. They participated in three separate driving sessions, each using a powered wheelchair, full-cabin simulator, and desktop simulator. The driving characteristics, such as driving distance, mean speed, and standard deviation (SD) of speed, were obtained and analyzed to assess differences and correlations. Statistically significant differences were found in the driving distance and the SD of speed, respectively. However, for the mean speed, there was no statistically significant difference among the three types of driving. The intraclass correlation coefficient (ICC) for the driving distance was 0.154, which was not statistically significant. However, for mean speed, the ICC was 0.752, indicating a strong correlation. The ICC for the SD of speed was 0.562, indicating a moderate correlation. We demonstrated that the two types of simulators have characteristics that are similar to real-world driving characteristics. The mean speed showed the highest similarity, and the SD of the speed showed a moderate degree of similarity. These results highlight the significant potential of employing simulator-based driving to evaluate the use of PMDs.

This study investigates the row hammer tolerance and potential degradation by capacitive crosstalk (CC) and parasitic bipolar junction transistor (BJT) effect in vertically stacked dynamic random-access memory (VS-DRAM) using technology computer-aided design (TCAD) simulations. The close arrangement of word lines in VS-DRAM results in a subthreshold leakage by the CC effect. Furthermore, as VS-DRAM has a floating body, hole accumulation in the body occurs via gate-induced drain leakage (GIDL) at the storage node in the cell that stores ’1’. This can be accelerated by activating the bit-line (BL). The accumulated holes cause leakage current ( $I_{BJT}$ ) by the parasitic BJT when the BL state becomes low and it is found that $I_{BJT}$ can be enhanced by the CC effect in this study. The row hammer effect and $I_{BJT}$ by the CC and parasitic BJT effects can be mitigated by reducing Si width.

Conventional text recommendation models often overlooked user contextual information, resulting in low accuracy and efficiency. Therefore, we propose a new model to integrate user spatiotemporal contextual information, so as to improve the recommendation effect. The study extracts spatiotemporal contextual information of users from their behavioral data, including location information, as well as behavioral characteristics. By processing textual semantic analysis, we define graph adjacency matrix and spatial dependencies. Hence, a spatiotemporal graph convolutional neural network model can be constructed to learn the correlation between users’ spatiotemporal contextual features and text items. The proposed model can effectively capture changes in user preferences and interests towards text in different contexts, thereby leading to better recommendation results. After that, some experiments are conducted on real-world dataset to make performance evaluation. The experimental results show that the proposal can achieve significant improvement in recommendation efficiency compared with typical methods. This indicates that proposed model can better understand contextual user information, and improve the user experience in recommendation systems.

The effects of single event upset (SEU) by alpha particles and heavy ions on the data flip of a 3 nm technology node gate-all-around (GAA) nanosheet field-effect transistor (NSFET) 6T static random-access memory (SRAM) cell was studied through technology computer-aided design (TCAD) simulations. It was found that the sensitivity to radiation in the “off” pull-down transistor varies depending on the position of the alpha particle and heavy ions in the incident. The most significant radiation-induced increase in electron density occurs at the drain-channel junction. Heavy ion strikes lead to the lowest threshold linear energy transfer (LETth) value during the hold operation compared to read and write operations. The partial bottom dielectric isolation (PDI) scheme demonstrates lower radiation sensitivity than the conventional scheme, as the PDI layer acts as a physical barrier preventing charge migration from the substrate to the drain.

Elderly care becomes more and more important as elderly populations increase. Many solutions are proposed for monitoring elderly activities of daily living (ADL) or healthcare related data. There are two aspects to be tackled to enable monitoring for elderly care. On the one hand, the up-to-date ADL sensors and biosensors greatly facilitate collection of various ADL and healthcare related data. On the other hand, data captured by sensors need to be gathered from residences of elderly people to local community center for monitoring and analysis. Therefore, gathering sensor data plays a significant role for building elderly care systems or architectures. Most current elderly care systems or architectures are developed based on reliable network infrastructures. However, such presumption is less practical in network deficient environments such as geographically depopulated rural and remote areas. In this paper, we propose a data gathering system model for elderly care based on autonomous device-to-device (D2D) communications, which is self-contained and independent from other network infrastructures. The proposed system model doesn’t need central control; thus, it can be simply deployed in network deficient environments or for users without access to existing networks. The idea behind is to use ad hoc D2D links by introducing moving devices. Procedures of data gathering using D2D are proposed and configured by modifying the two-way untargeted discovery protocols of IEEE Std 802.15.8. To examine the feasibility of the proposed system model, a prototype D2D system is developed using off-the-shelf radio modules and ADL sensors. We examine the efficacy of the prototype system through field trials in a mountainous village. Our obtained results confirm the usefulness of the proposed D2D solution for gathering ADL sensor data for elderly care in network deficient environments. Examples of using the gathered ADL data to detect anomalous events are presented.

Antineoplastic medication compounding is conducted in most large hospitals and involves the repetitive task of extracting a specific quantity of medication from vials using syringes and transferring it into intravenous infusion bags. When a large syringe is used during medication compounding, manipulating the syringe plunger requires considerable force because of the negative pressure inside the vial as well as the viscosity of the drug solution. This repetitive and intense plunger manipulation can lead to musculoskeletal disorders of the arms, shoulders, and other areas. Implementing motorized syringe manipulation and automating the drug compounding process are beneficial approaches for enhancing pharmacist convenience and work efficiency. This study introduces the development of an automated assistance system for medication compounding designed to manipulate syringes and medication containers accurately and reliably. The developed equipment ensures compatibility with various syringes, vials, and intravenous (IV) infusion bags without the need for consumables. It is designed for the diversified, small-quantity production of personalized drugs, featuring a compact, affordable structure that can be easily accommodated within existing drug compounding facilities. The performance of the developed equipment is validated through simulations and experiments. The results prove the developed equipment has high dispensing accuracy (error range: ±0.09%) and high repeatability (standard deviation: 21 mg).

Various training-based spatial filtering methods have been proposed to decode steady-state visual evoked potentials (SSVEPs) efficiently. However, these methods require extensive calibration data to obtain valid spatial filters and temporal templates. The time-consuming data collection and calibration process would reduce the practicality of SSVEP-based brain-computer interfaces (BCIs). Therefore, we propose a temporally local weighting-based phase-locked time-shift (TLW-PLTS) data augmentation method to augment training data for calculating valid spatial filters and temporal templates. In this method, the sliding window strategy using the SSVEP response period as a time-shift step is to generate the augmented data, and the time filter which maximises the temporally local covariance between the original template signal and the sine-cosine reference signal is used to suppress the temporal noise in the augmented data. For the performance evaluation, the TLW-PLTS method was incorporated with state-of-the-art training-based spatial filtering methods to calculate classification accuracies and information transfer rates (ITRs) using three SSVEP datasets. Compared with state-of-the-art training-based spatial filtering methods and other data augmentation methods, the proposed TLW-PLTS method demonstrates superior decoding performance with fewer calibration data, which is promising for the development of fast-calibration BCIs.

