Search

Resume information extraction technologies are crucial for automating the shortlisting and evaluation of resumes, benefiting both enterprises and job seekers. Resume block classification plays a pivotal role in the process of resume information extraction, as it significantly impacts the accuracy of the subsequent intra-block information extraction. However, as reported in the existing research, the classification of resume blocks often uses traditional and general-purpose text classification models, failing to consider the intrinsic contextual order relationships amongst resume blocks. Moreover, there are few studies that consider the transferability of classification models. Therefore, to address these limitations, we propose a series of methods to enhance the performance of resume block classification. Our approach focuses on three key aspects. Firstly, we introduce a novel sequence encoder that effectively extracts the sequence features among blocks of the same resume, leveraging the intrinsic contextual order relationships to improve classification accuracy. Secondly, considering the large classification variance of the existing models in different scenarios, we enhance the sequence encoder with a feature fusion strategy, which combines multiple features to improve the model’s robustness and transferability. Thirdly, from the perspective of ensemble learning, we propose a dynamic weighted hybrid model that dynamically generates weighting for each participating sub-model, enabling adaptive integration of different classification models. Finally, to alleviate the huge workload of cross-domain relabeling, we develop a transfer learning model specifically designed for the resume block classification task, facilitating the application of our approach across different domains. To evaluate the effectiveness of our proposed methods, we release three Chinese datasets that include 4,500 Chinese resumes on https://github.com/xqqhelloword/resume-block-classification. Experimental results show that our hybrid model achieves 97.6% accuracy, outperforming existing methods and establishing a new state-of-the-art in this field.

To assess various segmentation methods for the anterior and posterior chambers in ultrasonic biomicroscopy (UBM) images. UBM images were collected from 102 patients diagnosed with primary angle-closure glaucoma, and the corresponding intraocular pressure (IOP) values were measured. The UBM images are manually segmented using ImageJ software as the gold standard, while automatic segmentation employs six distinct methods: Otsu threshold, K-means, fuzzy C-means, robust self-sparse fuzzy clustering algorithm, spatial intuitionistic fuzzy C-means algorithm, and fast robust fuzzy C-means (FRFCM) algorithm. The segmentation results were used to quantify the anterior chamber depth and the area, perimeter, and height of the posterior chamber. Quantitative analyses of the accuracy and reliability of the segmentation quantification results were conducted using relative error and intraclass correlation coefficients (ICC). A total of 408 clear enough UBM images were used for segmentation. The ICC values of the quantitative results of the FRFCM method are 0.996, 0.970, 0.986, and 0.970, outperforming the other five segmentation methods. It excels in accuracy, reliability, precision comparable to manual techniques, and surpasses them in reproducibility and efficiency. Thus, it can be used for semi-automatic analysis of both anterior and posterior chamber regions in UBM images. Additionally, Spearman correlation analysis was performed to explore the relationship between IOPs and the four quantification results. The Spearman correlation coefficients between IOP and anterior chamber depth, posterior chamber area, posterior chamber perimeter, and posterior chamber height were -0.0991, -0.0043, -0.0321, and 0.0757, respectively. There is no significant correlation between IOP and anterior chamber depth, posterior chamber area, perimeter, or height.

The output fluctuation of the high proportion of photovoltaic new energy requires introducing energy storage units for compensation and adjustment, but the voltage stability performance of energy storage port converters under complex working conditions is often not effectively guaranteed. Therefore, this paper proposes an active disturbance rejection voltage stabilization strategy considering observation state correction for energy storage port converters. By deriving the high-precision paradigm of observation deviation, the important structural factors of observation deviation in the system are pointed out, and the ignored high-order terms of deviation are compensated into the observer to achieve accurate and rapid convergence of tracking deviation. Furthermore, through the theoretical analysis of the strategy in four dimensions of stability performance, tracking performance, anti-disturbance performance, and robust performance, the reasons why the strategy can effectively improve the anti-disturbance and robustness of the converter are explained. Finally, the simulation and experimental results jointly demonstrate that the strategy performs better anti-disturbance and robust-ness performance than traditional active disturbance rejection control under various complex working conditions of microgrids, which is of great significance for improving the stability of microgrid operation.

Real-time RGB-D semantic segmentation is crucial for tasks like dynamic environment analysis of mobile robots or autonomous vehicles, real-time and efficient segmentation results can enhance the accuracy of subsequent tasks such as free space detection, mapping and navigation. It requires models to balance computational cost and performance by employing more efficient mechanisms to effectively recognize differences in RGB-D multimodal information, retain complementary information, and reduce redundancies. Currently, although convolutional neural network (CNN) methods are less accurate than Transformer-based methods, they offer stronger real-time performance under the same computational load. Therefore, in this study, we proposed the Efficient Feature Interaction Network (EFINet), a real-time RGB-D segmentation method that uses a lightweight CNN encoder and incorporates encoder blocks with a lightweight upsampling method Dysample and the carefully optimized number of ConvNeXt V2 blocks, to redesign the decoder and minimize redundant computations. Additionally, we propose a robust cross-modal interaction mechanism that facilitates the exchange of useful information and the elimination of noise between modalities with minimal computational overhead. By incorporating lightweight convolutions, the proposed method can acquire multi-scale global information with minimal additional computation, thus providing the model with rich contextual information. The proposed EFINet has 38.0 M parameters and 40.0 GFlops, achieving 55.5% and 50.6% mIoU on the challenging NYUDv2 and SUN RGB-D datasets, respectively, and real-time inference at 31.2 FPS. EFINet outperforms several state-of-the-art methods in the real-time RGB-D segmentation task while achieving a better balance between segmentation performance and computational efficiency.

This paper introduces a novel neural network adaptive fractional-order sliding mode control strategy based on the minimal learning parameter method (MLPNN-FOSMC). This method aims to solve the problem of relative position control in satellite formation flying (SFF), especially in the presence of model uncertainties, external disturbances, input saturation, and unknown actuator gains. First, the mathematical model for satellite relative positioning is derived. Then, a fractional-order sliding mode controller is introduced and integrated with a radial basis function (RBF) neural network and the minimal learning parameter (MLP) strategy to compensate for errors caused by model uncertainties and external disturbances. At the same time, adaptive control is employed to mitigate the impact of unknown actuator gains on the system. To address the non-smooth input saturation nonlinearity problem, a new saturation function with the smoothing properties of the hyperbolic tangent function is introduced. The system’s stability is ensured by employing the Lyapunov theorem. Finally, a comparative analysis with traditional sliding mode control (SMC) and neural network sliding mode control based on minimal learning parameter (MLPNN-SMC) highlights the superiority of the proposed method.

