Search

This paper proposes a real-time multi-scale semantic segmentation network (MsNet). MsNet is a combination of our novel multi-scale fusion with matching attention model (MFMA) as the decoding network and the network searched by asymptotic neural architecture search (ANAS) or MobileNetV3 as the encoding network. The MFMA not only extracts low level spatial features from multi-scale inputs but also decodes the contextual features extracted by ANAS. Specifically, considering the advantages and disadvantages of the addition fusion and concatenation fusion, we design multi-scale fusion (MF) that balances speed and accuracy. Then we creatively design two matching attention mechanisms (MA), including matching attention with low calculation (MALC) mechanism and matching attention with strong global context modeling (MASG) mechanism, to match varying resolutions and information of features at different levels of a network. Besides, the ANAS performs the deep neural network search by employing an asymptotic method and provide an efficient encoding network for MsNet, releasing researchers from those tedious mechanical trials. Through extensive experiments, we prove that MFMA, which can be applied to numerous recognition tasks, possesses excellent decoding ability. And we demonstrate the effectiveness and necessity of implementing the ``matching'' attention mechanism. Finally, the proposed two versions, MsNet₋ANAS and MsNet₋M achieve a new state-of-the-art trade-off between accuracy and speed on the CamVid and Cityscapes datasets. More remarkably, on the Nvidia Tesla V100 GPU, our MsNet₋ANAS achieves 74.1% mIoU with the speed of 184.2 FPS on the CamVid while 72.9% mIoU with the speed of 119.9 FPS on the Cityscapes.

Although many graph convolutional neural networks (GCNNs) have achieved superior performances in semisupervised node classification, they are designed from either the spatial or spectral perspective, yet without a general theoretical basis. Besides, most of the existing GCNNs methods tend to ignore the ubiquitous noises in the network topology and node content and are thus unable to model these uncertainties. These drawbacks certainly reduce their effectiveness in integrating network topology and node content. To provide a probabilistic perspective to the GCNNs, we model the semisupervised node classification problem as a topology-constrained probabilistic latent space model, probabilistic graph convolutional network (PGCN). By representing the nodes in a more efficient distribution form, the proposed framework can seamlessly integrate the node content and network topology. When specifying the distribution in PGCN to be a Gaussian distribution, the transductive node classification problems can be solved by the general framework and a specific method, called PGCN with the Gaussian distribution representation (PGCN-G), is proposed. To overcome the overfitting problem in covariance estimation and reduce the computational complexity, PGCN-G is further improved to PGCN-G+ by imposing the covariance matrices of all vertices to possess the identical singular vectors. The optimization algorithm based on expectation-maximization indicates that the proposed method can iteratively denoise the network topology and node content with respect to each other. Besides the effectiveness of this top-down framework demonstrated via extensive experiments, it can also be deduced to cover the existing methods, graph convolutional network, graph attention network, and Gaussian mixture model and elaborate their characteristics and relationships by specific derivations.

In this letter, we investigate the performance of reconfigurable intelligent surface (RIS)-aided networks in the presence of non-identically distributed interferers at the destination. A mathematical framework is proposed to statistically characterize the cascaded channels and to evaluate the performance in terms of outage probability (OP) and average bit error rate (BER). The high signal-to-interference-plus-noise ratio (SINR) analysis reduces to easy-to-compute expressions for the outage probability and BER, which reveals inside information for the system design. We also extend the analysis to study multi-RIS networks with two schemes. Representative results show that the RIS element number and interference setting have a significant impact on the overall system performance.

Differential frequency hopping (DFH) technique is widely used in wireless communications by exploiting its capabilities of mitigating tracking interference and confidentiality. However, electronic attacks in wireless systems become more and more rigorous, which imposes a lot of challenges on the DFH sequences designed based on the linear congruence theory, fuzzy and chaotic theory, etc. In this article, we investigate the sequence design in DFH systems by exploiting the equivalence principle between the G-function algorithm and the encryption algorithm, in order to achieve high security. In more details, first, the novel G-function is proposed with the aid of the Government Standard algorithm and the Rivest–Shamir–Adleman algorithm. Then, two sequence design algorithms are proposed, namely, the G-function-assisted sequence generation algorithm, which takes the full advantages of the symmetric and asymmetric encryption algorithms, and the high-order G-function-aided sequence generation algorithm, which is capable of enhancing the correlation of the elements in a DFH sequence. Moreover, the security and ergodicity performance of the proposed algorithms are analyzed. Our studies and results show that the DFH sequences generated by the proposed algorithms significantly outperform the sequences generated by the reversible hash algorithm and affine transformation in terms of the uniformity, randomness, complexity, and the security.

Congestion is an economic phenomenon of overinvestment that occurs when excessive inputs decrease the maximally possible outputs. Although decision-makers are unlikely to decrease outputs by increasing inputs, congestion is widespread in reality. Identifying and measuring congestion can help decision-makers detect the problem of overinvestment. This paper reviews the development of the concept of congestion in the framework of data envelopment analysis (DEA), which is a widely accepted method for identifying and measuring congestion. In this paper, six main congestion identification and measurement methods are analysed through several numerical examples. We investigate the ideas of these methods, the contributions compared with the previous methods, and the existing shortcomings. Based on our analysis, we conclude that existing congestion identification and measurement methods are still inadequate. Three problems are anticipated for further study: maintaining the consistency between congestion and overinvestment, considering joint weak disposability assumption between desirable outputs and undesirable outputs, and quantifying the degree of congestion.