The paper presents a novel recognition framework for incomplete noninvasive Electroencephalography (EEG) signals relying on the recent advances in tensor algebra, named as An Innovative Tensor Algebra-based Detection Framework for Incomplete Noninvasive Electroencephalography (iTa-DFiE). iTa-DFiE is motivated to improve the diagnostic performance by tackling the major problems shared by a variety of noninvasive EEG-based Brain-Computer Interfaces (BCIs) application is tensorial structured time series with occlusions. The aforementioned challenge setting is solved on two major thrusts, including: 1) tensor completion: discovering hidden patterns and learning their evolving trends to offer missing values imputation via improvement of standard Kalman Filter approach and 2) tensor decomposition: extracting essential hidden information from multi aspects data tensor via extending the most well-known tensor factorization Tucker. The effectiveness and efficiency of the proposed tensor-based framework is proved via successfully improving the pattern classification results on two real-world noninvasive EEG-based motor imagery BCI with diverse corrupted data scenarios, especially in occurrence of consecutive missing observations. Strikingly, iTa-DFiE also outperforms the conventional matrices-based methods and the state-of-the-art tensor techniques in terms of missing reconstructions and feature extraction as well.

Cast-resin transformers are affected by deterioration due to manufacturing defects and continuous load. Studying PD, which is capable of detecting defects or degradation in advance, is important. With the rapid advancement of AI technologies, research on PD classification using CNN models is being actively conducted. However, due to the black box problem, it is impossible to explain the reasoning behind the learning outcomes. Therefore, relying solely on predictive outcomes of learning for PD classification raises issues of reliability. Recent studies in various fields are progressing with the application of XAI to address the black box issue of CNNs, aiming to identify the criteria used for making predictions. However, research on applying XAI in AI-based PD classification is currently insufficient. Therefore, further study on the implementation of XAI is necessary. In this paper, an excellent CNN model was applied to image classification for PD classification of cast-resin transformers, and the grad-cam model was used for XAI. This approach proposes a method for humans to comprehend the rationale behind the learning outcomes. The training data includes artificial defects created in laboratory settings and noise captured in cast-resin transformers using UHF sensors. Our research demonstrated that PD and noise due to defects can be identified with an accuracy of approximately 97%. The reasons for successful and failed results were analyzed through XAI. Consequently, it was observed that the application of XAI to CNN models leads to the construction of a more reliable model.

Recently, multi-hop flying ad hoc networks (FANETs) have received considerable attention because of their various advantages, such as ease of deployment, scalability, cost, and latency reduction. However, FANETs still encounter serious problems such as a highly dynamic network topology and the limited on-board battery of unmanned aerial vehicles (UAVs). Thus, to increase the coverage probability while reducing unnecessary packet transmissions, realizing efficient and effective relay selection is important. To resolve these issues, this paper proposes a proactive and probabilistic uncovered neighbor-aware relay selection (P2URE) method that considers FANET’s topological dynamics. Specifically, the relay decision factor to identify relay UAV candidates and the probabilistic relay decision to reduce coverage overlaps and minimize redundant packet transmissions are devised. Through extensive simulations, it has been validated that the proposed P2URE method surpasses the benchmark methods in terms of the total number of transmissions, coverage probability, and energy efficiency. This superiority is observed across various network dynamics resulting from variations in the UAV density, speed, neighbor-information update period, and delay interval.

The local dimming mini-LED backlight has become one of the most important technologies for liquid crystal display (LCD), facilitating the improvement on high dynamic range (HDR) and power saving. The halo effect still limits the performance of LCDs employing mini-LED backlight. The impact of the number of local dimming zones on the contrast ratio has been investigated. The halo effect has also been investigated under four typical local dimming algorithms and different local dimming zones. It has been experimentally demonstrated that the halo area gradually shrinks with increasing the number of local dimming zones, contributing to the improvement on the HDR of LCD.

Fluoronitrile (C4F7N) has many advantages, such as high dielectric strength, low global warming potential (GWP), and short atmospheric lifetime compared to sulfur hexafluoride (SF6). C4F7N is mixed with carbon dioxide (CO2) to prevent liquefaction at the operating pressure of high-voltage power equipment. Therefore, it is essential to investigate the dielectric characteristics of the C4F7N/CO2 mixture gas for the development of environmentally friendly power equipment. The dielectric characteristics of C4F7N/CO2 at lower pressures have been previously reported. However, research into the breakdown characteristics of C4F7N/CO2 under high pressure environments is insufficient. Furthermore, the effect of oxygen (O2) content contributing to suppression of carbon soot formation by arc on breakdown characteristics of C4F7N/CO2 has rarely been investigated. Therefore, in this paper, the breakdown characteristics of C4F7N/CO2 according to the mixing ratio and O2 content were investigated while considering a high–pressure environment. The electrode arrangements were determined based on the field utilization factor in order to consider two types of electric field distribution. Alternating current (AC) and negative lightning impulse (-LI) breakdown tests were performed. From these experiments, under an extremely non-uniform electric field, it was confirmed that the dielectric strength of C4F7N 5%/CO2 95% at high pressure was about 75% of the strength of SF6. In a weakly non-uniform electric field, increasing the C4F7N content contributed to the improvement of dielectric strength by 31-77% relative to pure CO2 at the high pressure. Furthermore, it was investigated that the presence of O2 had no significant influence on improving the insulation performance.

Preventing Use-After-Free (UAF) bugs is crucial to ensure temporal memory safety. Against UAF attacks, much research has adopted a well-known approach, lock-and-key, in which unique, disposable locks and keys are first assigned respectively to objects and pointers, and then on every memory access, checked for a match. Attention has been drawn again to this approach by recent work that capitalizes on a vast abundance of virtual address (VA) space in the lock assignment, thus being able to prevent UAFs in stripped binary. However, as this VA-based lock-and-key scheme tends to rapidly consume virtual space, it is likely to suffer from high performance overhead. In this paper, we propose a new scheme, called the VA tagging, whose goal is to tackle this performance problem with the support of the Memory Tagging Architecture (MTA) introduced in several commodity processors. In our scheme, the original VA-based locks are augmented with tags of MTA. As a VA-based lock can be assigned to multiple objects with different tags, the same VA is reused for many objects without compromising temporal safety. We have observed in our experiments that this tagging scheme lowers the VA consumption rate drastically by one order of magnitude. We implement a light-weight memory allocator, Vatalloc, by modifying existing allocators, dlmalloc and jemalloc, to employ the VA tagging scheme for efficient prevention of UAFs. Our evaluation shows that Vatalloc with allocator modifications only incurs 1.70% (on dlmalloc) and 3.05% (on jemalloc) of runtime overhead without considering performance degradation of MTE. As a result of simulating the tagging architecture assuming the worst-case, postulating MTE precise trapping mode incurs performance overhead of 30.9% based on dlmalloc, and 25.5% based on jemalloc. If imprecise mode is assumed, the slowdown is measured 16.9% for dlmalloc and 12.0% for jemalloc respectively. Vatalloc only incurs 19.0% and 3.0% memory overhead for dlmalloc and jemalloc respectively.