Predicting the wideband radiation of the high-power interconnection cable network within the system is necessary for the electromagnetic compatibility design verification of power electronic systems, and it is urgent to accurately model and analyze its radiation characteristics. The current modeling methods, such as the equivalent dipole method, have numerous constraints, making it difficult to effectively analyze multi-core cable network with connectors, transmission lines of arbitrary arrangement, and large-scale branches. To address this problem, this paper proposes a hybrid modeling method, which integrates multi-conductor transmission line (MTL) equations, dipole radiation theory, and multi-port network theory. Firstly, by using multi-port network theory, this method creates an equivalent circuit for the connection by extracting the S-parameters of the connector. Then, a cascaded transmission model is established using MTL equations, which can analyze the radiation characteristics of various interconnecting cable networks between electromechanical devices, including both straight and bent configurations. Next, the current distribution along the cable network is solved through termination conditions. Furthermore, dipole array radiation theory is used to calculate the radiated field of the complicated cable network with connectors, large-scale branches, and transmission lines of arbitrary layout. Finally, some examples are used to validate the validity and accuracy of this method.

Human man Activity Recognition (HAR) using wearable sensors, particularly wrist-worn devices, has garnered significant research interest. However, challenges such as sensor displacement and variations in wearing habits can affect the accuracy of HAR systems. Two compensation strategies for sensor displacemnt are proposed to address these issues. The first strategy is hybrid data fusion, which involves merging sensor data collected from different displacement locations on the wrist. This technique aims to mitigate the discrepancies in data distribution that result from the multiple wearing positions along the wrist, thereby enhancing the overall accuracy of HAR models. The second strategy is cross-location transfer fine-tuning, which involves pretraining a model with data from typical wrist locations and then fine-tuning it with data from a new sensor location. This approach improves the model’s ability to adapt and perform accurately when the sensor is placed in a different position, significantly enhancing its performance and generalization capabilities. To verify the effectiveness of these proposed compensation strategies, we built an LSTM baseline model and introduce a new Multi-stage Feature Extraction (MSFE) model that integrates 1D CNN and attention. Experiments on common activities such as walking, standing, using stairs, and lying down, with data recorded at multiple locations along the wrist, have shown that both hybrid data fusion and cross-location transfer fine-tuning strategies notably improve the recognition accuracy of HAR models. The proposed MSFE model achieves higher recognition accuracies than the LSTM model in all six experimental scenarios, particularly in Scenario 5 involving sensor displacement, with an improvement of up to 31.65%. Additionally, thecross-location transfer fine-tuning strategy enhances the recognition accuracy by 9.19% for Subject 3 with sensor displacement at the right wrist location. These advancements in handling sensor displacement and wearing variations are crucial for developing more reliable and versatile wearable technologies.INDEX TERMS Sensor displacement compensation, human activity recognition, deep learning, wrist.

In this paper, a calculation of restraining flow separation of the high-speed UAV airfoil with dielectric barrier discharge (DBD) plasma is studied. Based on the simplified model of the DBD plasma, the effect of plasma on neutral gas is considered as a body force, which is introduced into the Navier-Stokes (N-S) equation as a source term by using the user-defined functions interface in computational fluid dynamics (CFD) software. In this way, the distribution of the flow field of the plasma can be obtained, allowing for the analysis of the mechanism of suppression by the plasma in flow separation. The lift coefficient and the drag coefficient were chosen as two characterizations, then the situations of the plasma actuator in different positions and the airfoil at different angles of attack were quantitatively analyzed. The digital simulation test results indicate that for the high-speed airfoil selected in this paper, when the actuator is at 0.6C (C is the chord length) at the trailing edge in high-speed flight, the effect of the actuator will increase the lift-drag ratio of the high-speed airfoil, and the smaller the angle of attack is, the greater the increment will be, with the maximum increment being about 2.28%.

This work presents an amplitude extraction method based on eigenvalue decomposition (EVD), which can substitute the role of phase-sensitive demodulation in electrical impedance tomography (EIT) demodulation systems. An m $\times $ m square signal matrix is initially constructed according to a series of samples and an m $\times $ n Hankel matrix. Then, the EVD operation is performed on the square signal matrix, and as a result, its eigenvalues are sequentially obtained. Finally, a mathematical relation among the sinusoid amplitude, effective eigenvalues and the number of rows of the constructed signal matrix is provided. Compared to traditional methods, the proposed method offers the following merits: 1) since no reference signal is required, there is no need to consider synchronization between the received and reference signals; 2) it can extract the signal amplitude with high signal-to-noise ratio (SNR), even in the case of non-integer periodic sampling; 3) EVD itself can achieve a filtering function, thereby additional low-pass filters are unnecessary for the signal conditioning module. The feasibility and accuracy of the proposed method are verified by simulations and experimental results, demonstrating that it can provide the amplitude information with excellent robustness and SNR performances.

In this paper, we investigated an experimental analysis of the degradation caused by low dose rate irradiation in GaN-based high-electron-mobility transistors (HEMTs) with a p-type gate. Combined with experimental frequency-dependent conductance ( $\text{G}_{\mathrm {p}}/\omega)$ analyses and TCAD simulations, it has been demonstrated that the negative shifts in both $\text{I}_{\mathrm {d}}$ - $\text{V}_{\mathrm {gs}}$ and $\text{V}_{\mathrm {th}}$ are primarily due to the formation of donor-like traps near the p-GaN/AlGaN interface, which is a result of the dehydrogenation of pre-existing defects during low dose rate irradiation. Additionally, the results of the TCAD simulations, indicate that the trap-assisted tunneling (TAT) process, which involves the recombination of trap-assisted holes with electrons in the p-GaN layer, may dominate the physical mechanisms responsible for the increase in gate leakage current ( $\text{I}_{\mathrm {g}}$ - $\text{V}_{\mathrm {gs}}$ ). These results may provide a basis for understanding the role of radiation-induced traps on electrical parameters degradations for p-GaN gate HEMTs.

Deep learning models have demonstrated their effectiveness in capturing complex relationships between input features and target outputs across many different application domains. These models, however, often come with considerable memory and computational demands, posing challenges for deployment on resource-constrained edge devices. Knowledge distillation is a prominent technique for transferring the expertise from an advanced yet heavy teacher model to a more efficient leaner student model. As ensemble methods have exhibited notable enhancements in model generalization and have achieved state-of-the-art performance in various machine learning tasks, we adopt ensemble techniques to perform knowledge distillation from BERT using multiple lightweight student models. Our approach applies lean architectural paradigms of spatial and sequential networks including LSTM, CNN and their fusion to perform data processing from distinct perspectives. Instead of using contextual word representations which require more space in natural language processing applications, we take advantage of a single static pre-trained and low-dimensional word embedding space to be shared among student models. Empirical studies are conducted on the sentiment classification problem and our model outperforms not only other existing techniques but also the teacher model.