Communication between nanomachines is still an important topic in the construction of the Internet of Bio-Nano Things (IoBNT). Currently, molecular communication (MC) is expected to be a promising technology to realize IoBNT. To effectively serve the IoBNT composed of multiple nanomachine clusters, it is imperative to study multiple-access MC. In this article, based on the molecular division multiple access technology, we propose a novel multiuser MC system, where information molecules with different diffusion coefficients are first employed. Aiming at the user fairness in the considered system, we investigate the optimization of molecular resource allocation, including the assignment of the types of molecules and the number of molecules of a type. Specifically, three performance metrics are considered, namely, min-max fairness for error probability, max-min fairness for achievable rate, and weighted sum-rate maximization. Moreover, we propose two assignment strategies for types of molecules, i.e., best-to-best (BTB) and best-to-worst (BTW). Subsequently, for a two-user scenario, we analytically derive the optimal allocation for the number of molecules when types of molecules are fixed for all users. In contrast, for a three-user scenario, we prove that the BTB and BTW schemes with the optimal allocation for the number of molecules can provide the lower and upper bounds on system performance, respectively. Finally, numerical results show that the combination of BTW and the optimal allocation for the number of molecules yields better performance than the benchmarks.

In this work, the authors consider the security performance for the mixed radio frequency link and visible‐light link cooperative relaying system, where the relay has the energy collection ability and can extract the direct current part existing in the visible‐light signal. Then, the relay uses the collected energy to demodulate the received visible‐light signal and retransmit it through the radio frequency link. With this equipment, some analytical expressions of the average secrecy capacity and the secrecy outage probability are derived. Furthermore, the asymptotic expression is also derived to examine the influence of the various configuration parameters on the secrecy performance of the mixed relaying system. At last, some numerical and simulation results are given to verify the theoretical analysis.

In this letter, a performance study of a mixed dual-hop communication system consisting of free-space optical/underwater wireless optical communication (FSO-UWOC) links is presented. In particular, considering both heterodyne detection and intensity modulation with direct detection (IM/DD), we assume that the FSO channel suffers a Gamma-Gamma fading in the presence of pointing error impairments, while the UWOC channel experiences a mixture Exponential-Generalized Gamma (EGG) fading with pointing errors. Assuming a fixed-gain amplify-and-forward (AF) protocol, closed-form expressions for the outage probability, average bit-error-rate, and average capacity in terms of the bivariate Fox’s H-function are derived. To verify the derived analytical expressions, numerical results are provided to verify the accuracy of our results.

Humor and pun are widely used in our daily life. Due to the incongruous and inconsistent expressions in humor and pun, humor recognition and pun detection are challenging tasks in sentiment analysis and natural language processing. Pun is usually used as one of the humorous rhetorical devices. Thus, determining the puns in the texts can be pertinent to the successful recognition of humor and vice versa. Existing models of humor recognition and pun detection usually dealt with the two tasks independently. Therefore, we propose an attention-based multi-task learning model (ABML) to train the two tasks jointly. Our model unifies the two highly pertinent tasks, including the humor recognition and pun detection. In the ABML model, we design a Co-Encoder module to capture the common features between the two tasks by weight sharing. Apart from the Co-Encoder module, we also design two Private Encoder modules for the two tasks, respectively. The Private Encoder module is used to capture the private semantic feature of the two tasks. We conduct sufficient experiments on two widely used datasets, including a humor dataset and a pun dataset, to verify the effectiveness of our model. The experimental results show our model can get comparable performance compared with state-of-the-art methods.

In this paper, we investigate the performance of a mixed radio-frequency-underwater wireless optical communication (RF-UWOC) system where an unmanned aerial vehicle (UAV), as a low-altitude mobile aerial base station, transmits information to an autonomous underwater vehicle (AUV) through a fixed-gain amplify-and-forward (AF) or decode-and-forward (DF) relay. The analysis accounts for the main factors that affect the system performance, such as the UAV height, air bubbles, temperature gradient, water salinity variations, detection techniques, and pointing errors. Employing the fixed-gain AF and DF relays, expressions for some key performance metrics are derived, e.g., outage probability (OP), average bit error rate (ABER), and average channel capacity. Moreover, in order to get further insights, asymptotic analyses for the OP and ABER are also carried out. Additionally, for the two relaying systems, we derive analytical expressions for the optimal UAV altitude that minimizes the OP. Simulation results show that the UAV altitude influences the system performance and there is an optimal altitude which ensures a minimum OP. Furthermore, it is demonstrated that the diversity order of the fixed-gain AF relaying depends on the RF link and pointing errors, while the diversity order of the DF relaying depends on the detection techniques and pointing errors.

This paper is focused on code-domain non-orthogonal multiple access (CD-NOMA), which is an emerging paradigm to support massive connectivity for future machine-type wireless networks. We take a comparative approach to study two types of overloaded CD-NOMA, i.e., sparse code multiple access (SCMA) and dense code multiple access (DCMA), which are distinctive from each other in terms of their codebooks having sparsity or not. By analysing their individual diversity orders (DO) in Rayleigh fading channels, it is found that DCMA can be designed with the aid of generalized sphere decoder (i.e., a nonlinear multiuser detector) to enjoy full DO which is equal to the maximum number of resource nodes in the system. This is in contrast to SCMA whose error rate suffers from limited DO equal to the codebook sparsity (i.e., the effective number of resource nodes occupied by each user). We conduct theoretical analysis for the codebook design criteria and propose to use generalized sphere decoder for DCMA detection. We numerically evaluate two types of multiple access schemes under “ $4\times 6$ ” (i.e., six users communicate over four subcarriers) and “ $5\times 10$ ” NOMA settings and reveal that DCMA gives rise to significantly improved error rate performance in Rayleigh fading channels, whilst having decoding complexity comparable to that of SCMA.