Recently, research on autonomous driving has focused on the advent of various deep learning algorithms. The main sensors for autonomous driving include cameras, LiDAR, and radar, but these algorithms primarily focus on image and LiDAR data. This is because radar data is limited compared to image and LiDAR data. To address the lack of data problem, GAN-based translation methods have been proposed. However, these methods also focus only on image and LiDAR data, such as day-to-night translation or sunny-to-adverse weather translation. Since radar data differs depending on radar sensors and radar points are too sparse to learn patterns compared to LiDAR, translation with radar data is a challenging task. Radar is usually utilized as a sensor that is nearly unaffected by the weather. However, it has been confirmed through JARI data collected by us that rain has a negative effect. CycleGAN is useful for data translation in traffic scenes where pair data is difficult to acquire, since CycleGAN is a network specialized in style translation. KP-Convolution is a module specialized in feature extraction of points while maintaining location information. Therefore, we propose a radar translation network between sunny and rainy domains by combining KP-Convolution and CycleGAN. In this process, we address the adverse effects of radar data by rain, establishing the training format of radar data, KP-Convolution which can learn patterns despite a small number of points, and CycleGAN which is the basis of the translation method.

This paper deals with the torque vectoring control problem for uncertain in-wheel motor vehicles. The main objective of this study is to enhance the cornering performance of vehicles at high speed driving and this is done by generating possibly different motor torques applied to wheels. The proposed controller consists of a model predictive control based outer-loop controller and a disturbance observer based inner-loop controller. The outer-loop controller finds the optimal longitudinal tire force considering state and input constraints. The inner-loop controller estimates the difference between the nominal tire model that is used by the outer-loop controller and the unmodeled disturbance, and compensates the disturbance so that the actual vehicle follows the nominal model. The proposed controller is validated via numerical experiments for circular driving, double lane change, and skidpad scenarios and it is shown that the state tracking performance is improved by using the proposed controller.

The black-box arc model is a tool that facilitates efficient research based on empirical studies in breaker design and analysis. However, its application has primarily focused on the design of ACCBs, with only a few instances of its implementation in the design of low-voltage (LV) DC circuit breakers (DCCBs) using mechanical methods. In DCCBs with non-zero crossing characteristics, several factors need to be considered in the design process. Therefore, it is crucial to have a reliable and accurate DC arc model to effectively apply black-box arc model studies to actual DCCB designs. Especially, DC Current limiting circuit breaker (CLCB), which has fast operation characteristics through its own over current relays (OCR), requires reliable and accurate modeling because minor factors in the topology can contribute significantly to the interrupting performance. This paper presents a study on the characterization and modeling approach for the DC CLCB, which exhibits somewhat different arc voltage characteristics compared to the general Air Circuit Breaker (ACB). The black-box arc model incorporates existing schwarz and kema arc models, and the Levenberg-Marquardt Algorithm (LMA) is employed to optimize the parameters of each model. An efficient parameter optimization method for the CLCB model is proposed, and the characteristics of factors that should be considered in the design are identified. Consequently, the applicability and reliability of the model are verified through a comparative study with the short-circuit test results of the actual DC CLCB.

Due to the aggressive scaling down of logic semiconductors, the difficulty of semiconductor component processes has increased. As the structure of components becomes more complex, the time and cost of processes and simulations have risen. Machine learning is now being used to analyze the electrical characteristics data of semiconductor components and apply the trained machine learning to next-generation semiconductor development. Machine learning trained on process data and simulation results can quickly and accurately predict which electrical characteristics change significantly when the component’s structure changes and which parameters have a significant impact on the electrical characteristic changes. This paper presents suitable machine learning models for analyzing and predicting the electrical characteristics (on-current ( $I_{on}$ ), off-current ( $I_{off}$ ), threshold voltage ( $V_{th}$ ), subthreshold swing (SS), and drain induced barrier lowering (DIBL)) and statistical distribution (mean and standard deviation of the electrical characteristics) resulting from geometrical variability (sheet thickness ( $T_{wire}$ ), sheet diameter ( $D_{wire}$ ), oxide thickness ( $T_{ox}$ ), gate length ( $L_{g}$ ), spacer length ( $L_{sp}$ ), gate metal work-function (WF)) in nanosheet field-effect transistor (NSFET), which are a next-generation logic device. Machine learning models, including regulation-based models (Ridge and LASSO) and tree-based models (decision tree (DT), random forest (RF), extreme gradient boost (XGBoost), and light gradient boost machine (LGBM)), are trained on technology computer-aided design (TCAD) simulation data. The LGBM more accurately predicts the electrical characteristics and statistical distribution of the NSFET than the other models. Additionally, we analyze the effect of geometrical variability on the NSFET based on feature importance.

Various vulnerable road user (VRU) protection systems have been proposed based on the edge server paradigm to take advantage of the reduced latency as well as computational offloading to servers. In most existing studies, the authors presume that each edge server receives data from its associated users and takes care of the collision risks among them. Because of this presumption, the collision risks between users associated with different edge servers can be overlooked until one of the users at risk crosses the boundary of the server. Therefore, users located at or near the boundary of the edge server domain can receive late alerts or, more seriously, miss the alert entirely until a collision occurs. To address this hazardous scenario, we propose a scalable VRU protection system (SVPS) with an edge server cooperation mechanism. SVPS minimizes additional communication and computational overhead while maintaining satisfactory service accuracy even if users are moving. The numeric results demonstrate that SVPS effectively predicts users’ risks associated with different edge servers. Furthermore, SVPS is demonstrated to be scalable: The larger the edge server coverage area, the lower the overhead. Therefore, the coverage area should be set as large as possible while still satisfying latency requirements.

In this work, the effect of displacement defect (DD) owing to cosmic rays on six-transistor (6T) static random access memory (SRAM) with a 3 nm node nanosheet field-effect transistor (NSFET) is investigated using technology computer-aided design (TCAD) simulation. In order to comprehensively study the uncertainty of the radiation of NSFET 6T SRAM, the shape of the DD cluster cross-section and the transistor damaged by the DD in 6T SRAM are considered. Read static noise margin (RSNM) degradation (19 %) is the highest when the rectangular cross-section of the DD cluster (rectangular-DD cluster) is located in the pull-down1 (PD1) transistor. To mitigate the rectangular-DD cluster damage, we studied the variation in the DD cluster influence on the sheet shape and the source/drain (S/D) overlap length fluctuation. The sheet shape resulted in 2.3 % lower RSNM degradation in NS compared with nanowire (NW). Under the worst conditions (PD1 transistor damaged rectangular-DD cluster, NW structure), the S/D underlap structure showed 3.7 % lower RSNM degradation than the S/D overlap structure.