Text-conditioned outfit recommendation aims to recommend a whole fashion outfit that satisfies the compatibility between the recommended items and given items and adheres to the text condition like “Paradise Tropical Vacation” or “60s Style”. Using text description as a condition can provide users with a flexible and accurate way to retrieve and recommend fashion items but this problem is underexplored by existing studies. A challenge of text-conditioned outfit recommendation is how to encode and fuse the outfit text description and fashion item images and text. To solve this, this paper proposes a framework for this task which features a hybrid attention layer that constructs the relationship between outfit text description and fashion items for condition compliance, and the relationship between fashion items for internal compatibility. To encode fashion item features, our method uses pre-trained FashionCLIP as an extractor which significantly reduces the trainable parameters compared to previous methods training CNN from scratch. The whole outfits are generated by iteratively adding compatible items based on a given partial outfit. Compared with state-of-the-art methods on polyvore disjoint and non-disjoint datasets, our approach can achieve 3% relative improvement in compatibility prediction AUC, achieve 5% relative improvement in fill-in-the-blank accuracy; achieve 19% relative improvement on complementary item retrieval recall at different ranks in average. Besides, We demonstrate that our approach can recommend a whole outfit with inner compatibility and adhere to the text description.

The applications of multiple underwater robots are becoming more widespread as the oceans continue to be explored. However, the cooperative control of multiple underwater robotic systems lacks an effective way to interact with information in the presence of weak communication in the water. This study proposes a brief binary optical-acoustic dual signals hybrid cooperative control method for overcoming this difficulty. It simplifies the information interaction rule to improve the efficiency of cooperative control while guaranteeing the range of underwater communication. This method includes an acoustic communication mechanism that enables the underwater robot to transmit signals over long distances and an optical communication mechanism that enables the transmission of signals in close situations. Avoiding information redundancy due to traditional underwater real-time communication, the underwater robots send brief binary signals through optical and acoustic sensors to realize cooperative control. An underwater robot equipped with a dual optical-acoustic system is designed in this paper combining the leader and follower formation structure. The triangular formation experiment and diamond formation experiment were also conducted in the pool to verify the effectiveness of this cooperative control method. The study provides a method for the cooperative control of multiple underwater robotic systems.

Since the original signature model did not account for the possible threat of the forging signature attack, the majority of existing multivariate polynomial public key based signature schemes are at risk of forging signatures by equivalent key attacks. In this research, an enhanced signature model, Triple Unlocking Mechanism (TUM), is provided to resist all the possibility that the signature generated by the equivalent keys. The additional public keys are generated in the sake of enhancing the security by verifying internal unit information in the scheme after three designed unlocking operations. As a result, the signature can only be generated by the user who has the real legal key and the threat of the key recovery attack can be eliminated. The security of TUM signature based on random oracle model is proved. Experimental results demonstrate that, using an MI system coupled with TUM as an example, the advantages of the enhanced signature model are more secure than the original one at the expense of taking a little more time for signing. Moreover, influences of different parameters in multivariate polynomial public key signatures are investigated. Based on the original state-of-art model, the presented model is generic construction and applicable for existing multivariate signature scheme’s construction.

Learning-based image registration approaches typically learn to map from input images to a transformation matrix. Regarding the current deep-learning-based image rigid registration approaches learn a transformation matrix in a one-shot way. Our purpose is to present a deep reinforcement learning (DRL) based method for image registration to explicitly model the step-wise nature of the human registration process. We cast an image registration process as a Markov Decision Process (MDP) where actions are defined as global image adjustment operations. Then we train our proxy to learn the optimal action sequences to achieve a good registration. More specifically, we propose a DRL proxy incorporating an attention mechanism to address the challenge of large differences in appearance between images from different modalities. Registration experiments on 3D CT-MR image pairs of patients with nasopharyngeal carcinoma and on publicly available 3D PET-MR image pairs show that our approach significantly outperforms other methods, and achieves state-of-the-art performance in multi-m-modal medical image registration.

Transient stability emergency control (TSEC) enhances power system transient stability during large disturbances but faces challenges in the high-penetration wind power grid where the wind power forecast error still cannot be ignored even with state-of-the-art forecasting methods. In this paper, the TSEC problem is modeled as robust nonlinear programming with the objective of maintaining rotor angle stability by generator tripping and load shedding while the uncertain wind power outputs are regarded as intervals. Interval programming is employed to solve the robust TSEC problem where the trajectory sensitivity analysis is applied to approximate the bounds of transient stability constraints. Numerical results on two test systems demonstrate that the proposed method improves computational efficiency and shows good performance on robustness.

Weakly supervised semantic segmentation (WSSS) generally utilizes the Class Activation Map (CAM) to synthesize pseudo-labels. However, the current methods of obtaining CAM focus on salient features of a specific layer, resulting in highlighting the most discriminative regions and further leading to rough segmentation results for WSSS. In this paper, we rethink the potential of the ordinary classifier and find that if features of all the layers are applied, the classifier will obtain CAM with complete discriminative regions. Inspired by this, we propose Fully-CAM for WSSS, which can fully exploit the potential of the ordinary classifier and yield more accurate segmentation results. Precisely, Fully-CAM firstly weights feature with their corresponding gradients to yield CAMs of each layer, then fusing these layers’ CAMs could generate an ultimate CAM with complete discriminative regions. Furthermore, Fully-CAM is encapsulated into a plug-in, which can be mounted on any trained ordinary classifier with convolution layer, and it exceeds its previous performance without extra training.

A multimodal biometric authentication framework based on Index-of-Max (IoM) hashing, Alignment-Free Hashing (AFH), and feature-level fusion is proposed in this paper. This framework enjoys three major merits: 1) Biometric templates are secured by biometric template protection technology (i.e., IoM hashing), thus providing strong resistance to security and privacy invasion; 2) It flexibly adopts a variety of biometric feature representations (e.g., binary, and real-valued), thus generalizing to a wide range of biometric features for fusion; 3) Feature-level fusion, which has low template storage, low matching computational complexity, and low privacy risks, which can be accomplished without alignment via AFH. Specifically, the proposed framework works as a drag-and-drop model that can quickly adopt all popular biometric modalities with different feature distributions for feature-level fusion. The fused templates are produced using operators AND, OR and XOR in binary domain. To evaluate the proposed framework, benchmarking datasets from four widely deployed biometric modalities (i.e., FVC 2002 for fingerprint, LFW for face, CASIA-v3-Interval for iris, and UTFVP for finger-vein) are used. The experimental results presented in Table 5 suggest that the proposed framework can achieve state-of-the-art performance in most of the datasets while offering additional folds, such as template protection and generalization to variable features. Moreover, biometric template protection criteria (irreversibility, unlinkability, and revocability) are also analyzed. The results of the analysis indicate satisfaction in terms of the security and privacy of the templates generated from the proposed framework.