In-band full-duplex (IBFD) systems require promising resource allocation (RA) strategies to fully exploit the available time-frequency resources. Furthermore, the acquisition of channel state information and signal reception in IBFD systems are significantly impacted by insufficient self-interference cancellation (SIC), impeding the applications of IBFD in practical wireless systems. Multicarrier-division duplex (MDD), which benefits low-budget SI mitigation in digital domain and flexible subcarrier assignment, is expected to be a promising transitional technique from half-duplex (HD) to IBFD. To demonstrate the advantages of MDD over HD, this paper first compares the upper-bound performance of MDD and HD by applying unfair greedy RA. Then, considering the millimeter-wave (mmWave) with hybrid beamforming, we propose the RA optimization with the quality-of-service constraints on both downlink (DL) and uplink (UL) mobile-stations (MSs). To solve this non-convex RA problem, we divide it into a suboptimal subcarrier allocation problem solved by the proposed improved fair greedy (IFG) algorithm, and a convex power allocation problem. Furthermore, we design two general hybrid precoder based on matrix factorization and direct approach, and a combiner having high SIC capability. Our results show that the proposed RA algorithm can achieve the performance near the upper-bound achieved by the unfair greedy algorithm, while guaranteeing the proportional fairness among all DL/UL MSs. The performance of the two precoding schemes is depended on the number of radio frequency chains supported. Finally, the proposed SIC algorithm is able to provide sufficient SI mitigation, which can be implemented without impacting the RA operation.

This article aims at proposing a multistage linear Gauss pseudospectral method (MS-LGPM) for solving the piecewise continuous nonlinear optimal control problem (OCP) with interior-point constraints and terminal constraints. First, the first-order necessary conditions for the multistage nonlinear OCP are derived, and a typical multipoint boundary value problem is obtained. Second, the original problem can be solved iteratively by integral prediction and quasi-linearization. Then Lagrange interpolation polynomials are used to approximate the state, control, and costate variables, to transfer those differential equations into a set of linear algebraic equations. Therefore, the control update coming closer to the optimal solution can be derived in an analytical manner. Additionally, those linear algebraic equations are formulated in regular banded matrix forms to make the code as concise as possible. Finally, the proposed method is applied to the midcourse guidance design for a dual-pulse missile to evaluate its performance. Simulation results show that the proposed method performs well in providing the optimal control with high accuracy and computational efficiency. Furthermore, a comparison with other typical guidance method and Monte Carlo simulations are also provided. Results demonstrate, even with some random uncertainties, the proposed method still has strong robustness and superior performance.

Instance segmentation has a wide range of applications, including video surveillance, autonomous driving, and behavior analysis. Nevertheless, as a type of pixel-level segmentation, its prediction performance in practice is substantially affected by low-resolution (LR) images resulting from the limitations of image acquisition equipment and poor acquisition conditions. Moreover, because their immense computational costs prevent the implementation of existing segmentation models on embedded devices, the development of a lightweight segmentation model has become an urgent necessity. However, it is challenging to achieve sound results with high efficiency and portability. From another perspective, to improve understanding of detailed objects, an architecture is needed that promotes an advanced interpretation of the segmentation, that is, a refined mask with texture. Our main contribution, called TextureMask, consists of the MobileNet-FPN for Mask R-CNN methods, segmentation with cropping, and a gradient sensitivity map, which are then merged into a unified map to refine and enrich the mask with texture information. Furthermore, preprocessing and post-processing algorithms are incorporated. Experiments demonstrated that our technique exhibits good pixel-level segmentation performance in terms of both accuracy and computational efficiency for a given LR input, and it can be easily implemented in embedded platforms.

Animals can use the maneuver of a flipping body and flapping wings to reduce the normal rebound force of impact during landing, decreasing the adsorption force required by the contact point. This capability aids aerial vehicles with landing not only on vertical surfaces, but also on inclined surfaces. In this study, a “ flip-and-flap ” biomimetic strategy is presented that enables a high-speed flying object to perch on inclined surfaces without speed reduction before touchdown. During perching, the kinetic energy of the aerial vehicle is first converted to potential energy by flipping upward with a customized extended leg. Then, with a controlled thrust, the potential energy is consumed during the downward-swing phase. Thus, the aerial vehicle can land smoothly on a wall. Simplified and multilink models were designed to simulate the landing of a flying object, and the dynamic interaction between the flying object with an extended-leg mechanism and the surface was elaborated. A set of experiments on perching performed with the proposed strategy and customized extended leg, demonstrating the effectiveness and robustness of the strategy.

It has been proved that hyperspectral image (HSI) classification task benefits from introducing additional spatial information. However, how to classify high-dimensional hyperspectral images under the condition of limited training samples is still a challenge. In this paper, we propose a hyperspectral image classification framework based on discriminant manifold broad learning system (DMBLS). We introduce the manifold structure and discriminant information of the data samples as prior knowledge into broad learning system (BLS), which effectively solves the problem of insufficient learning caused by BLS with limited training samples. Firstly we construct two different types of manifold structures for class modelling: the intra-class manifold structures and the inter-class ones. Secondly, we integrate these local discriminant manifold structures to build a manifold regularization framework. Finally, we add this framework to the DMBLS via minimizing the intra-class manifold structures, while maximizing the inter-class ones. And we think this operation can optimize the projection direction of output weights in the DMBLS, enhancing the discrimination ability of these weights. Experimental results on three benchmark HSI datasets show that DMBLS can effectively improve the classification accuracy of hyperspectral images under the condition of limited training samples.