In this study, experiments were performed to measure the inception and breakdown voltages of SF6/CO2 mixed gases with varying mixing ratios at two different pressures (1 and 2 bar). Using the preliminary data obtained from the experiments, a 3D particle model was used to investigate positive streamer initiation in SF6/CO2 with field ionization. Field ionization using Zener’s model was assumed to be similar to the electron detachment in air. The experimental results showed that at $P=1$ bar, the discharge inception voltage increased for CO2 concentrations lower than 50% and decreased for CO2 concentrations higher than 50%. The inception voltage of 80%CO2/20%SF6 was close to the inception voltage of pure SF6. A similar trend of the inception and breakdown voltages was observed at $P=2$ bar. The simulation results show the evolution of the thin filaments of the streamer into three distinct stages. The streamer morphology was similar irrespective of the type and concentration of the mixed gas. Other parameters such as the electric field, electron density, and streamer length increased with increasing CO2 concentration in SF6 at both pressures. Zener’s field ionization could be considered a primary ionization mechanism for SF6/CO2 in the case of very high non-uniform electric fields; however, finding the experimental validation of Zener’s field ionization for electronegative gases is still a challenging task.

In steady-state visual evoked potential (SSVEP)-based brain-computer interfaces (BCIs), various spatial filtering methods based on individual calibration data have been proposed to alleviate the interference of spontaneous activities in SSVEP signals for enhancing the SSVEP detection performance. However, the time-consuming calibration session would increase the visual fatigue of subjects and reduce the usability of the BCI system. The key idea of this study is to propose a cross-subject transfer method based on domain generalization, which transfers the domain-invariant spatial filters and templates learned from source subjects to the target subject with no access to the EEG data from the target subject. The transferred spatial filters and templates are obtained by maximizing the intra- and inter-subject correlations using the SSVEP data corresponding to the target and its neighboring stimuli. For SSVEP detection of the target subject, four types of correlation coefficients are calculated to construct the feature vector. Experimental results estimated with three SSVEP datasets show that the proposed cross-subject transfer method improves the SSVEP detection performance compared to state-of-art methods. The satisfactory results demonstrate that the proposed method provides an effective transfer learning strategy requiring no tedious data collection process for new users, holding the potential of promoting practical applications of SSVEP-based BCI.

To investigate impacts of intelligent and fashion factors of sports bras on consumers’ emotions, decision-making and behavior, a quantitative analysis method combing macro affective computing and micro emotion data was proposed. The context where a consumer purchased sports bras was first simulated. In this process, an eye tracker and a multi-channel physiological recorder were utilized to collect physiological signal data from participants in an experimental setting. Then, big data and machine learning were both adopted to macroscopically perform data pre-processing, build a computational model, fulfill relevant prediction and evaluation, analyze correlations in physiological data features, and explore potential values existing in data. Furthermore, highly correlated data features were extracted to investigate micro causalities and identify reasons why consumer behavior and decision-making were supported by data about emotional physiology. The proposed method may provide considerably reliable data support for designers, product service providers, and other practitioners. As an innovative and universal integration approach, it has the potential to be applied in medical science, psychology, management science and other fields.

Various website fingerprinting attacks (WF) have been developed to detect anonymous users accessing illegal websites in Tor networks by analyzing Tor traffic. These attacks consider several traffic features, such as packet length, number of packets, and time, to identify users who attempt to access prohibited content. Due to the advance of artificial intelligence (AI) technologies, machine learning or deep learning techniques have been widely adopted for WF to generate an accurate model to break the privacy of illegal users. Nevertheless, such state-of-the-art approaches to WF assumed that entire data from various Tor nodes are collected and trained in a centralized way to generate the model: However, training data sets from Tor nodes may contain sensitive information that the Tor nodes may not want to share. In addition, significant computing and network bottleneck at the centralized server is inevitable in collecting and training various data in a centralized manner. Correspondingly, this paper proposes a novel framework using federated learning (FL) for WF in the Tor network (denoted as FedFingerprinting), enabling Tor nodes to generate the global model collaboratively without exposing their local data sets. Specifically, to alleviate the burden for local training of selected Tor nodes in the FL process, the importance of various handcrafting features used for WF is firstly evaluated through the analysis of the accuracy of features under the ensemble of tree machine learning methods. Then, to balance the accuracy and training time, the combination of selected top-ranked features is trained using FL approaches rather than raw data in the model. Moreover, considering the local model accuracy of each Tor node, effective Tor node selection for the FL process is also designed. Finally, under closed-world settings with the real-world Tor data sets, we empirically demonstrate the comparisons of the proposed FedFingerprinting with raw data and feature selection compared to various benchmarks in terms of the training time and accuracy. Then, the superior performance of the FedFingerprinting with Tor node selection is evaluated in terms of convergence speed.

Vision Transformer (ViT) has been proposed as a new image recognition method in the field of computer vision. ViT applies a Transformer structure with excellent performance in the field of natural language processing to recognize images. Unlike existing Convolutional Neural Network (CNN) models, ViT can achieve State-Of-The-Art (SOTA) image recognition without inputting Inductive Biases into the model, demonstrating that the Transformer is a useful structure in the field of computer vision. However, ViT requires large datasets such as ImageNet-21K and Joint Foto Tree (JFT) for learning. In addition, it takes a lot of time to train. Moreover, there is a problem that location information is lost by inputting images in patch units. To improve such issues, many models are being proposed. In this paper, a new model is proposed by restructuring the Convolutional neural networks Meet vision Transformers (CMT) model by applying Coordinate Attention Block, a CNN model, to improve problems of the Vision Transformer family of models. The proposed model combines Transformer, which has shown excellent performance in Long Range, and CNN, which has shown excellent performance in Local Feature, to achieve higher performance than existing models. We also compared performance of the proposed model with those of existing models with relatively small datasets such as Canadian Institute For Advanced Research-10 (CIFAR-10), Self-Taught Learning-10 (STL-10), and Tiny-ImageNet to facilitate researchers’ access to the evaluation. Despite being restructured from the smallest CMT-Tiny model, the proposed model showed better accuracy than CMT-Tiny, CMT-XS, CMT-S, and CMT-B models with CIFAR-10, STL-10, and Tiny-ImageNet datasets. The proposed model showed an accuracy of 90.21% with the CIFAR-10 dataset, higher than existing CMT models except for the CMT-S model with an accuracy of 90.6%. It had the lowest loss value of 0.3967. The proposed model is expected to be utilized as a backbone in Object Detection and Segmentation fields in the future.

Medium voltage direct current (MVDC) transmission system are growing due to their assistive quality in conventional grid and compatibility with renewable power network. MVDC distribution links with “Mixed” overhead (OH) & underground (UG) sections could be devised based on urban planning. UG Cables or substations are indirectly exposed to lightning strikes due to adjacent tower sections. In case of MVDC converter or cable, present researchers do not specify lightning voltage impulse level for related system voltage. Therefore, preluding electromagnetic (EM) transient investigation are required to determine the maximum lightning overvoltages for system peripherals i.e. cable & Modular Multilevel Converter (MMC) substation. This research focuses on analyzing lightning performance of OH transmission towers-cable junction & tower-substation link in case of a shielding failure (SF) and back flashover (BF) for a ±35kV short-length mixed MMC-MVDC transmission scheme. This article provides broad-band modeling method for MMC substation for lightning investigation. In addition, based on a detailed time-domain parametric evaluation in PSCAD/EMTDC program, lightning impulse voltage across the transmission line’s pole insulator and embedded cable section are estimated along with numerical validation relying on travelling wave theory. Effect of project parameters such as tower grounding resistance, riser section surge impedance (which connects cable & OH line) and cable length on lightning overvoltage impacting the cable and connected tower section has been demonstrated.