Knowledge synchronization in robot swarm systems is a challenging task under communication constraints. Swarm robots cannot maintain the complete communication structure with classic fixed communication network, which leads to the decline of motion decision effects. It is expected to improve the performance of swarm robot motion decisions under communication constraints with more researches on dynamic complex networks. This study established a feedback search control model for robot swarm systems for target search tasks in a 3D environment. Inspired by the WS model, which showed good synchronization performance in complex systems owing to its small average shortest path length, a novel dynamic small-world network model called decay small-world was proposed for negative impact of communication constraints. Decay small-world model reconnects robot communication links by controlling the rewired probability based on the half-life formula. It realizes dynamic network topology by decentralized computing. This new model maintains a small-world pattern over time without degenerating it into a random network. Simulations show that good knowledge synchronization of swarm robots can be realized using the decay small-world model. And the results also show that target search performances are promoted by this method.

Generative Adversarial Network (GAN) based techniques can generate and synthesize realistic faces that cause profound social concerns and security problems. Existing methods for detecting GAN-generated faces can perform well on limited public datasets. However, images from existing datasets do not represent real-world scenarios well enough in terms of view variations and data distributions, where real faces largely outnumber synthetic ones. The state-of-the-art methods do not generalize well in real-world problems and lack the interpretability of detection results. Performance of existing GAN-face detection models degrades accordingly when facing data imbalance issues. To address these shortcomings, we propose a robust, attentive, end-to-end framework that spots GAN-generated faces by analyzing eye inconsistencies. Our model automatically learns to identify inconsistent eye components by localizing and comparing artifacts between eyes. After the iris regions are extracted by Mask-RCNN, we design a Residual Attention Network (RAN) to examine the consistency between the corneal specular highlights of the two eyes. Our method can effectively learn from imbalanced data using a joint loss function combining the traditional cross-entropy loss with a relaxation of the ROC-AUC loss via Wilcoxon-Mann-Whitney (WMW) statistics. Comprehensive evaluations on a newly created FFHQ-GAN dataset in both balanced and imbalanced scenarios demonstrate the superiority of our method.

Active flow control (AFC) based on plasma synthetic jet actuator (PSJA) has been attracted a lot of researchers since Anderson and Knight put forward the concept of PSJA for the first time. Since then the characteristic of PSJA has been studied by means of many numerical simulations and experiments. This article showed a series of studies that manipulated a vehicle without moving control surfaces. Firstly, we research the characteristics of the aviation actuator, including the effects of structure and parameters on performance. Secondly, we explore the influence of the actuator on the air field around the aircraft by simulation, including the influence of the actuator on the aerodynamic coefficients and pitching moment coefficient. Finally, since these changes are similar to those caused by the deflection of the rudder surface, we can equivalent the effect of the PSJA to the deflection of the rudder surface, and simulate the three-degree-of-freedom pitch control channel of the aircraft to illustrate the effectiveness of PSJA on the flapless control system.

The fifth-generation (5G) communications enables various promising applications that was once impossible, e.g. remote healthcare with the help of fast and reliably delivery of medical data. Post-partum hemorrhage (PPH) refers to the massive blood loss after the birthing stage (within 24 hours), i.e. >500ml for the vaginal delivery, and >1000ml for the cesarean section. PPH is by far the most common cause of the mortality rate of pregnant women, as well as a primary cause of current pregnant mortality in China. Despite the great potential of prediction of PPH, there is currently no effective tool based on the limited raw data from the clinical trials. In the study, we retrospectively study the 3842 vaginal delivery cases in 2017 collected from Beijing Obstetrics and Gynecology Hospital, Capital Medical University. In particular, we obtain the prediction based diagnostic model relying on machine learning, and we adopt the ensemble learning to accomplish this task, by combining the results of various candidate methods. According to the experimental results, the accuracy of correct PPH diagnosis would approach 96.7%; the total disseminated intravascular coagulation (DIC) prediction accuracy approaches 90.3%. In this regard, we may conclude the proposed model based on machine learning would allow us to predict successfully the risk of PPH, and assess the critical level of PPH patient. We anticipate our study results would contribute to the reduction the mortality of pregnant women.

The core problem in unmanned/intelligent working face of coal mining is the automatic adjustment of shearer arm where the coal-rock interface detection is the key. The cutting location of shearer drum affect the proportion of coal and rock powder around the cutting teeth of shearer drum. Therefore, the method of on-line coal rock interface characterization using Terahertz Time Domain spectroscopy (THz-TDs) we proposed aims to estimate the ratio of rock by Terahertz response. Firstly, anthracite and quartz sandstone were uniformly mixed according to 39 different ratios in this study, the samples' responses were obtained by terahertz system, and then the obtained time domain data was converted into frequency domain data by fast Fourier transform. The absorption coefficient spectrum and the refractive index profile of the 39 samples were calculated by optical parametric model. Secondly, corresponding quantitative model between mixed coal/rock powder and THz signal was built by using back propagation neural network (BPNN) and least squares support vector machine (LSSVM). We expected to use the ratio of rock powder detected by the model to estimate the depth of shearer drum teeth embedded in the rock layer. Finally, we found that both two mathematical arithmetic is feasible to quantitatively detect different proportion of coal and rock mixtures. The results show that the depth of shearer drum teeth embedded in the rock layer could be estimated by the novel method, which means the coal-rock interface could be on-line characterized by using THz-TDs and the height of the drum could be adjusted in time.

This paper studies the problem of H filtering for nonlinear Markov jump systems based on Takagi-Sugeno model. Firstly, we propose a hybrid event-triggered mechanism with an adjustable threshold, which not only helps to save more limited communication resources, but also excludes Zeno behavior while preserving the merits of continuous triggering. Secondly, given the threat of cyber-attacks to network security, a stochastic variable is introduced to describe the considered deception attacks in filter design. Thirdly, a less restrictive Lyapunov-Krasovskii functional (LKF), which is not required to be continuous and positive definite in a triggering interval, is constructed to establish sufficient condition on the exponential mean-square stability for the filtering error system with a weighted H∞ performance. Meanwhile, co-design of the desired filter and event-triggered mechanism is achieved. Finally, a tunnel diode circuit system is provided to illustrate the effectiveness and advantage of the obtained results.