This article aims at proposing a multisystem gradient method (MSGM) for solving the integrated nonlinear optimal control problem with multiple systems and unspecified terminal time. This method is based on the influence functions, which are used to predict changes in performance indices and terminal conditions caused by changes in control and terminal time, so as to update the improvements of control and terminal time. To improve the computational efficiency, the linear Gauss pseudospectral technology is employed, which successfully transfers the nonlinear differential equations into a set of linear algebraic equations. In addition, all systems are discreted in the orthogonal points, which leads to the fact that the coupling degree among the systems is very low. In most calculations, each system is independent of the other, and only the multiplier associated with terminal time needs to be solved simultaneously. In other words, the computational complexity will be significantly reduced, which is suitable for online applications. Finally, MSGM is applied to the multimissile cooperative guidance with the same attack time. The simulation results show that the algorithm has high computational efficiency and fast convergence rate. In addition, the stability and robustness of the algorithm are also excellent.

Session-based recommendation is proposed to predict user preferences in a short anonymous interaction session. There are two kinds of characteristic in the session: sequential dependencies and collective dependencies. Sequential dependencies mean that user-item interactions in the session are strictly ordered, while collective dependencies mean that multiple interactions with a flexible order jointly determine the user’s next behaviors. However, existing methods only take one of the two characteristics into account, which leads to the fact that they can’t model user behavior well. Furthermore, few methods emphasize the first interaction in the session, although user’s initial intention can be reflected by it to a large extent. To this end, we propose a novel model, Dual Part-pooling Attentive Networks for session-based Recommendation (DPAN4Rec), which is capable of capturing the sequential dependencies and collective dependencies of sessions simultaneously. By proposing a part-pooling attention mechanism, DPAN4Rec explores user’s initial intention from first click and filters out noisy clicks in the session. Extensive experiments have been conducted on two benchmark e-commerce datasets, Yoochoose and Diginetica, and the experimental results show that DPAN4Rec outperforms state-of-the-art methods. Furthermore, our study demonstrates that it is necessary to consider both sequential dependencies and collective dependencies in session-based recommendation.

Sarcasm is a common way of rhetoric in our daily life. It is used to express the opposite of the literal meaning, which makes it a challenging task in sentiment analysis of natural language processing (NLP). The formation mechanism of sarcasm is usually caused by the contrast between the positive sentiment and the negative situation. In this paper, we propose a contextual dual-view attention network (CDVaN) for sarcasm detection according to the formation mechanism of sarcasm. A Contrast Understanding Unit is proposed to effectively extract the contrast between the positive sentiment and the negative situation from the view of formation mechanism of sarcasm. Apart from it, we further use a Context Understanding Unit to extract the contextual semantic information from the contextual semantic view. Our experiments on the IAC-V1 dataset and IAC-V2 dataset demonstrate that the proposed CDVaN model can distinguish sarcasm effectively. The results show that our model achieves state-of-the-art or comparable results.

Graph neural networks have achieved tremendous success in semi-supervised node classification. In this paper, we firstly analyse the propagation strategies in two milestone methods, Graph Convolutional Network (GCN) and Graph Attention Network (GAT), to reveal their underlying philosophies. According to our analysis, the propagations in GAT can be interpreted as learnable and asymmetric local attribute augmentations, while that of GCN can be interpreted as fixed and symmetric local attribute smoothing. Unfortunately, the local attribute augmentations in GAT is not adequate in certain circumstances, because the nodes tend to possess similar attributes in local neighbourhoods. With a toy experiment, we manage to demonstrate the necessity to incorporate global information. Therefore, we propose a novel Graph Co-ATtention Network (Graph-CAT), which performs both the local and global attribute augmentations based on two different yet complementary attention schemes. Extensive experiments in both the transductive and inductive tasks demonstrate the superiority of our Graph-CAT compared to the state-of-the-art methods.

BACKGROUND: There is a great demand for the extraction of organ models from three-dimensional (3D) medical images in clinical medicine diagnosis and treatment. OBJECTIVE: We aimed to aid doctors in seeing the real shape of human organs more clearly and vividly. METHODS: The method uses the minimum eigenvectors of Laplacian matrix to automatically calculate a group of basic matting components that can properly define the volume image. These matting components can then be used to build foreground images with the help of a few user marks. RESULTS: We propose a direct 3D model segmentation method for volume images. This is a process of extracting foreground objects from volume images and estimating the opacity of the voxels covered by the objects. CONCLUSIONS: The results of segmentation experiments on different parts of human body prove the applicability of this method.

In nano-networking, many nano-machines need to share common communication media, in order to achieve information exchange and data fusion. Multiple-access is an important technique for multiple nano-machines to send information to one access point (AP) or fusion center. Built on the Molecular Shift Keying (MoSK) modulation, this article proposes a novel Molecular Type Spread MoSK (MTS-MoSK) scheme for achieving multiple-access transmission in diffusion-based molecular communications (DMC). Correspondingly, two detection schemes are introduced and investigated, which are the Maximum Selection assisted Majority Voter Detection (MS-MVD) and Equal-Gain Combining Detection (EGCD). The error performance of the MTS-MoSK DMC systems with respectively the two detection schemes is analyzed, when both Multiple-Access Interference (MAI) and Inter-Symbol Interference (ISI) are taken into account. Furthermore, the performance of MTS-MoSK DMC systems is investigated and compared with the aid of analytical results as well as Monte-Carlo and particle-based simulations. Our studies and performance results show that MTS-MoSK constitutes a promising candidate for implementing multiple-access DMC, and MS-MVD has the potential to outperform EGCD in DMC.