In this work, the methodology using Graph Attention Network (GAT) for the reserve design in GaN power MIS-HEMTs based on hand-drawn characteristics is demonstrated for the first-time. The hand-drawn ID-VG characteristic is constructed by Ramer-Douglas-Peucker algorithm. Then, the extracted information is sent to the Graph Attention Network to receive the corresponding device design variables, including tAlGaN, recessed depth, Al%, Lg, Lgd, and Lgs. Less than 30 seconds is consumed to generate the design variables and less than 8% of the differences in the key extracted parameters, such as threshold voltage (Vth), On-state current (Ion), and subthreshold slope (SS), can be achieved by comparing hand-drawn ID-VG and simulated ID-VG characteristic based on the design variables from GAT model. Therefore, the developed GAT approach is promising for the reverse design of GaN power MIS-HEMTs, which can provide users with efficient and valuable design suggestions to optimize the devices toward the targeting performance.

Recently, neural video compression networks have obtained impressive results. However, previous neural video compression models mostly focus on low-delay configuration with the order of display being the same as the order of coding. In this paper, we propose a hierarchical random access coding approach that exploits bidirectionally temporal redundancy to improve the coding efficiency of existing deep neural video compression models. The proposed framework applies a video frame interpolation network to improve inter-frame prediction. In addition, a hierarchical coding structure is also proposed in this paper. Experimental results show the proposed framework improves the coding efficiency of the base deep neural model by 48.01% with the UVG dataset, 50.96% with the HEVC-class B dataset, and outperforms the previous deep neural video compression networks.

Fault diagnostics are extremely important to decide proper actions toward fault isolation and system restoration. The growing integration of inverter-based distributed energy resources imposes strong influences on fault detection using traditional overcurrent relays. This paper utilizes emerging graph learning techniques to build new temporal recurrent graph neural network models for fault diagnostics. The temporal recurrent graph neural network structures can extract the spatial-temporal features from data of voltage measurement units installed at the critical buses. From these features, fault event detection, fault type/phase classification, and fault location are performed. Compared with previous works, the proposed temporal recurrent graph neural networks provide a better generalization for fault diagnostics. Moreover, the proposed scheme retrieves the voltage signals instead of current signals so that there is no need to install relays at all lines of the distribution system. Therefore, the proposed scheme is generalizable and not limited by the number of relays installed. The effectiveness of the proposed method is comprehensively evaluated on the Potsdam microgrid and IEEE 123-node system in comparison with other neural network structures.

For several years, mobile edge computing (MEC) has been highlighted as a promising technique to support emerging computation-intensive applications in cellular networks, e.g., 5G and 6G. Most previous studies have mainly focused on jointly optimizing communication (i.e., radio) and computation resources to improve offloading performance without considering specific communication protocols required for operating the MEC systems in practice. In this article, we newly design a contention-based access reservation protocol (ARP) for efficiently supporting simultaneous task offloading requests from a number of edge devices (EDs), and further define necessary signalings required to determine an optimal offloading factor. Thereafter, we formulate an optimization problem to find the optimal offloading factor that minimizes the task-completion latency. Through simulations, we mainly evaluate the task-completion latency performance of the MEC system incorporated with our proposed access reservation protocol under a multiple-input multiple-output (MIMO) environment. Particularly, we thoroughly investigate several practical aspects such as the effect of the number of edge devices attempting to join the task offloading, the number of antennas equipped at the edge device, and the amount of reserved radio resources required for configuring access reservation protocol on the optimal offloading decision. From the results, we verify the validity of our approach, and provide meaningful insights regarding how the network should be configured to fruitfully exploit the offloading with the MEC systems.

In recent years, deep neural networks (DNNs) have brought revolutionary progress in various fields with the advent of technology. It is widely used in image pre-processing, image enhancement technology, face recognition, voice recognition, and other applications, gradually replacing traditional algorithms. It shows that the rise of neural networks has led to the reform of artificial intelligence. Since neural network algorithms are computationally intensive, they require GPUs or accelerated hardware for real-time computation. However, the high cost and high power consumption of GPUs result in low energy efficiency. It recently led to much research on accelerated digital circuit hardware design for deep neural networks. In this paper, we propose an efficient and flexible neural network training processor for fully connected layers. Our proposed training processor features low power consumption, high throughput, and high energy efficiency. It uses the sparsity of neuron activations to reduce the number of memory accesses and memory space to achieve an efficient training accelerator. The proposed processor uses a novel reconfigurable computing architecture to maintain high performance when operating Forward Propagation and Backward Propagation. The processor is implemented in Xilinx Zynq UltraSacle+MPSoC ZCU104 FPGA, with an operating frequency of 200MHz and power consumption of 6.444W, and can achieve 102.43 GOPS.

Estimating the time of arrival (TOA), pulse width (PW), and pulse repetition interval (PRI) is critical for identifying the low probability of intercept (LPI) radar signals as it is common in electronic warfare (EW) to experience a pulse train being received. To determine these parameters, we need to differentiate between pulse and noise in a pulse train. However, separating the LPI radar pulse train into pulse and noise is challenging owing to its characteristics and modulation scheme. Therefore, we propose a method to accurately estimate TOA, PW, and PRI using change point detection, which can effectively classify the pulse train into pulse and noise. The main contributions of this study are as follows. First, a denoising method robust to various modulation schemes is introduced. Second, an algorithm capable of estimating the parameters without considering the threshold calculation is presented. Moreover, most prior studies have entailed parameter estimations for a single pulse. In contrast, we focus on estimating parameters when a pulse train is intercepted and generating superior results even in the signal-to-noise ratio under 0 dB. We present experiments performed on the following eight modulation schemes: linear frequency modulation, Costas, Barker, Frank, P1, P2, P3, and P4 codes. Furthermore, we compare the proposed method with the wavelet-based method, which has received much attention in EW. The results demonstrate that the proposed method outperforms the conventional approaches.