Mission-critical sensors and sensor networks (MC-SSN) have been applied to missions such as battlefield, border patrol, search and rescue, critical structure monitoring, and surveillance. To support critical missions, sensors and sensor networks need to be flexible and interactive and continuously work despite limited bandwidth, intermittent connectivity, and with a large number of devices on the network. Sometimes, humans will be the elements within mission-critical sensors and sensor networks that are most vulnerable to deception, and humans will be handicapped when they are concerned about the received information from the network is not trustworthy, even if their concern is misplaced. In MCSSN, the advantages of linking multiple electronic support measures and electronic attack assets to achieve improved capabilities across a networked mission-critical force have yet to be quantified. Algorithms are sought for fused and/or coherent cross-platform radio frequency (RF) sensing. The MC-SSN algorithms should be capable of utilizing RF returns from multiple aspects in time-coordinated sensors and sensor networks. Such adaptation, management, and reorganization of information sources, devices, and networks must be accomplished almost autonomously in order to avoid imposing additional burdens on the humans, and without much reliance on support and maintenance services. Moreover, humans, under extreme cognitive and physical stress, will be strongly challenged by the massive complexity of the MC-SSN and the information it will provide and carry. Advances in technologies that capitalize on the benefits of the MC-SSN have to assist humans in making effective use of this massive, complex, confusing, and potentially deceptive ocean of information while taking into account the ever-changing mission. New approaches and low-complexity algorithms are expected to enable MC-SSN to automatically manage and effect risk and uncertainty in a highly deceptive, mixed cooperative/adversarial, information-centric environment.

Communication systems deployed in harsh environments, such as high-speed trains, subways, deserts, trenches, forests, and underground mines, should be specially constructed to withstand extreme conditions such as high or low temperatures, corrosive humidity, extreme weather, or excessive dust and dirt. Such applications require specially designed wireless communication, fiber-optic communications, satellite communication, or signal processing techniques that can perform under extreme conditions and meet the QoS, security, and reliability requirements. Quite often, communications in harsh environments have an extremely low signal-to-noise ratio (SNR), high Doppler shift, and long latency, and often consume more power and energy. Bandwidth limitation in harsh environments requires spectrum efficient communications. The current state of art technologies such as massive MIMO, advanced modulation and channel coding, artificial intelligence, and signal processing provide different venues to explore this challenging area.

Efficient decoding for general linear channel codes has been a long standing problem. Message passing type algorithms can achieve near-optimal performance for channel codes with sparse check matrices and long code-lengths, but they may suffer from obvious performance loss when the aforementioned two conditions are not satisfied. In this paper, we propose an alternative hypergraph filtering based decoding scheme based on recently developed hypergraph signal processing (HGSP) frameworks. Specifically, we model the coded sequence as a hypergraph signal by regarding each coded bit as a vertex and representing each check constraint by a hyperedge. We then map the message passing in iterative decoding algorithms to the signal shifting in HGSP, based on which a hypergraph filtering (HF) based decoding scheme is developed. By further combining the modified random redundant iterative decoding (mRRD) technique, we finally develop a novel HF-mRRD scheme to approach the near-optimal decoding performance. Simulation results on both Hamming codes and BCH codes show that the proposed scheme obtains significant performance gains over standard message passing type algorithms, and achieve comparable or even better performance than the state-of-the-art deep learning based schemes.

This paper proposes a non-weighted maximal approach of multi-attribute risk assessment for production accidents, which comes from the Chinese practice of risk management rather than the theoretical weighted multi-attribute approach. The existing literature for risk assessment of pipeline accidents, there is an absence of or lack of explicit consideration of some special dimensions, i.e., environmental pollution as the important derivative disaster. The non-weighted maximal approach is described the maximum function among multiple criteria, which include fatalities, serious injuries, direct economic loss, and environment pollutions. The approach comes from the Chinese government official achievement assessment system with the characteristics of “one ticket veto system for production safety”, and has applied to ex ante assessing likelihood of the accident, and ex post holding the responsible for accidents. At last, applying the case of the Chinese Qingdao oil pipeline accident, the maximal approach is compared with the FN curve criterion, the ALARP principle and the ELECTRE TRI method. The results show that the maximal approach of production safety accident criterion pays more attention to the risk density or risk consequences, which follows the “cask principle” and is much more useful controlling the risk when targeting the vulnerable links of engineering systems.

The rise of model-based and machine learning methods have created increasingly realistic opportunities to implement personalized, patient-specific mechanical ventilation (MV) in the ICU. These methods require monitoring of real-time patient ventilation waveform data (VWD) during MV treatment. However, there are relatively few non-invasive and/or non-proprietary systems to monitor and record patient-specific lung condition in real-time. In this paper, we present a CARE network data acquisition and monitoring system (CARENet) to automate data collection and to remotely monitor patient-specific lung condition and ventilation parameters. The automated system acquires VWD from a mechanical ventilator using a data acquisition device (DAQ), stores data in network-attached storage (NAS), and processes VWDs via a data management platform (DMP) web application. The web application enables real-time and retrospective model-based monitoring and analysis of clinical MV data. In particular, CARENet provides detailed breath-by-breath patient-specific respiratory mechanics, as well as the overall trends assessing changes in patient condition. Validation testing with a retrospective data set illustrated how breath-to-breath and time-varying patient-ventilator interaction during MV can be monitored, and, in turn, can be used to guide MV treatment. The network data acquisition system design presented is low-cost, open, and enables continuous, automated, scalable, real-time collection and analysis of waveform data, to help improve decision making, care, and outcomes in MV.

This paper investigates the distributed resilient consensus control problem for a class non-linear heterogeneous multiagent systems under unreliable communication environment, which includes time-varying interaction network and intermittent denial of service (DoS) attacks. In order to handle heterogeneous nonlinear unknown dynamics in each agent subsystem, a novel distributed adaptive neural networks (NNs) control approach is developed exploiting the neighborhood information. By introducing the topology-dependent Lyapunov function and adaptive switching control mechanism, we show that the uniformly bound convergence of the consensus errors can be guaranteed by the proposed topology average dwell time constraints. A simulation example is provided to validate the effectiveness of the proposed resilient control algorithm.

Bad data may lead to performance degradation or even instability of a power system, which can be caused by various factors: unintentional PMU abnormalities, topology error, malicious cyber-attacks, electromagnetic interference, temporary loss of communication links, external disturbances, extraneous noise biases, etc. In order to develop a more resilient and reliable state estimation technique, this manuscript presents a novel two-step fault tolerant extended Kalman filter framework for discrete-time stochastic power systems, under bad data, PMU failures, external disturbances, extraneous noise, and bounded observer-gain perturbation conditions. The failure mechanisms of multiple phasor measurement units are assumed to be independent of each other with various bad data or malfunction rates. The benchmark IEEE standard test systems are utilized as a demonstrative example to carry out computer simulation studies and to examine different estimation algorithms. Experimental results demonstrates that the proposed second-order fault tolerant extended Kalman filter provides more accurate estimation results, in comparison with traditional first- and second-order extended Kalman filter, and the unscented Kalman filter. The proposed two-step fault-tolerant extended Kalman filter can serve as a powerful alternative to the existing dynamic power system state estimation techniques.