Recognition of malposition and location of high-temperature forgings play a critical role in the realisation of robotised die forging, which is an ongoing trend in intelligent manufacturing. This study is aimed at the robotised die forging of the scraper beam of armoured face conveyor, which is the only transporting equipment used in the coal mine workface. Firstly, a novel process to recognise the malposition and location of high-temperature forgings using two monocular cameras, one placed horizontally and another vertically, is proposed. Secondly, a novel image preprocessing algorithm combining the algorithms of grey linear transformation, exponential transformation, and median filtering is proposed. After processing the high-temperature forging image using the proposed image preprocessing algorithm, the grey difference between the target region and background region of the processed image is highly increased. This is conducive to the subsequent image segmentation and contour extraction processes conducted in the forging region. Thirdly, after the comparison and analysis of the three commonly used image segmentation methods, including edge detection, threshold segmentation, and region growing methods, it is discovered that the region growing method is suitable for image segmentation of high-temperature forgings. Fourthly, a two-way modified and blob analysis based forging location algorithm is proposed to reduce the location error caused by the axial and radial asymmetric flash during the forging process. Finally, the proposed algorithms are validated by experiments and the location recognition error of the proposed location algorithm is only 0.86059 mm. This study provide technical support for the realisation of robotised die forging.

In this letter, we consider a network assisted by multiple reconfigurable intelligent surfaces (RISs). Assuming that the RIS with the highest instantaneous end-to-end signal-to-noise ratio (SNR) is selected to aid the communication, the outage probability (OP) and average sum-rate are investigated. Specifically, an exact analysis for the OP is developed. In addition, relying on the extreme value theory, a closed-form expression for the asymptotic sum-rate is derived, based on which the capacity scaling law is established. Our results are corroborated through simulations and insightful discussions are provided. In particular, our analysis shows that the number of RISs as well as the number of reflecting elements play a crucial role in the capacity scaling law of multiple RIS-aided networks. Also, comparisons with relay-aided systems are carried out to demonstrate that the proposed system setup outperforms relaying schemes both in terms of the OP and average sum-rate.

Existing WiFi recognition schemes are capable of discovering patterns of indoor semantics, such as human activity, identity, indoor environment, and so on. We note that channel state information (CSI) presents an opportunity for hackers to learn indoor privacy, however, currently there is a lack of security research on CSI. In this article, we are the first to discuss and define the security problem of CSI signals, which is further extended to the problems of nontargeted protection and targeted protection. To solve them, we present two types of adversarial autoencoder networks (AAENs). Through replacing the original signals with the generated adversarial ones, the protected semantic features are modified, and the significant features of the other semantics required to be recognized are reserved. Intensive evaluations demonstrate that with the proposed AAENs, the recognition accuracy of the protected semantic can be significantly decreased, while still maintaining the other semantics to be identified correctly.

This letter investigates user fairness in wireless powered communication networks (WPCNs) with non-orthogonal multiple access (NOMA) with the aim of maximizing the equal individual data rates. Initially, it is proposed the bisection method to obtain the optimal power allocation coefficients. Then, Lagrange dual-decomposition is considered for evaluating the optimal time-splitting ratio. Next, Newton’s method combined with the sub-gradient method are employed to achieve the optimal solution. In addition, two sub-optimal solutions that provide a good balance between performance and complexity are proposed. Lastly, simulation results are presented to evaluate the performance of the proposed solutions.

Benefiting from the high resolution in beamspace, millimeter wave (mmwave) communication has been regarded as a high-accuracy localization solution, where the location information is embedded in the channel via angle and time delay, for example. In this paper, to locate a user equipment (UE) and scatterers, we present the localization model in mmwave communications as a compressed sensing assisted channel estimation problem, which is solved using a proposed two-stage channel estimation based localization scheme. During the first stage, a sparse Bayesian learning (SBL) algorithm is operated to attain a coarse estimation. Then during the second stage, a multi-stage grid refinement assisted fine estimation is achieved by a distributed compressed sensing simultaneous orthogonal matching pursuit (DCS-SOMP) algorithm. Moreover, in our approach, the few-bit analog to digital converters (ADCs) are utilized by the receiver of UE so as to attain a good trade-off among performance, complexity and energy-efficiency. Finally, the performance of channel estimation and positioning is comprehensively investigated and compared. It can be shown that our proposed two-stage approach is capable of achieving centimeter-level accuracy with the required number of quantization bits of ADCs less than four.

In this paper, we investigate a nonconvex fractional quadratically constrained quadratic problem (fractional QCQP), which has a wide application to the resource allocation optimization in wireless communication systems. Different from the state-of-the-art methods needing to solve semidefinite programmings or second-order cone programmings, we propose a fast algorithm for solving fractional QCQP by combining the successive convex approximation method and the consensus alternating direction method of multipliers, which has only simple computations and works very fast in applications with modest accuracy. We also apply the proposed fast algorithm to secure beamforming design for enhancing physical layer security in cognitive nonorthogonal multiple access (NOMA) networks, where secrecy rate optimization problems in both underlay and overly cognitive NOMA networks are typical fractional QCQPs and have not been well studied. Simulation results have shown that our proposed fast algorithm achieves almost the same performance as the state-of-the-art methods, however, the proposed fast algorithm has very low computation complexity.

To provide high data rate for high-speed trains (HSTs), it is required to use emerging wireless communication systems, such as the fifth generation (5G). An asymmetric 5G mobile relay system is investigated for HSTs, where the mobile relay is deployed at the HST to avoid high penetration loss of the direct link between the base station (BS) and the users (UE) inside carriages. The sub-6GHz frequency is utilized for the BS-relay link while the relay-UE link adopts the millimeter wave frequency. Therefore, the BS-relay link experiences $\kappa $ - $\mu $ fading and the relay-UE link experiences static fluctuating two-ray fading. Moreover, the channel aging effect is considered due to the mobility of HST. For the considered system, we first study the exact statistical characterizations of the end-to-end signal-to-noise ratios. Then, we derive exact closed-form expressions for key performance metrics, such as outage probability, average bit-error rate, and average achievable rate per unit bandwidth. The significant effects of channel aging, system and channel parameters on the mobile relay system are revealed from theoretical analysis and are further illustrated by simulation results. Our investigation reveals that the mobile relay system is a promising network architecture for HST communications and can provide steady and high-speed data provisioning to HST passengers against the significant bottleneck of channel aging.