Device-to-device (D2D) communication in cellular networks refers to a technology that enables direct transmission and reception between devices in proximity without infrastructures such as base stations (BSs). It has consistently attracted attention due to its key role extending cellular coverage through the configuration of clusters, in which the D2D devices belonging to a cluster can either be a cluster head or a cluster member. In this paper, we investigate the effect of channel state information (CSI) forgery attacks of a single malicious user (i.e., attacker) belonging to a cluster. Particularly, we investigate two major threat models: 1) CLUSTERING FAILURE ATTACK, where an attacker reports its overestimated CSI instead of the original one to the BS during clustering, 2) quality of service (QoS) degradation attack where an attacker reports its underestimated CSI instead of the original one to the cluster head during the intra-cluster D2D communications (i.e., D2D multicasting). We define metrics to measure each CSI forgery attack as CLUSTERING FAILURE PROBABILITY and CLUSTER SUM-RATE, respectively, and further derive a closed-form expression for the CLUSTERING FAILURE PROBABILITY. In addition, we propose threshold-based defense mechanisms as countermeasures against CSI forgery attacks and find suboptimal threshold values. Through simulations, we evaluate the performance of the defense mechanisms in terms of CLUSTERING FAILURE PROBABILITY and CLUSTER SUM-RATE, comparing with optimal simulation results.

In this paper, we present a digital processing in memory (DPIM) configured as a stride edge-detection search frequency neural network (SE-SFNN) which is trained through spike location dependent plasticity (SLDP), a learning mechanism reminiscent of spike timing dependent plasticity (STDP). This mechanism allows for rapid online learning as well as a simple memory-based implementation. In particular, we employ a ternary data scheme to take advantage of ternary content addressable memory (TCAM). The scheme utilizes a ternary representation of the image pixels, and the TCAMs are used in a two-layer format to significantly reduce the computation time. The first layer applies several filtering kernels, followed by the second layer, which reorders the pattern dictionaries of TCAMs to place the most frequent patterns at the top of each supervised TCAM dictionary. Numerous TCAM blocks in both layers operate in a massively parallel fashion using digital ternary values. There are no complicated multiply operations performed, and learning is performed in a feedforward scheme. This allows rapid and robust learning as a trade-off with the parallel memory block size. Furthermore, we propose a method to reduce the TCAM memory size using a two-tiered minor to major promotion (M2MP) of frequently occurring patterns. This reduction scheme is performed concurrently during the learning operation without incurring a preconditioning overhead. We show that with minimal circuit overhead, the required memory size is reduced by 84.4%, and the total clock cycles required for learning also decrease by 97.31 % while the accuracy decreases only by 1.12%. We classified images with 94.58% accuracy on the MNIST dataset. Using a 100 MHz clock, our simulation results show that the MNIST training takes about 6.3 ms dissipating less than 4 mW of average power. In terms of inference speed, the trained hardware is capable of processing 5,882,352 images per second.

Burst Buffer is widely used in supercomputer centers to bridge the performance gap between computational power and the high-performance I/O systems. The primary role of Burst Buffer is to temporarily absorb the bursty I/O and reduce the heavy access on Parallel File System (PFS). However, the job resource manager on High-Performance Computer (HPC) systems prefers to use a dedicated Burst Buffer allocation approach, which eventually leads to the severely underutilized Burst Buffer resource. To improve the efficiency of using the expensive Burst Buffer resource, we analyze the I/O patterns on Burst Buffer in depth. We propose Burst Buffer over-subscription allocation method, which improves Burst Buffer utilization by allowing each job to access Burst Buffer only during its I/O phases so that the jobs can overlap each other. Furthermore, we develop a new I/O congestion-aware scheduler and a transparent data management system between Burst Buffer and PFS. Our approach also reduces the memory overhead and improves the data persistence of the data management system by adapting the persistent memory. With the proposed approach, not only the Burst Buffer utilization can be improved, but also HPC applications can achieve high I/O performance by exploiting the powerful Burst Buffer hardware capabilities. Experimental results show that BBOS can improve Burst Buffer utilization by up to 120% while more stable and higher checkpoint performance is guaranteed even under high I/O loads compared to other state-of-the-art schedulers. Besides, our approach can improve the hit ratio of restart requests by up to 96.4% and provides up to 210% higher restart throughput on Burst Buffer.

The vacuum degree of Vacuum Interrupter (VI) will deteriorate because of gases emitted by arc heat, and leakage through the joints. It is important to select an appropriate maintenance criteria for a VI by monitoring the vacuum degree. However, monitoring of a VI installed inside a solid insulation apparatus in a transmission system is impossible. Only studies on the insulation deterioration because of the decrease of the vacuum degree have been conducted, and important factors such as the maintenance criteria have not been investigated. Moreover, most studies have only been performed using AC voltage, not DC voltage. Therefore, in this paper, AC and DC partial discharge were compared by monitoring of the vacuum degree in a distribution class VI to propose maintenance criteria according to voltage type. Through this, for efficient maintenance, maintenance criteria of VI for each voltage type are presented. The vacuum degree at which the dielectric strength rapidly decreases was confirmed according to voltage type. The coupling capacitor was installed directly on the floating shield of VI. Moreover, to measure the pattern and apparent charge of partial discharge according to the vacuum degree, phase resolved partial discharge (PRPD) for AC partial discharge and pulse sequence analysis (PSA) for DC partial discharge were adopted. It is found that maintenance should be performed when the vacuum degree is in the range of $10^{-3} \boldsymbol {\sim } 10^{-2}$ torr regardless of the voltage type. Moreover, the usefulness of the technology is higher at AC voltage than at DC voltage.

Due to the high robustness to artifacts, steady-state visual evoked potential (SSVEP) has been widely applied to construct high-speed brain-computer interfaces (BCIs). Thus far, many spatial filtering methods have been proposed to enhance the target identification performance for SSVEP-based BCIs, and task-related component analysis (TRCA) is among the most effective ones. In this paper, we further extend TRCA and propose a new method called Latency Aligning TRCA (LA-TRCA), which aligns visual latencies on channels to obtain accurate phase information from task-related signals. Based on the SSVEP wave propagation theory, SSVEP spreads from posterior occipital areas over the cortex with a fixed phase velocity. Via estimation of the phase velocity using phase shifts of channels, the visual latencies on different channels can be determined for inter-channel alignment. TRCA is then applied to aligned data epochs for target recognition. For the validation purpose, the classification performance comparison between the proposed LA-TRCA and TRCA-based expansions were performed on two different SSVEP datasets. The experimental results illustrated that the proposed LA-TRCA method outperformed the other TRCA-based expansions, which thus demonstrated the effectiveness of the proposed approach for enhancing the SSVEP detection performance.

Indoor location localization using the time of arrival (TOA) of ultra-wideband (UWB) radio waves faces a challenge in obtaining accurate TOA values in a non-line-of-sight (NLOS) environment owing to obstacles and multipath reception. The accuracy of localization using such sensor data in an NLOS environment is significantly degraded. To address this problem in NLOS environment, we proposed a new localization method that utilizes different sensor combinations for estimation and sensor placement topology. Based on the conventional localization method using various combinations of sensors, the estimation results containing an NLOS ranging value were first removed. Subsequently, we introduced a new parameter called the topological expectancy of localization accuracy (TELA), which represents the effect of sensor placement on the accuracy of localization. The final estimation result was then obtained using the weighted average of the remaining localization results obtained by TELA. Computer simulations and experimental results showed that the performance of the proposed method is superior by 4 to 37% compared with that of conventional methods, even in the presence of many NLOS environment sensors.

