Search

As a soft computing method, applying fuzzy cognitive map (FCM) to time series prediction has become a timely issue pursued by numerous researchers. Although many FCM construction methods have emerged, most of them exhibit obvious limitations in weight learning especially for long-term or complex time series. Either the weight calculation is computationally expensive, or it cannot achieve gratifying accuracy. In this paper, a new method for constructing FCM is proposed which extracts concepts from data by exploiting triangular membership function, and the weights of high-order FCM are subtly obtained by transforming the learning problem of FCM into a convex optimization problem with constraints. Since then, FCM with optimized weights is used to represent fuzzy logical relationships of time series and implement prediction further. Fifteen benchmark time series,such as Soybean Price time series, Yahoo stock time series, Condition monitoring of hydraulic systems time series etc. are applied to verify prediction performance of the proposed method. Accordingly, experiment results show that the proposed numerical prediction method of time series is effective and can acquire better prediction accuracy with lower computation time than other recent advanced methods. In addition, the influence of parameters of the method is analyzed individually.

Geiger-mode avalanche photodiode (Gm-APD) array LiDAR has become the current research focus due to its sensitive response, high precision, and easy integration. However, due to the limitations of the fabrication process and manufacturing cost, the images collected by Gm-APD array LiDAR have very serious image low-resolution problems. Since the intensity map and depth map of the target can be obtained simultaneously by relying on single-source data from Gm-APD array LiDAR, we propose a Gm-APD LiDAR single-source data self-guided method. We propose to first perform Gm-APD LiDAR intensity map superresolution, and then use the processed intensity map with the corresponding depth map for guided superresolution. The advantage of this process is that instead of using high-resolution (HR) imaging devices from different domains, it relies only on single-source data from Gm-APD LiDAR to obtain HR depth map, thus eliminating the need for additional image registration work and providing wider applicability. We investigate the feasibility of the proposed single-source data processing method and evaluate our method on the real Gm-APD LiDAR single-source data with an average peak signal-to-noise ratio of 42.21, which better preserves the original distance information of the targets while providing visually sharper outputs.

Electroencephalogram (EEG) signals have emerged as an important tool for emotion research due to their objective reflection of real emotional states. Deep learning-based EEG emotion classification algorithms have made encouraging progress, but existing models struggle with capturing long-range dependence and integrating temporal, frequency, and spatial domain features that limit their classification ability. To address these challenges, this study proposes a Bi-branch Vision Transformer- based EEG emotion recognition model, Bi-ViTNet, that integrates spatial-temporal and spatial-frequency feature representations. Specifically, Bi-ViTNet is composed of spatial-frequency feature extraction branch and spatial-temporal feature extraction branch that fuse spatial-frequency-temporal features in a unified framework. Each branch is composed of Linear Embedding and Transformer Encoder, which is used to extract spatial-frequency features and spatial-temporal features. Finally, fusion and classification are performed by the Fusion and Classification layer. Experiments on SEED and SEED-IV datasets demonstrate that Bi-ViTNet outperforms state-of-the-art baselines.

We present an effective modes theory for studying the strong light-matter interaction and better understanding the non-Markovian dynamics of a quantum emitter (QE) coupled to two-dimensional materials. Specifically, we investigate the spontaneous emission of a V-type QE in close proximity to a MoS$_{2}$ nanodisk. The non-Markovian population dynamics is observed and found linked to the initial states, indicating that the coupled system enters the strong-coupling regime. The effective modes analysis shows that though a dozen of resonant modes appear in the photonic density of states, only a few modes have significant contribution to the population dynamics. Our work demonstrates the effective modes theory as a powerful tool to explore and optimize the quantum states control of QE by two-dimensional materials.