Sharing short personalized videos to various social media networks has become quite popular in recent years. This raises the need for digital retouching of portraits in videos. However, applying portrait image editing directly on portrait video frames cannot generate smooth and stable video sequences. To this end, we present a robust and easy-to-use parametric method to reshape the portrait in a video to produce smooth retouched results. Given an input portrait video, our method consists of two main stages: stabilized face reconstruction, and continuous video reshaping. In the first stage, we start by estimating face rigid pose transformations across video frames. Then we jointly optimize multiple frames to reconstruct an accurate face identity, followed by recovering face expressions over the entire video. In the second stage, we first reshape the reconstructed 3D face using a parametric reshaping model reflecting the weight change of the face, and then utilize the reshaped 3D face to guide the warping of video frames. We develop a novel signed distance function based dense mapping method for the warping between face contours before and after reshaping, resulting in stable warped video frames with minimum distortions. In addition, we use the 3D structure of the face to correct the dense mapping to achieve temporal consistency. We generate the final result by minimizing the background distortion through optimizing a content-aware warping mesh. Extensive experiments show that our method is able to create visually pleasing results by adjusting a simple reshaping parameter, which facilitates portrait video editing for social media and visual effects.

This paper investigates the performance of cooperative networks based on non-orthogonal multiple access (NOMA) with multiple full-duplex decode-and-forward relays. Taking into account rate fairness between the users and assuming imperfect successive interference cancellation (SIC), the problem is formulated so that power allocation is performed adaptively by the multiple relays in order to maximize the minimum users’ achievable rate. We demonstrate that the proposed optimization problem is convex and a closed-form expression for the power allocation factor is derived. The attained results show that the proposed scheme achieves satisfactory results in terms of achievable rate, fairness, and outage probability when compared to other adaptive power allocation strategies.

Extracting 3D structures from voxel based images can make doctors more directly observe the situation of the target in the clinic, making it easier for doctors to diagnose the condition and make the medicine teaching more directly and easier to understand. For this purpose, we propose a 3D volume image segmentation method based on the max-flow/min-cut algorithm. Our segmentation method can be applied directly to 3D volume image. After users marking small amount tags (foreground and background pixels), we put forward a method to use a directed connected graph structure to represent the volume image. In the directed connected graph, in order to speed up the efficiency of the segmentation in subsequent steps, we divide each voxel node in the graph into different color ranges, and each color range match up with an auxiliary node. In order to divide the color range more finely, we propose a method to calculate the color similarity. We then use the max-flow/min-cut algorithm to segment the directed connected graph. The result of experiments performed in multiple sets of slice images shows that our proposed method improves the efficiency, reduces human error on the 3D volume image segmentation task, and the result is complete and accurate.

In this letter, a reconfigurable intelligent surface (RIS)-assisted dual-hop visible light communication (VLC)/radio frequency (RF) system in indoor scenarios is proposed, in which the first link is a VLC link while an RIS is deployed in the second RF link. More specifically, a relay first converts the received optical signal into the RF signal blue using either decode-and-forward or amplify-and-forward protocols and then forwards it to the RIS. The RIS alters the electromagnetic response of the impinging waves to provide connectivity between the RF destination and optical source. Based on this model, closed-form expressions for the outage probability and bit error rate (BER) are derived. Also, an asymptotic analysis for the outage probability is presented. Finally, simulation results are provided to verify our analytical results.

In the helicopter hover refuelling system, the emergency breakaway device is an important component to ensure the safety of the helicopter. It requires automatic disconnection and sealing of the pipeline under the tension of 1700–2000 N. The claw type emergency breakaway device was designed, the three-dimensional model is established for dynamic analysis, and the optimal parameters of the breakaway device are obtained by orthogonal experiment. Then, the calculation equation of the braking force is obtained through analysis. The sealing structure of the emergency breakaway device adopts a pair of poppet valves. In order to verify the reliability of the braking force and the seal, the emergency breakaway device is subjected to a water-passing test. The results show that the emergency breakaway device is separated with the required breaking force and the leakage amount is zero after the breakaway.

The rapid growth in nursing demand in P.R.China and the slow increase in the number of qualified nurses has led to a serious increased in workload of existing nurses. Most of their works include heavyweight handling which causes low back pains. Thus a mobile, flexible, comfortable and wearable waist assist device, the Mobile Wearable Waist Assist Robot using pneumatic artificial muscles as power actuators has been developed. The wearer wears the device like a backpack, which can provide wearer the necessary assistance when he lifts loads or performs a static maintenance work, which will reduce the risk of lower back pain. The device is evaluated through an experiment which include three tests - heavy load handling, maximum weight lifting test, and Center of gravity (CoG) trajectory test during static weight lifting. Surface electromyography of erector spinae was recorded during the first test. The iEMG is significantly reduced by about 39% and 27%, respectively (p

In this paper, we mainly solve the adaptive control problem of robot manipulators with uncertain kinematics, dynamics, and actuators parameters, which has been a long-standing, yet unsolved problem in the robotics field, because of the technical difficulties in handling highly coupled effect between control torque and the mentioned uncertainties. To overcome the difficulties, we propose a new Lyapunov-based adaptive control methodology, which effectively fuses the inverse Jacobian technique and the actuator adaptation law, with which the chattering in tracking errors caused by actuator parameter perturbation is well suppressed. It is demonstrated that the asymptotic convergence of all closed-loop signals is guaranteed. Moreover, the effectiveness of our control scheme is illustrated through simulation studies.