Music generation using deep learning has received considerable attention in recent years. Researchers have developed various generative models capable of imitating musical conventions, comprehending the musical corpora, and generating new samples based on the learning outcome. Although the samples generated by these models are persuasive, they often lack musical structure and creativity. For instance, a vanilla end-to-end approach, which deals with all levels of music representation at once, does not offer human-level control and interaction during the learning process, leading to constrained results. Indeed, music creation is a recurrent process that follows some principles by a musician, where various musical features are reused or adapted. On the other hand, a musical piece adheres to a musical style, breaking down into precise concepts of timbre style, performance style, composition style, and the coherency between these aspects. Here, we study and analyze the current advances in music generation using deep learning models through different criteria. We discuss the shortcomings and limitations of these models regarding interactivity and adaptability. Finally, we draw the potential future research direction addressing multi-agent systems and reinforcement learning algorithms to alleviate these shortcomings and limitations.

In recent years, extensive research and applications have demonstrated that a more accurate and stable indoor positioning system (IPS) can be realized through the complementarity of different technologies. However, improving the accuracy of a single positioning technology would significantly impact the subsequent fusion. In this paper, we proposed a smartphone positioning system based on pedestrian dead reckoning (PDR) fusing with a floor estimation module (FEM) and a directional particle filter (PF) extracted by the idea of a regular PF to improve positioning accuracy. Generally, particles are generated by circle-shaped distribution by a regular PF, leading to producing some useless particles that waste computing resources. To tackle this issue, our proposed directional PF uses the particle more effectively by adjusting the distribution shape of particles with an estimated moving direction. Comprehensive experiments are conducted to verify the positioning accuracy and robustness of the proposed method. The experimental data show that the positioning error of our proposed method is only 1.53 m in a multi-floor path with a total length of 306 m.

Objective: A neurodegenerative disease (NDD) detection algorithm using a convolutional neural network (CNN) and wavelet coherence spectrogram of gait synchronization was developed to classify NDD based on gait force signals. The main purpose of this research was to help physicians with screening for NDD for early diagnosis, efficient treatment planning, and monitoring of disease progression. Methods: The NDD detection algorithm was evaluated using the existing online database from Physionet by Hausdorff et al., called gait in neurodegenerative disease database, comprised of windowing, feature transformation, and classification processes. Force pattern variations among healthy control (HC) and patients with ALS, HD, and PD were distinctly observed from feature-extracted wavelet coherence spectrogram images. Results: HC was balanced because their left and right feet supported each other when walking. In patients with ALS, the left-right foot correlation was weaker than that in HC. In patients with HD, walking velocity varied, which indicated that only one foot (right or left) was dominant and sustained the entire body’s balance during movement. The left and right feet of patients with PD were correlated and coordinated in terms of supporting lower-body movements. The right foot was always on the ground to support the entire body when walking. Conclusion: The proposed NDD detection algorithm effectively differentiates gait patterns on the basis of a time-frequency spectrogram of gait force signals between HC and NDD patients with an overall sensitivity of 94.34%, specificity of 96.98%, accuracy of 96.37%, and AUC value of 0.97 using 5-fold cross-validation.

A common belief in designing deep autoencoders (AEs), a type of unsupervised neural network, is that a bottleneck is required to prevent learning the identity function. Learning the identity function renders the AEs useless for anomaly detection. In this work, we challenge this limiting belief and investigate the value of non-bottlenecked AEs. The bottleneck can be removed in two ways: (1) overparameterising the latent layer, and (2) introducing skip connections. However, limited works have reported on the use of one of the ways. For the first time, we carry out extensive experiments covering various combinations of bottleneck removal schemes and datasets using variants of Bayesian AEs. In addition, we propose the infinitely-wide AEs as an extreme example of non-bottlenecked AEs. Their improvement over the baseline implies learning the identity function is not trivial as previously assumed. Moreover, we find that non-bottlenecked architectures (highest AUROC=0.905) can outperform their bottlenecked counterparts (highest AUROC=0.714) on a recent benchmark of CIFAR (inliers) vs SVHN (anomalies), among other tasks, shedding light on the potential of developing non-bottlenecked AEs for improving anomaly detection.

To overcome the problems caused by the limited battery lifetime in multiple-unmanned aerial vehicle (UAV) wireless networks, we propose a hierarchical multi-agent reinforcement learning (RL) framework to maximize the energy efficiency (EE) of UAVs by finding the optimal frequency reuse factor and transmit power. The proposed algorithm consists of distributed inner-loop RL for transmit power control of the UAV terminal (UT) and centralized outer-loop RL for finding the optimal frequency reuse factor. Specifically, the proposed algorithm adjusts these two factors jointly to effectively mitigate intercell interference and reduce undesired transmit power consumption in multi-UAV wireless networks. We show that, for this reason, the proposed algorithm outperforms conventional algorithms, such as a random action algorithm with a fixed frequency reuse factor and a hierarchical multi-agent Q-learning algorithm with binary transmit power controls. Furthermore, even in the environment where UTs are continuously moving based on the mixed mobility model, we show that the proposed algorithm can find the best reward when compared to conventional algorithms.

We investigate the uplink non-orthogonal multiple access (NOMA)-based Internet-of-Things (IoT) networks, where each IoT device (station: STA) equipped with a single antenna exploits the constellation-rotated space-time line code (CR-STLC) to send signals to an access point (AP) equipped with two receive antennas. The AP decodes the signals transmitted from the STAs with the optimal joint maximum-likelihood (JML) detector. As a main result, we mathematically analyze both the bit-error-rate (BER) performance and the spatial diversity order of each STA in the two-user uplink CR-STLC NOMA system. In particular, it is shown that the two-user uplink CR-STLC NOMA system achieves the optimal diversity order regardless of the constellation rotation angle in the high signal-to-noise ratio (SNR) regime. We also consider two different rotation angle optimization techniques (dynamic rotation & fixed rotation) to improve the BER performance in the practical SNR regime, and found that the fixed rotation approach is simple but yields almost the same performance as the dynamic approach. Finally, we show that the uplink CR-STLC NOMA system achieves the spatial diversity order of 1 even when more than two STAs send packets simultaneously. It is worth noting that all STAs obtain an improved spatial diversity gain compared to that without constellation rotation.

A gate driver in array (GIA) design based on the amorphous indium gallium zinc oxide (a-IGZO) thin-film transistor (TFT) is developed for narrow border displays. In the design, each TFT in the gate driver circuits is divided into a certain number of smaller size devices, which can be placed in different subpixels. Therefore, the pixel aperture ratio loss is minimized, and uniform placement of the gate driver circuits over the pixel array area is able to be achieved. The proposed step-like repeating block structure further reduces the occupied area of the signal interconnects. A 12.4-inch fringe field switching (FFS) liquid crystal display (LCD) panel of ultra-narrow border (0.5 mm) is demonstrated with reliable operation based on this GIA design, proving its potential for practical applications.