The problem of cooperative arraying to jam is an important part of EW mission planning. Aiming at the problem that multi-objective optimization algorithm is easy to fall into local optimum and converge in three-objective optimization, a multi-aircraft jamming and cooperative arraying method based on improved multi-objective Moth-flame optimization algorithm is proposed. Firstly, the simulation environment is established by using digital elevation map and radar detection model. Then, based on the multi-objective Moth-flame optimization algorithm, the population initialization is completed by using Logistic-Tent chaotic map, which increases the diversity and uniformity of the solution and improves the search ability of the algorithm; Then, the decision factor and Gaussian difference mutation are introduced, which makes the algorithm not only accept the current solution with a certain probability, but also jump out of the current solution and search again according to the disturbance, thus enhancing the search ability of the algorithm; Finally, by comparing with NSGA-II, MOEA/D, MOPSO and NSMFO algorithms on test functions of ZDT and DTLZ series, the performance of the algorithm is verified, and it is proved that multi-objective Moth-flame optimization algorithm is better than other algorithms in both convergence and diversity. In addition, compared with the NSMFO and MOEA/D algorithms in the arraying simulation experiment. The values of the interference power, the width of the route safety zone and the detection area of the radar obtained by the algorithm in this paper, are 117.9 kw, 46 km, and 1727 km2. Compared with the results of the other two algorithms, the effectiveness of interference is improved by 39.8%, 22.8% and 41.9% respectively.

Driving stress is the demand for reserved cognitive space after a driver perceives changes in vehicle, road and environmental factors during driving, which has been proven to affect driving behaviour, interfering with driving safety. Traditional stress prediction relies extensively on psychological data and is limited by the unpopularity of psychological data collection technology, which cannot be applied in daily life on a large scale. In recent years, advances in high-precision visual analysis technology represented by deep learning have laid the foundation for automated and large-scale visual environment analysis. This study proposes a framework for the quantitative analysis of highway driving stress based on multiple vehicle, road, and environmental factors. A dilated residual network model and other methods were used to extract visual environmental indexes. Combined with multisource data such as traffic volume and road design parameters, the LightGBM method was used to construct an expressway driving stress prediction model with high accuracy. The MAE, RMSE and $R^{2}$ values of the proposed model are 0.042, 0.004 and 0.881, respectively, demonstrating the usefulness for scaled and efficient assessment of expressway stress loads. The SHAP method was used to explore the relationship between different influencing factors and driving stress to quantify the mechanism of vehicle, road and environment influences on stress load, and to propose recommendations for highway design and planning from the perspective of reducing stress load. This study provides a new way of thinking to quantitatively investigate the link between multiple road traffic factors and driving stress, providing efficient and large-scale assessment of expressway driving stress, as well as proposing some suggestions for highway design and planning to enhance stress reduction.

Metasurfaces analogues of Fano resonances provide a powerful platform for high sensitivity sensing, nonlinear optics, and light manipulation. However, previous Fano-resonant metasurfaces usually are not compatible with silicon complementary metal-oxide semiconductor circuits due to their hybrid material structures and large non-radiative loss. Herein, we theoretically demonstrate a silicon-on-insulator metasurface (SOIM) enhancing Fano resonances by using a tandem neural network design. Multiple Fano resonances with high Q-factor have been observed in the symmetry-breaking SOIM. The Fano-resonant mechanism of the SOIM is analyzed. Additionally, the spectral features of the Fano-resonant SOIM as a function of the symmetry tuning factor of the double silicon nanobars and the environment refractive index are also investigated. The result shows that the Fano-resonant SOIM as a methanol sensor with a sensitivity of 310 nm/RIU can achieve an overall figure of merit of 195 in the near-infrared spectral regime. The designed Fano-resonant SOIM shows enormous potential applications in highly sensitive sensors and light-matter interaction enhancement.

In the above paper [1] there are errors which we correct here.

For reaching efficient deterministic synchronous Byzantine agreement upon partially connected networks, the traditional broadcast primitive is extended and integrated with a general framework. With this, the Byzantine agreement is extended to fully connected bipartite networks and some bipartite bounded-degree networks. The complexity of the Byzantine agreement is lowered and optimized with the so-called Byzantine-levers under a general system structure. Some bipartite simulation of the butterfly networks and some more refined properties of bipartite bounded-degree networks are also provided for building efficient incomplete Byzantine agreement under the same system structure. It shows that efficient real-world Byzantine agreement systems can be built with the provided algorithms with sufficiently high system assumption coverage. Meanwhile, the proposed bipartite solutions can improve the dependability of the systems in some open, heterogeneous, and even antagonistic environments.