Emotion cause detection is mainly to identify the emotion cause with an emotion expression text, which plays a critical role in building NLP applications. This task is much more difficult than other emotion classification and emotion extraction problems. However, most existing methods only focus on partial information to extract the emotion cause. In this study, we present a new approach to combine the emotion word with its synonyms in order to discover the deep and semantic information by enhanced-representation and attention mechanism. Meanwhile, we propose a new mechanism to introduce the hierarchical context behind emotion word information for extracting emotion cause inspired by a convolution operation. Our proposed framework can extract both enhanced represented emotion level features and context level features to better detect the emotion cause. We have conducted extensive experiments on the emotion cause dataset. Experimental results demonstrate the effectiveness of our proposed model, outperforming a number of competitive baselines by at least 3.39% in F1-measure.

In this paper, we present a system for estimating the trajectory of a moving RGB-D camera with applications to building maps of large indoor environments. Unlike the current most researches, we propose a ‘feature model’ based RGB-D visual odometry system for a computationally-constrained mobile platform, where the ‘feature model’ is persistent and dynamically updated from new observations using a Kalman filter. In this paper, we firstly propose a mixture of Gaussians model for the depth random noise estimation, which is used to describe the spatial uncertainty of the feature point cloud. Besides, we also introduce a general depth calibration method to remove systematic errors in the depth readings of the RGB-D camera. We provide comprehensive theoretical and experimental analysis to demonstrate that our model based iterative-closest-point (ICP) algorithm can achieve much higher localization accuracy compared to the conventional ICP. The visual odometry runs at frequencies of 30 Hz or higher, on VGA images, in a single thread on a desktop CPU with no GPU acceleration required. Finally, we examine the problem of place recognition from RGB-D images, in order to form a pose-graph SLAM approach to refining the trajectory and closing loops. We evaluate the effectiveness of the system on using publicly available datasets with ground-truth data. The entire system is available for free and open-source online.

Humans have the ability to perceivekinetic depth effects, i.e., to perceived 3D shapes from 2D projections of rotating 3D objects. This process is based on a variety of visual cues such as lighting and shading effects. However, when such cues are weak or missing, perception can become faulty, as demonstrated by the famous silhouette illusion example of thespinning dancer. Inspired by this, we establish objective and subjective evaluation models of rotated 3D objects by taking their projected 2D images as input. We investigate five different cues: ambient luminance, shading, rotation speed, perspective, and color difference between the objects and background. In the objective evaluation model, we first apply 3D reconstruction algorithms to obtain an objective reconstruction quality metric, and then use quadratic stepwise regression analysis to determine weights of depth cues to represent the reconstruction quality. In the subjective evaluation model, we use a comprehensive user study to reveal correlations with reaction time and accuracy, rotation speed, and perspective. The two evaluation models are generally consistent, and potentially of benefit to inter-disciplinary research into visual perception and 3D reconstruction.

Recent studies have shown that reconfigurable Intelligent surfaces (RiSs) can help In extending the wireless coverage. However, no work has quantitatively analyzed the coverage. In this letter, we present the coverage analysis for an RiS-aided communication system. For comparison, the coverage analysis for relaying systems and direct-link transmission systems are also derived. Furthermore, we provide the analysis for the probability of signal-to-noise ratio (SNR) gain of the RiS over the direct link transmission. Finally, we analyze the delay outage rate of the RiS In achieving ultra-reliable low-latency transmissions. The results show that using the RiS can expand the coverage, Increase the SNR gain, and reduce latency of the links. National Natural Science Foundation of China;Guangzhou Science and Technology Program key projects;Natural Science Foundation of Hunan Province; Department of Education of Guangdong Province; Changsha Science and Technology Project Scopus

Sarcasm detection is a challenging natural language processing task for sentiment analysis. Existing deep learning based sarcasm detection models have not fully considered sentiment semantics, even though sentiment semantics is necessary to improve the performance of sarcasm detection. To deal with the problem, we propose a multi-level memory network using sentiment semantics to capture the features of sarcasm expressions. In our model, we use the first-level memory network to capture sentiment semantics, and use the second-level memory network to capture the contrast between sentiment semantics and the situation in each sentence. Moreover, we use an improved convolutional neural network to improve the memory network in the absence of local information. The experimental results on the Internet Argument Corpus (IAC-V1 and IAC-V2) and Twitter dataset demonstrate the effectiveness of our model.

As a competitive machine learning algorithm, extreme learning machine (ELM), with its simple theory and easy implementation, has been widely used in the field of pattern accuracy. Recently, researchers have proposed related research algorithms to accommodate noise and outlier data. With a proper fuzzy membership function, a fuzzy ELM can effectively reduce the effects of outliers when solving the classification problem. However, how to apply ELM for learning and accuracy in the presence of noise is still an important research topic. A novel fuzzy ELM (ANFELM) technique is proposed in this paper. In the algorithm, the membership degree of the sample is calculated in a feature mapping space instead of the data input space. The algorithm provides good performance in reducing the effects of outliers and significantly improves classification accuracy and generalization. Experiments on UCI datasets and textual datasets show that the proposed algorithm significantly improves the classification capability of ELM and is superior to other algorithms.