Air pollution, especially the continual increase in atmospheric particulate matter (PM), is a global environmental challenge. To reduce the PM concentration, a remarkable amount of machine learning-based research has been proposed. However, increasing the accuracy of the predictions and providing clear interpretations of the predictions are challenging. In particular, no studies have addressed models that predict and interpret PM before and during the COVID-19 pandemic. In this paper, we present a two-step predictive and explainable model to obtain insights into reducing PM. We first use attentive multi-task learning to predict the air quality of cities. To accurately predict the concentration of particles with sizes of $\sim 10~\mu \text{m}$ or $\le 2.5~\mu \text{m}$ (PM10 and PM2.5, respectively), we demonstrate a performance difference between single-task and multi-task learning, as well as among the state-of-the art methods. The proposed attentive model with multi-task learning outperformed the others in terms of accuracy performance. We then used Shapley additive explanations, a representative explainable artificial intelligence framework, to interpret and determine the significance of features for predicting PM10 and PM2.5. We demonstrated the superiority of the proposed approach in predicting and explaining both PM10 and PM2.5 concentrations, and observed a statistically significant difference in air pollution before and during the COVID-19 pandemic.

To provide reliable connectivity in recent and future wireless communication systems, it is necessary to deploy several relay nodes. Further, a full-duplex (FD) technique has been in the spotlight since it can significantly improve spectral efficiency, and thus recent studies in relaying networks have considered FD relays. In relaying networks, confidentiality between the source and destination nodes from the relay node should be carefully kept since the relay node cannot be fully trusted, so called untrusted relay node. To this end, in this paper, we consider physical-layer security taking into account an untrusted FD relay node. We investigate a secure relaying protocol against the untrusted relay node where the destination generates artificial noise to prevent the untrusted relay from decoding the source information. We derive the analytical expression of the lower bound of the ergodic secrecy rate ( $\bar {R}$ ). We find two main factors affecting the secrecy performance: residual self-interference (RSI) at the FD-available nodes (i.e., relay and destination), and outdated channel state information (CSI) at the destination. Thereafter, we evaluate their effects on $\bar {R}$ and suggest the algorithm to find the sub-optimal artificial noise power level at the destination for maximizing $\bar {R}$ . Through simulations, we have verified our mathematical derivation and shown that our secure relaying protocol can achieve near-optimal secrecy performance. Numerical results imply that the artificial noise power level should be carefully considered when the channel is severely outdated and RSI is irresistible.

This paper presents a dual-circularly polarized (dual-CP) antenna with simple configuration, low-profile, and wide bandwidth. Its primary radiating element is an aperture-coupled patch loaded with a stacked patch for a wide operational bandwidth. A single-section hybrid coupler is utilized as a feeding structure of the antenna to generate dual-CP radiation. Different from the priors normally requiring both wideband radiator and wideband feeding structure, this work demonstrates that the wideband dual-CP radiation can be obtained by combining a wideband radiator and a conventional hybrid coupler. A design operating at the center frequency of 5.5 GHz has been fabricated and tested. The measurements result in a −10-dB reflection coefficient bandwidth of 30.0% (4.66 – 6.31 GHz), a 10-dB isolation bandwidth of 30.9% (4.66 – 6.35 GHz), a 3-dB axial ratio (AR) bandwidth of 29.1% (4.70 – 6.30 GHz), and a realized broadside gain of 5.5 – 7.05 dBic. For a more robust dual-CP radiation, a low-sidelobe 4 $\times $ 4 element array of the proposed antenna has been designed, fabricated, and measured. Compared to the single element design, the array prototype yields similar impedance and 3-dB AR bandwidths, but a higher isolation. The measurements on the array prototype result in an isolation of $\geq 14$ dB, a peak gain of 16.3 dBic, a side-lobe level of $\leq -20$ dB, and a cross-polarization of $\leq -16$ dB across the frequency range of 4.7 – 6.3 GHz (29.1%).

Due to the dynamic and competitive market environment, it is widely recognized that the development of new products and processes has become the critical point of attention for many companies. The first step in the product development process is to define the nature and function of the product, which is to classify the new product. The traditional product classification process only focuses on the product and market, which is developed by designers on the basis of the products of past dynasties. This way is labor-consuming, inefficient, and has become a bottleneck or constraint for these enterprises to improve their productivities and regulate production. In recent years, Artificial Intelligence has been applied in a wide range of fields including product classification. How to apply machine learning technologies to solve the classification problem of product classification has been widely concerned by researchers. In this paper, a fast and effective product classification, called MdmNet, is proposed, which is based on a novel attempt that embeds the innovation idea of human in machine learning technologies. MdmNet includes three modules: a target customer modeling method based on the deep learning technologies a consumer information deduction method based on the MDM that builds a consumer feature closed loop and output the classification result of the consumers’ perspective, and a weighted fusion module. Experiments conducted on benchmark datasets Cars demonstrate the impressive performance of the proposed MdmNet. This paper first attempts to add consumer motivation analysis to traditional machine learning method, which has a strong application prospect.

This study firstly proposes and experimentally demonstrates a phase demodulation method with high sensitivity for interferometric fiber sensors (IFSs) on the basis of the fast Fourier transform of a wavenumber domain. The phase information of IFSs triggered by changes in environmental parameters is obtained by calculating the initial phase variation of a specific Fourier-transformed spatial frequency. Theoretically, phase sensitivity can be improved by n times when the optical path difference (OPD) of a spatial frequency peak is increased by a multiple (n times) of the OPD of other spatial frequency domain peaks. To verify the method experimentally, this study designed a large laterally offset spliced sensor formed by common single-mode fibers on the basis of mode interference. The sensing characteristics of temperature and strain in each set of two-beam interference are analyzed simultaneously by calculating the phase sensitivities of the frequency domain peaks. Furthermore, the highest reported temperature and strain sensitivities of 0.0795 rad/°C and −0.0088 rad/μϵ, respectively, with good repeatability are realized.

A portion of the data in video captioning datasets are noisy and unsuitable for models to learn at early stages, e.g., there could be a generic 4-word-long caption lacking distinctive details of video content and a 19-word-long description with rare words and complex structure given the average caption length of around 9. The conventional training method, i.e., learning by random sampling indiscriminately from the whole training set, may cause data bias problems and undermine the model performance. In this work, we present a novel learning strategy, Adaptive Curriculum Learning (ACL), to alleviate the adverse effect of such problems. The main idea of our approach is to allow a model to learn from the data within its competence. Specifically, a difficulty measurement is first defined to evaluate the learning difficulty of video-caption pairs, and training data can be ranked accordingly. Then, based on the learning difficulties and model competence, an adaptive sampling approach is developed to provide suitable training subsets for video captioning models in different training stages. Notably, our proposed ACL is applicable to most existing video captioning works as it requires no modifications of the model architecture. Extensive experiments are conducted on mainstream benchmarks, i.e., MSVD and MSR-VTT datasets. The results show that both RNN-based and Transformer-based models achieve consistent performance improvements with our ACL strategy.