In recent years, more and more deep neural network methods have been used in the forecasting research of multivariate time series. Comparing to the traditional methods such as autoregressive models, methods based on neural networks have achieved superior results. However, the sequence data comes from all aspects, and the data contains long-term and short-term models, which is not a small challenge for data prediction research. We builds a model based on a deep neural network, and uses a convolutional network and a recurrent network in the long-term sequence data input and short-term sequence data input of the model. Convolutional networks are mainly used to extract short-term patterns in sequence data, and recurrent networks are mainly used to extract long-term patterns in sequence data. In addition, in order to solve the problem that the output of the neural network is not sensitive to the change of the input scale, we also added an autoregressive network to the model. When constructing the input of the model, we use the corresponding period characteristics to construct the two input of the model for different datasets: long-term historical data and short-term historical data. Our model has achieved better results than traditional prediction models on four public data sets: out of 32 indicators in 8 methods, 24 optimal results have been achieved.

Video prediction has developed rapidly after the booming of deep learning. As an important part of unsupervised representation learning, it plays an important role in anomalous behavior detection, autonomous driving, video games, and other fields. However, the prediction method based on optical flow estimation is susceptible to brightness change and camera shake, and it is difficult to predict occluded objects. While the prediction method based on pixel generation is difficult to fit ambiguous and complex scenes, which leads to a blurry prediction. In this work, we proposed an end-to-end video prediction framework that combines the optical flow estimation module with the pixel generation module by a learnable mask weight to predict high-fidelity videos. To further improve the prediction effect, we introduce adversarial training to the framework. We introduced a frame discriminator and a sequence discriminator to ensure the consistency of the spatio-temporal distribution of predicted video frames and real video frames. The results of experiments on challenging datasets demonstrate the practicability and effectiveness of our proposed video prediction framework. On the one hand, our proposed framework has achieved an equal quality compared with the current latest model, which requires fewer parameters and has a faster prediction speed. On the other hand, the results of ablation experiments demonstrate the effect of fusing different modules and the effectiveness of adversarial training.

Recent developments in Neural Architecture Search (NAS) resort to training the supernet of a predefined search space with weight sharing to speed up architecture evaluation. These include random search schemes, as well as various schemes based on optimization or reinforcement learning, in particular policy gradient, that aim to optimize a parametric architecture distribution and the shared model weights simultaneously. In this paper, we focus on efficiently exploring the important region of a neural architecture search space with reinforcement learning. We propose Deep Deterministic Architecture Sampling (DDAS) based on deep deterministic policy gradient and the actor-critic framework, to selectively sample important architectures in the supernet for training. Through balancing exploitation and exploration, DDAS is designed to combat the disadvantages of prior random supernet warm-up schemes and optimization schemes. Gradient-based NAS approaches require the execution of multiple short experiments in order to combat the random stochastic nature of gradient descent, while still only producing a single architecture. Contrary to this approach, DDAS employs a reinforcement learning-based agent and focuses on discovering a Pareto frontier containing many architectures over the course of a single experiment requiring 1 GPU day. Experimental results for CIFAR-10 and CIFAR-100 on the DARTS search space show that DDAS can depict in a single search, the accuracy-FLOPs (or model size) Pareto frontier, which outperforms random sampling and search. With a test accuracy of 97.27%, the best architecture found on CIFAR-10 outperforms the original second-order DARTS while using 600M fewer parameters. Additionally, DDAS finds an architecture capable of achieving 82.00% test accuracy on CIFAR-100 while using only 3.14M parameters and outperforming GDAS.

Mode division multiplexing (MDM) silicon photonic integrated circuits (PICs) have been widely developed for achieving high-speed optical interconnects and communications. As an excellent nonlinear optical platform, silicon PICs also receive great attention in applications of optical parametric devices and nonlinear optical signal processing. However, it is still challenging to develop MDM optical parametric devices due to the strong mode dependence of the group velocity dispersion (GVD) in a silicon waveguide. Here, we theoretically design a convex waveguide exhibiting almost the same GVD profiles for quasi-TE0 and quasi-TE1 modes based on the standard fabrication flow of silicon photonic foundries. Specifically, flat GVD curves varying from -1500 ps/nm/km to -1000 ps/nm/km are obtained for the two modes in a convex waveguide within a spectral region of 1.37 μm to 1.75 μm covering from E-band to U-band. The study is expected to open an avenue for exploring unprecedented MDM nonlinear applications.