The insulation of high-frequency transformers (HFTs) has a significant impact on the safety and reliability of high voltage power electronic transformers (PETs). The transient voltages from the rapid turn-off and turn-on of the high voltage power semiconducting device increase the insulation stress and loss, resulting in thermal stress distortion. However, few studies have presented an accurate evaluation of the dielectric loss, especially under high dV/dt square waves. This paper focuses on the dielectric loss calculation in HFT insulation by analyzing the loss from a high-frequency square wave with spike voltages. A step function that is equivalent to spike voltage superposition is proposed for the quantitative calculation of the dielectric loss of epoxy resin. A cutoff frequency ( $f_{\mathrm {c}}$ ) is defined to indicate high dV/dt characteristics. A similar narrow step function is used for simulating spike voltage at the rising/falling edge. The additional dielectric loss increases with the spike voltage and the duration. The electric field and temperature distributions are simulated in a 10 kW, 10 kHz HFT by considering the calculation of dielectric loss. The spike voltages enhance the electric field and temperature in the insulation. The increase in the external temperature (close to the glass transition temperature ( $T_{\mathrm {g}}$ ) of epoxy resin) causes a significant increase in the insulation temperature. This paper is of great importance to the design and application of HFT insulation.

With the development of wind turbines to the large scale with high rated powers, how to reduce fatigue loads is the key factor for modern wind turbine control. Aiming at reducing the wind turbine structural fatigue loads, an individual pitch control (IPC) method based on SVM estimation of blade root loads was proposed in this paper. The blade root loads were dynamically estimated by a SVM model according to light detection and ranging (LIDAR) wind speed data. Two approximate decoupling systems for yaw moments and pitch moments were then obtained through Coleman transform. Meanwhile, two individual pitch controllers were designed to reduce the loads and a wind speed feedforward controller was adopted to improve the pitch control performance. The proposed method was verified by a FAST, LIDAR and RT-LAB semi-physical real-time combined simulation system. On this basis, a master control unit was developed and the overall unit was further verified by a CRRC wind turbine master control and pitch joint test platform. Compared with collective pitch control (CPC) and the IPC based on azimuth, the proposed method is effective on reducing fatigue loads and vibration.

Rainy days usually degrade the visual effect of images and videos. At present, most deraining models for single images adopt gradual optimization or elimination to remove rain streaks, but actually with relatively low efficiency in real tasks. An efficient one-stage deraining model, Efficient Transformer Derain Network (ETDNet), is proposed to remove rain streaks in single images efficiently. A new Transformer architecture is designed to provide rich multiple scales and context information, making the model extract features in a coarse-to-fine way. Multiple expansion filters with different expansion rates are embedded to predict the suitable kernel for each pixel of the rainy image in a multi-scale way. A multi-scale Loss Function is introduced to restore features with high-fidelity and detail textures. Experiments on Rain100L, Rain100H, and SPA datasets show that the proposed ETDNet reaches the highest PSNR and SSIM values at the fastest speed compared with other models.

Named Entity Recognition (NER) is one key step for constructing power domain knowledge graph which is increasingly urgent in building smart grid. This paper proposes a new NER model called Att-CNN-BiGRU-CRF which consists of the following five layers. The prefix Att means the model is based on attention mechanism. A joint feature embedding layer combines the character embedding and word embedding based on BERT to obtain more semantic information. A convolutional attention layer combines the local attention mechanism and CNN to capture the relationship of local context. A BiGRU layer extracts higher-level features of power metering text. A global multi-head attention layer optimizes the processing of sentence level information. A CRF layer obtains the output tag sequences. This paper also constructs a corresponding power metering corpus data set with a new entity classification method. The novelties of our work are the five layer model structure and the attention mechanism. Experimental results show that the proposed model has high recall rate 88.16% and precision rate 89.33% which is better than the state-of-the-art models.

The emerging link flooding attacks (LFAs) are one type of attacks that attract significant attention in both academia and industry against the routing infrastructure. The attack traffic flows originating from bots (e.g., compromised IoT devices) are deliberately aggregated at upstream critical links and grow intensified, gradually making a network connected to the critical links disconnected. Although LFAs are far more sophisticated than traditional DDoS attacks, whether such sophistication comes without a downside has never been investigated. In this paper, by modeling link flooding attacks and defenses, we tackle a series of questions concerning the practical issues of LFAs. Specifically, from the perspective of attacks, we advance a novel notion of strike precision, and reveal that LFAs may exhibit attack interference (i.e., unexpectedly interfere the connectivity of innocent networks) which might undermine the stealthiness and persistence of LFAs. From the perspective of defenses, we make the first step to study attack intention, i.e., inversely inferring the target network to disconnect based on the identified links under attack. Furthermore, we consider a strong defender who employs traffic engineering to mitigate LFAs, and formulate the game-theoretic interactions between attackers and defenders. The experiment results show that attack interference is pervasive, and our proposed SPAH flooding strategy can substantially lower attack interference and increase strike precision. Moreover, we demonstrate that LFAs can be effectively mitigated based on traffic engineering from a game-theoretic perspective, wherein the defender can adopt non-cooperative measurement techniques to achieve light-weight and multi-protocol-based robust probe deployment.

Urban network traffic congestion can be caused by disturbances, such as fluctuation and disequilibrium of traffic demand. This paper designs a distributed control method for preventing disturbance-based urban network traffic congestion by integrating Multi-Agent Reinforcement Learning (MARL) and regional Mixed Strategy Nash-Equilibrium (MSNE). To enhance the disturbance-rejection performance of Urban Network Traffic Control (UNTC), a regional MSNE concept is integrated, which models the competitive relationship between each agent and its neighboring agents in order to improve the decision-making process of MARL. The learning rate is enhanced with a self-adaptive ability to avoid a local optimal dilemma; Jensen-Shannon (JS) divergence is utilized to define the learning rate of the modified MARL. A two-way rectangular grid network with nine intersections is modeled via a Cell Transmission Model (CTM). A probability distribution mechanism, which can update the turn ratio of each approach dynamically and discretely, is established to represent the segmented route-decision process of the vehicles. The effectiveness of the proposed control method is evaluated through simulations in the grid network. The results show the influence of major disturbances, such as fluctuation of vehicle arrival rate, fluctuation of traffic demand (e.g. a rapidly rising flow and extreme changes in origin-destination distribution), and disequilibrium of traffic demand (e.g. different arrival flows at each boundary of the urban network), on the performance of the suggested control method. The results can be used to improve the state of the art in order to reduce urban network traffic congestion due to these disturbances.