Visual structure and syntactic structure are essential in images and texts, respectively. Visual structure depicts both entities in an image and their interactions, whereas syntactic structure in texts can reflect the part-of-speech constraints between adjacent words. Most existing methods either use visual global representation to guide the language model or generate captions without considering the relationships of different entities or adjacent words. Thus, their language models lack relevance in both visual and syntactic structure. To solve this problem, we propose a model that aligns the language model to certain visual structure and also constrains it with a specific part-of-speech template. In addition, most methods exploit the latent relationship between words in a sentence and pre-extracted visual regions in an image yet ignore the effects of unextracted regions on predicted words. We develop a residual attention mechanism to simultaneously focus on the pre-extracted visual objects and unextracted regions in an image. Residual attention is capable of capturing precise regions of an image corresponding to the predicted words considering both the effects of visual objects and unextracted regions. The effectiveness of our entire framework and each proposed module are verified on two classical datasets: MSCOCO and Flickr30k. Our framework is on par with or even better than the state-of-the-art methods and achieves superior performance on COCO captioning Leaderboard.

In this letter, we propose a dual-hop reconfigurable intelligent surface (RIS)-based free space optical and radio frequency (FSO-RF) communication system, where an RIS is utilized to improve the coverage and system performance. Taking both the atmospheric turbulence and pointing errors into consideration, we derive exact closed-form expressions for the outage probability and bit error rate (BER) in terms of the Meijer’s G-function, generalized hypergeometric function, and Marcum Q-function. Furthermore, to obtain the diversity order, an asymptotic outage analysis is also obtained. Finally, the correctness of the analytical results is verified by Monte-Carlo simulation results.

The concrete aging problem has gained more attention in recent years as more bridges and tunnels in the United States lack proper maintenance. Though the Federal Highway Administration requires these public concrete structures to be inspected regularly, on-site manual inspection by human operators is time-consuming and labor-intensive. Conventional inspection approaches for concrete inspection, using RGB image-based thresholding methods, are not able to determine metric information as well as accurate location information for assessed defects for conditions. To address this challenge, we propose a deep neural network &#x0028 DNN &#x0029 based concrete inspection system using a quadrotor flying robot &#x0028 referred to as CityFlyer &#x0029 mounted with an RGB-D camera. The inspection system introduces several novel modules. Firstly, a visual-inertial fusion approach is introduced to perform camera and robot positioning and structure 3D metric reconstruction. The reconstructed map is used to retrieve the location and metric information of the defects. Secondly, we introduce a DNN model, namely AdaNet, to detect concrete spalling and cracking, with the capability of maintaining robustness under various distances between the camera and concrete surface. In order to train the model, we craft a new dataset, i.e., the concrete structure spalling and cracking &#x0028 CSSC &#x0029 dataset, which is released publicly to the research community. Finally, we introduce a 3D semantic mapping method using the annotated framework to reconstruct the concrete structure for visualization. We performed comparative studies and demonstrated that our AdaNet can achieve 8.41 &#x0025 higher detection accuracy than ResNets and VGGs. Moreover, we conducted five field tests, of which three are manual hand-held tests and two are drone-based field tests. These results indicate that our system is capable of performing metric field inspection, and can serve as an effective tool for civil engineers.

Humor as a lubricant for daily life is frequently used in language expressions. It is usually triggered by comparisons of metaphorical scenes or misunderstandings of ambiguous words. Hence, detecting and recognizing the humor implied in text is an interesting and challenging research problem in the field of natural language processing. To understand humor, we discuss incongruity and ambiguity in detail and then propose an internal and external attention neural network (IEANN) for the humor detection task. The IEANN integrates two types of attention mechanisms to capture the incongruity and ambiguity in humor text. Meanwhile, extensive experiments are conducted on two humor datasets to test the effectiveness and robustness of our model. The experimental results show that the proposed model not only achieves state-of-the-art performance but also has better interpretability.

Massive multiple input multiple output (MIMO) systems with hundreds of correlated antennas at base station (BS) are capable of offering abundant spatial resources. When spatial-domain modulation is applied to these systems, the spatial modulation (SM) [1] using only one radio frequency (RF) chain benefits practical implementation, but suffers from significantly reduced multiplexing gain offered by the massive number of antennas. By contrast, the generalized spatial modulation (GSM) [2] allows to simultaneously activate multiple transmit antennas, which improves spatial multiplexing gain, but degrades the achievable error performance, as its design pays no attention to the antenna correlation. In this paper, we propose an antenna grouped spatial modulation (GrSM) scheme, which is capable of circumventing the shortcomings of both the SM and GSM schemes suffered in massive MIMO scenarios. In the proposed GrSM scheme, transmit antennas are partitioned into multiple groups, where the relatively strongly correlated antennas within individual groups are used to implement component SM schemes, while the relatively weakly correlated antennas in different groups are beneficial to obtain multiplexing gain. In order to further improve the spectral efficiency, adaptive modulation (AM) is integrated with GrSM to form the adaptive GrSM (AGrSM). The achievable error and spectral efficiency performance of GrSM and AGrSM systems are investigated based on both mathematical analysis and Monte-Carlo simulations, which are also compared with that of the conventional SM and GSM schemes, when massive MIMO communication scenarios are considered. Our studies and performance results show that GrSM is a promising transmit scheme for massive MIMO, which can outperform both the SM and GSM schemes.

Humor refers to the quality of being amusing. With the development of artificial intelligence, humor recognition is attracting a lot of research attention. Although phonetics and ambiguity have been introduced by previous studies, existing recognition methods still lack suitable feature design for neural networks. In this paper, we illustrate that phonetics structure and ambiguity associated with confusing words need to be learned for their own representations via the neural network. Then, we propose the Phonetics and Ambiguity Comprehension Gated Attention network (PACGA) to learn phonetic structures and semantic representation for humor recognition. The PACGA model can well represent phonetic information and semantic information with ambiguous words, which is of great benefit to humor recognition. Experimental results on two public datasets demonstrate the effectiveness of our model.