Visual object tracking is a fundamental problem in computer vision, and has been greatly improved with the rapid development of deep learning. However, existing tracking methods with single model update strategy cannot guarantee the robustness of tracker in complex scenes. In this paper, we innovatively propose a novel real-time long-short-term multi-model based tracking method. For the fusion of long-term and short-term features contain more spatiotemporal information, three models with different update periods are designed to learn the long-term and short-term features to improve the tracking robustness. Besides, the hierarchical feature contain deep convolution features and handcraft features are used to represent the current object, which can further improve the tracking accuracy with richer semantic information. Finally, to solve the inaccurate prediction of object position due to the cosine window in the correlation filter, the bounding-box regression strategy is introduced to optimize the final object position. Extensive experiments on OTB, VOT, TC128, and UAV123 datasets demonstrate that the proposed method performs favorably against state-of-the-art algorithms while running at 24 fps.

The recent success of deep neural networks is attributed in part to large-scale well-labeled training data. However, with the ever-increasing size of modern datasets, combined with the difficulty of obtaining label information, semi-supervised learning (SSL) has become one of the most remarkable issues in data analysis. In this paper, we propose an Incremental Self-Labeling strategy for SSL based on Generative Adversarial Nets (ISL-GAN), which functions by constantly assigning unlabeled data with virtual labels for promoting the training process. Specifically, during the virtual labeling process, we introduce a temporal-based self-labeling strategy for safe and stable data labeling. Then, to dynamically assign more virtual labels to data during the training, we conduct a phased incremental label screening and updating strategy. Finally, to balance the contribution of samples with different loss during the training process, we further introduce the Balance factor Term (BT). The experimental results show that the proposed method gives rise to state-of-the-art semi-supervised learning results for the MNIST, CIFAR-10, and SVHN datasets. Particularly, our model performs well with fewer labeled conditions. With a dataset of only 1,000 labeled CIFAR-10 images with CONV-Large Net, a test error of 11.2% can be achieved, and nearly the same performance with a 3.5% test error can be achieved with both 500 and 1,000 image-labeled SVHN datasets.

The performances of heuristic algorithms are highly dependent on the parameters used, and usually difficult to determine subjectively. Thus, how to balance the relations between the qualities of the solutions and the values of the parameters has been a hot and lasting topic in the field of optimization research. This article presented a statistical method to estimate the efficient parameter values of three alternative neighborhood search techniques of simulated annealing when applied to forest spatial harvest scheduling problems, as an example. The neighborhood search techniques included: the conventional version of simulated annealing (Method1), and the swapping version (Method2) and the changing version (Method3) of 2-element optimization (2-opt) moves. Results indicated that the performances of different neighborhood search strategies highly depended on the problem size, in which the superiorities of Method2 increased from about 10% for smaller cases (400 units) to approximately 80% of larger cases (>3600 units) when compared the objective function values with Method1 and Method3. The efficient parameter values of the cooling rate (CR), the number of iterations per temperature (Ntem), and the number of iterations used for generating initial solution (Nsol) could be estimated using polynomial functions with the number of units, while the initial temperature (IT) should be estimated using exponential function, where the determination coefficient ($R^{2}$ ) of the fitted functions were all larger than 0.60 [except for Nsol and CR of Mehod1 ($R^{2}=0.32$ and 0.42), CR of Method2 ($R^{2}=0.20$ )]. The number of satisfactory solutions all increased linearly ($R^{2} > 0.85$ ) with the number of units, while the solution efficiency decreased linearly ($R^{2} > 0.30$ ). The verifications of two extra grid datasets indicated that the parameter optimization methods were valid.

In frequency-modulated continuous-wave (FMCW) radar interference suppression based on signal reconstruction, the pruned exact linear time (PELT) algorithm is used to detect the time positions of the interference. Due to the uncertain penalty factor of the PELT algorithm, the exactness of the position detection is reduced; thus, the suppression performance is degraded. We propose a PELT algorithm with a known change number (PELT-KCN), where a known change number is used to calculate the optimal penalty factor such that the high accuracy of the algorithm can be guaranteed. After interference recognition, the beat signal is separated into two parts: the undamaged signal and the damaged signal. The former is utilized to restore the latter through an autoregressive (AR) model. In simulations and field experiments, we applied our proposed PELT-KCN algorithm to the interference suppression method and verified its performance. Our method can accurately detect the time positions of interference and effectively improve the signal-to-noise ratio (SNR) of the detected targets.