In this paper, a novel circuit architecture is proposed to generate narrow pulses with high flexibility and without high cost. The desired narrow pulse signal is obtained via combining multiple sub-signals. The bandwidth of individual sub-signals is much smaller than the bandwidth of the desired pulse signal, and thus the sub-signals can be generated by low-speed digital-to-analog converters (DACs). When the sub-signals are combined, their spectra are stitched into the wideband spectrum of narrow pulse signal. Theoretically, arbitrary pulse waveforms can be synthesized through the proposed “spectrum stitching” technique without resorting to high-speed DACs, as long as the sub-signals are synchronized among each other in time. A prototype circuit is implemented to verify the fundamental scheme of the proposed “spectrum stitching” technique. The experimental results demonstrate that, Dirichlet pulses with temporal width of approximately 4 ns are successfully synthesized through combining the outputs of a regular 80-MHz waveform generator. In addition, numerical and experimental results indicate that, the proposed “spectrum stitching” circuit architecture is quite insensitive to various practical errors.

In this article, a direct neural network based adaptive backstepping control approach is proposed for a class of uncertain non-affine ship manoeuvring pure-feedback nonlinear systems. To carry out the backstepping design, the high fidelity 3 degrees of freedom Manoeuvring Modelling Group (MMG) model with external disturbances is transformed into the ship manoeuvring systems in non-affine pure-feedback form. Then, by combing the Implicit Function Theorem, Mean Value Theorem and dynamic surface control technique, the proposed approach is able to avoid completely the circularity problem and complexity growing problem exist in the adaptive neural network controller. During the controller design, the uncertain nonlinear functions are approximated by neural networks. Following this control approach, it is worth noting that the direct adaptive backstepping control for the high fidelity MMG model based ship manoeuvring nonlinear systems is achieved firstly, and the controller structure is simpler. Furthermore, it is shown via stability analysis that all signals in the closed-loop system are uniformly ultimately bounded. At last, two reference signals consist of a constant and a realistic performance requirement of ship are applied to simulation studies to illustrate the utility and merits of the proposed control scheme.

Seawater quality prediction has a tremendous potential of enabling future smart ocean. However, this time-sensitive application puts forward a strict delay requirement, thus easily leading to overwhelmed networks. Edge computing is emerging as an effective means of solving network overload, due to its edge-based distributed processing. Therefore, we develop a hybrid multivariate prediction model for seawater quality assessment in an edge computing environment, considering the combination of principal component analysis (PCA) and relevance vector machine (RVM). The PCA method is employed for dimension reduction of ten seawater quality factors in advance. Six principal components are extracted from multiple features, used as input variables of the subsequent predictor. Finally, a RVM is developed to predict the future trends of dissolved oxygen and pH, measuring seawater quality. Experimental results on the real-world ocean sensor data show that our PCA-RVM based multivariate prediction model outperforms single RVM, SVM and its extended version in prediction accuracy and efficiency, meanwhile statistical testings confirm this finding.

We propose a new method to achieve single-polarization guidance in a hollow-core negative-curvature fiber. Simulation results show that coating a layer of high-index material with a suitable thickness on the inner surface of one vertical cladding tube in a hollow-core negative-curvature fiber can effectively suppress y-polarized fundamental mode. Silicon is chosen as the high-index material in this paper. In the five-nested-tube silicon-coated negative-curvature fiber, the single-polarization region is located in the wavelength range of 1512-1587 nm with a broad bandwidth of 75 nm. The polarization extinction ratio is up to 1732 at the wavelength of 1550 nm. Moreover, the mode evolutions of xand y-polarized fundamental modes in the five-nested-tube silicon-coated negative-curvature fiber are investigated to explain why y-polarized fundamental mode is suppressed. Compared with the traditional method of using different tube thicknesses to achieve single-polarization guidance in a negative-curvature fiber, this method possesses the advantages of broad single-polarization bandwidth and high error tolerance.

The Dynamic Traffic Assignment (DTA) is one of the important measures to alleviate urban network traffic congestion. The congestions are usually caused by stochastic traffic demands, which are generally unassignable from time dimension in the real-world but are assumed to be assignable in existing DTA methods (i.e. real-time travel demands). In this paper, a distributed DTA method for preventing urban network traffic congestion caused by stochastic real-time travel demands by improving Multi-Agent Reinforcement Learning (MARL). A team structure, which consists of decision-makers and advisers, is designed to learn parallelly in realistic DTA tasks. To reduce the size of the solution space adaptively, the dynamic critical values advised by adviser agents are adopted as constraints for the strategy space of decision-makers (i.e. main agents). A collaborative heterogeneous-adviser mechanism is designed to avoid deviation of guidance. To enhance the adaptability of DTA to the changeable external environment, the mixed strategy concept is introduced to improve the decision-making process of main agents. The respective mapping mechanisms are designed to define adaptive learning rates to improve the sensitivity of MARL. The Sioux Falls (SF) network is established as a test platform via a Dynamic Network Loading (DNL). The effectiveness of the suggested DTA method is assessed through numerical simulations SF network. Under the influence of the scenario with stochastic real-time travel demands, the results show that the proposed method outperforms in terms of the throughput of the network and the individual average travel time among the overall network. Additionally, the ability of the proposed method in response to the external environment rapidly has also been demonstrated. Adopting the suggested method can improve the state of the art to assign stochastic real-time travel demands dynamically and to avoid potential traffic congestion fundamentally.

In this paper, the robust controller design problem of uncertain multi-input multi-output nonlinear non-minimum phase system is discussed. The nonlinear system is suffering from both uncertainty and input delay, so the controller design is difficult. The traditional stable inversion controller is utilized and extended to uncertain case. An integral of past control input is constructed and fuzzy logical system is utilized for approaching the unknown state matrix and input matrix. Then a robust adaptive control strategy is presented. Finally, a numerical simulation on vertical takeoff and landing aircraft is given to show the effectiveness of the proposed method.