Nowadays, fieldwork is often accompanied by tight schedules, which tends to strain the shoulder muscles due to high-intensity work. Moreover, it is difficult and stressful for the disabled to drive agricultural machinery. Besides, current artificial intelligence technology could not fully realize tractor autonomous driving because of a high uncertain filed environment and short interruptions of satellite navigation signal shaded by trees. To reduce manual operations, a tractor assistant driving control method was proposed based on the human-machine interface utilizing the electroencephalographic (EEG) signal. First, the EEG signals of the tractor drivers were collected by a low-cost brain-computer interface (BCI), followed by denoising using a wavelet packet. Then the spectral features of EEG signals were calculated and extracted as the input of Recurrent Neural Network (RNN). Finally, the EEG-aided RNN driving model was trained for tractor driving robot control such as straight going, brake, left turn, and right turn operations, which control accuracy was 94.5% and time cost was 0.61 ms. Also, 8 electrodes were selected by the PCA algorithm for the design of a portable EEG controller. And the control accuracy reached 93.1% with the time cost of 0.48 ms. To solve the incomplete driving data set in the actual world because some driving manners may cause dangerous or even death, RNN-TL algorithm was employed by creating the complete driving data in the virtual environment followed by transferring the driving control experience to the actual world with small actual driving data set in the field, which control accuracy was 93.5% and time consumption was 0.48 ms. The experimental results showed the feasibility of the proposed tractor driving control method based on EEG signal combined with RNN-TL deep learning algorithm which can work with the displacement error less than 6.7 mm when the tractor speed is less than 50 km/h.

Traffic flow forecasting is a critical task for urban traffic control and dispatch in the field of transportation, which is characterized by the high nonlinearity and complexity. In this paper, we propose an end-to-end deep learning based dual path framework, i.e., Spatial-Temporal Graph Attention Network (STGAT), for traffic flow forecasting. Specifically, different from previous structure-based approaches, STGAT can be directly generalized to the graph with arbitrary structure. Furthermore, STGAT is capable of handling long temporal sequence by stacking gated temporal convolution layer. The dual path architectures is proposed for taking both potential and existing spatial dependencies into account. By capturing potential spatial dependencies will naturally catch more useful information for forecasting. We design a gated fusion mechanism to combine the outputs from each path. The proposed model can be directly applicable to inductive learning tasks by introducing a graph attention mechanism into spatial-temporal framework, which means our model can be generalized to completely unseen graphs. Moreover, experimental results on two public real-world traffic network datasets, METR-LA and PEMS-BAY, show that our STGAT outperforms the state-of-the-art baselines. Additionally, we demonstrate the proposed model is competent for efficient migration between graphs with different structures.

In data science, how to depict data in the concise and comprehensive way is an important issue. To address the issue, the key is to construct descriptors that are highly interpretable and can be used to reveal the data structure. Information granules, as one important role in the field of granular computing, are entities that can be easily represented and abstracted from data. Therefore, by constructing a series of information granules, the characteristics of data can be captured and described, and the granular description of data is realized. A key part of the granular description of data is to explore the geometric characteristics (locations and shapes) of information granules used to describe data. Since distance measures directly affect the geometric characteristics of the constructed information granules, a comparative study based on three different distance measures is conducted in this paper. From the experimental results based on both synthetic and UCI repository datasets, it can be seen that the information granules constructed in the case where three different distance measures are used show different geometrical shapes, and can describe the data in a concise way. Furthermore, the data structure can be explored more comprehensively by using three distance measures.