From November 2017 to December 2018, Beidou navigation satellite system (BDS) completed the basic system construction of Beidou-3 system (BDS-3) and opened the initial service, BDS-3 satellites are equipped with Ka-band inter-satellite link (ISL) payloads, satellite autonomous orbit determination (OD) and time synchronization (TS) can be realized through ISL technology. At present, 18 BDS-3 medium earth orbit (MEO) satellites have been formally networked, and a series of ground-satellite link (GSL) test and connection test of Ka-band ISL signals are carried out, through the test, the working state and functional performance of the ISL payloads can be evaluated. This article focuses on the GSL measurements and performance evaluation of the ISL signals of BDS-3, the basic architecture of the GSL test platform is given, the principle of Ka-band GSL measurement and basic data processing methods and performance evaluation methods are introduced. Obtained the GSL observation data of 18 BDS-3 MEO satellites using the Ka-band ground station of the National Time Service Center (NTSC), and evaluate the performance of the GSL ranging and time synchronization. The result shows that GSL ranging accuracy of 18 MEO satellites is better than 0.02m and GSL time synchronization accuracy is better than 0.2ns. Thus, this meets the requirements of ISL system. The assessment methods and analysis results presented in this article can provide the reference of the function and performance verification for inter-satellite link system, and provide technical and data accumulation for the time synchronization extension application based on ISL signals.

Automated pavement crack image segmentation is challenging because of inherent irregular patterns, lighting conditions, and noise in images. Conventional approaches require a substantial amount of feature engineering to differentiate crack regions from non-affected regions. In this paper, we propose a deep learning technique based on a convolutional neural network to perform segmentation tasks on pavement crack images. Our approach requires minimal feature engineering compared to other machine learning techniques. We propose a U-Net-based network architecture in which we replace the encoder with a pretrained ResNet-34 neural network. We use a “one-cycle” training schedule based on cyclical learning rates to speed up the convergence. Our method achieves an $F1$ score of 96% on the CFD dataset and 73% on the Crack500 dataset, outperforming other algorithms tested on these datasets. We perform ablation studies on various techniques that helped us get marginal performance boosts, i.e., the addition of spatial and channel squeeze and excitation (SCSE) modules, training with gradually increasing image sizes, and training various neural network layers with different learning rates.

Ocean wave parameters, including wave height, direction and period, are necessary to improve forecasts of ocean wave conditions. Their accurate observation contributes to safety of both offshore operations and military operations. However, conventional wave observation methods that combine the three-axis accelerometer-based wave sensor with the gyroscope may show problems such as inner arm error, transverse sensitivity effect, and gyroscope drift. The first two are due to non-coincidence of centroids inside the accelerometer. Such problems will reduce reliability of data. To solve the problems, a new type of parallel six-dimensional accelerometer is developed to reduce principle errors and improve stability of wave buoys. The solution features a single mass and 12-chain redundancy. When compared with the performance of high-precision sensors in the wave pool, the wave buoy based on six-dimensional accelerometer is proved to be more accurate in wave height measurement than 3.75 % × measurement value, with wave direction accuracy of ± 1°, and the wave period measurement accuracy of ± 0.1s.

In this paper, a novel substrate integrated waveguide (SIW) filtenna with two controllable radiation nulls is proposed, which has the advantage of good radiation characteristics. Radiation nulls are generated by electric and magnetic mixed coupling structure and fundamental mode, respectively. And a quasi-elliptic characteristic can be obtained at the boresight gain direction. By changing the length of perturbation slot and the strength of electric coupling and magnetic coupling, the position of radiation nulls can be controlled independetly. Compared to conventional filtenna, the proposed SIW filtenna has no extra filtering circuit to increase the dimension and complexity. Two resonant modes in the passband can be excited through loading electric and magnetic mixed coupling structure in one SIW cavity. The measured results are in good agreement with simulated results. The good frequency selectivity and out-of-band rejection are achieved. The proposed antenna exhibits a maximum realized gain of 5.8 dBi at the operating bandwidth. The measured impedance bandwidth is from 9.44 GHz to 9.68 GHz and the cross-polarization levels are lower than -20 dB in both planes.

Multimodal image registration is a crucial step for a variety of medical applications to provide complementary information from the combination of various data sources. Conventional image registration methods aim at finding a suited similarity metric as well as a descriptive image feature, which is quite challenging due to the high diversity of tissue appearance across modalities. In this paper, we present a novel approach to register images via an asynchronously trained reinforcement learning agent automatically. Within this approach, convolutional gated recurrent units (ConvGRU) is incorporated after stacked convolutional layers to extract both spatial and temporal features of the neighboring frames and implicitly learn the similarity metric. Moreover, we propose a customized reward function driven by fixed points error (FPE) to guide the agent to the correct registration direction. A Monte Carlo rollout strategy is also leveraged to perform a look-ahead inference to the elimination of jitter in the test stage. Evaluation is performed on paired CT and MR images from patients diagnosed as nasopharyngeal carcinoma. The results demonstrate that our method achieves state-of-the-art performance in medical image registration.

Based on tapered Triple-Core Photonic Crystal Fiber (TCPCF), a high performance tunable Dual-Wavelength Erbium-Doped Fiber Laser (DW-EDFL) is proposed and experimentally demonstrated. The mode-coupling among three cores of TCPCF appears periodically and becomes further enhanced in tapered TCPCF, which are beneficial to form a filter to select lasing wavelength. A series of tapered TCPCFs with different taper lengths are fabricated and inserted into tunable DW-EDFL as filters. Multiple groups of tunable dual-wavelength lasing outputs are achieved by using the tapered TCPCF filter with a waist diameter of 126 μm. The maximum tunable range can reach up to 14.36 nm and the side mode suppression ratio of all lasing outputs are above 52 dB. In addition, the tunable single- and triple-wavelength lasers are also obtained. The tunable single-wavelength laser has an excellent linear response to strain and can be applied to realize a high-sensitivity strain sensor with the highest strain sensitivity of -13.1 pm/με.

The convergence of optical and wireless networks, and the marginalization of centralized data centers are required for next-generation telecommunications. According to the operators' concern of cost-efficient deployment combined with latency-sensitive requirements, network planning will lay a foundation for the development of next-generation networks. In particular, the network planning of mobile edge computing (MEC)-enabled cloud radio access networks (C-RAN) in a cost-effective and low-latency manner becomes exceedingly challenging. However, most research studies focus on the cost-effective deployment of C-RAN without considering the optimal placement of MEC servers simultaneously. Moreover, deployment costs and access latency will be directly affected by the optimal deployment of both the MEC servers and access networks. Therefore, we investigate how to coordinate the optimal deployment of MEC servers and C-RAN for latency-sensitive services. In detail, a MILP model is proposed to minimize the joint cost of latency and network deployment of time division multiplexing-passive optical network (TDM-PON) based MEC-enabled C-RAN. Given the model's complexity, a heuristic algorithm is also proposed to solve the mixed integer linear programming (MILP) model. Simulations are conducted to evaluate the performance comparisons of different approaches under different network scenarios. The impacts of some key parameters on comparisons of different approaches are also analyzed.