In this paper, a simple optical vector quadrature de-multiplexer (QD) scheme is proposed for de-aggregating input 10 Gbaud 16/32/64 quadrature amplitude modulation (QAM) signals into two pulse amplitude modulation (PAM) streams. The proposed QD is based on a single bi-directional degenerate phase-sensitive amplifier (PSA), where the wavelength and the polarization status of the extracted in-phase and quadrature components stay the same as the input signal. Since the proposed QD is realized using PSA based on semiconductor optical amplifier (SOA), it is also possible to realize an on-chip QD system, providing an integrated platform for optical signal processing. Through numerical simulation, the transfer characteristics of the proposed QD system show that the SOA-based PSA has a high phase-sensitive gain extinction ratio of 44.1 dB. The constellations, error vector magnitudes (EVMs), and bit-error rates (BERs) of the data streams after the de-aggregation are numerically investigated to verify the proposed QD scheme. For the 10 Gbaud 16/32/64 QAM signals with an input optical signal-to-noise ratio (OSNR) of 25 dB, the de-aggregated PAM4/PAM6/PAM8 signals show 6.1, 5.6, and 8.2 dB receiver OSNR improvements at the BER of 10−3, respectively. Also, to optimize the performance of the subsystem, the BER performance dependence on the phase difference between two arms in the subsystem is also examined. The simulation results reveal that the proposed QD can accomplish the function of optical vector de-aggregation well for the high-level QAM signals. The proposed QD can be applied to information de-aggregation, format conversion, and direct detection for optical vector signals, which may have great potential values for the flexible optical networks.

With the development of multimedia processing technology, it is becoming much easier to manipulate and tamper with digital video without leaving any visual clues. Because video compression is very common in digital videos, the tamper might employ powerful multimedia deblocking methods to cover up the video tampering traces. Motion JPEG (MJPEG) is one of the most popular video formats, in which each video frame or interlaced field of a digital video sequence is compressed separately as a JPEG image. By splitting the MJPEG video into JPEG image frames, the tamper might employ powerful multimedia deblocking methods to cover up the video tampering traces. To the best our knowledge, there is no existing method for the forensics of deblocking. In this paper, we propose a novel method to detect deblocking, which can automatically learn feature representations based on a deep learning framework. We first train a supervised convolutional neural network (CNN) to learn the hierarchical features of deblocking operations with labeled patches from the training datasets. The first convolutional layer of the CNN serves as the preprocessing module to efficiently obtain the tampering artifacts. Then, we extract the features for an image with the CNN on the basis of a patch by applying a patch-sized sliding-window to scan the whole image. The generated image representation is then condensed by a simple feature fusion technique, i.e., regional pooling, to obtain the final discriminative feature. The experimental results on several public datasets demonstrate the superiority of the proposed scheme.

Background subtraction has been widely used in the detection of a moving object from a still scene. Due to the uncertainty in the classification of the pixels in the foreground and background, we propose a novel fuzzy approach for background subtraction using fuzzy histograms based on fuzzy c-means clustering and the fuzzy nearness degree, called FCFN. In this method, the temporal characteristics of the pixels are described by a fuzzy histogram using the fuzzy c-means algorithm. The segmentation threshold is adaptively calculated according to the distribution of the fuzzy nearness degree of the individual pixel. Fuzzy adaptive background maintenance is adopted in the background update framework. The performance of the FCFN is evaluated against several state-of-the-art methods in the complex dynamic scenes. The experimental results demonstrate that the proposed method doubles the improvements in performance than the classic fuzzy background modeling methods and outperforms most state-of-the-art methods.

In this paper, we present a concept of interval granular fuzzy models and elaborate on their detailed development scheme and performance evaluation. The crux of the underlying design is that the information granules and information granularity which play a pivotal role in human cognitive and decision-making activities are incorporated into existing fuzzy modeling methods to realize system modeling at the level of information granules. The overall development process of interval granular fuzzy model includes two stages. At the first stage, we form input-error data by computing residual error resulting from the well-established numeric model and then establish granular structures positioned in the input space by clustering input-error data and invoking the principle of justifiable information granularity for weighted data. These granular structures discovered in the input space become condition parts of the rules of the developed interval granular fuzzy model. At the second stage, the interval information granules positioned in the error space are directly induced with the aid of granular structures discovered in the input space. These interval information granules describe the range of residual error produced by the numeric model and are exploited to form the conclusion part of the corresponding rules of the developed interval granular fuzzy model. So far, the interval granular fuzzy model is completely constructed, whose rules help compensate the discrepancies of the numeric model. The output of the developed model is an interval information granule showing a range of possible residual error produced by the numeric model. The several performance indices are presented to evaluate the developed interval granular fuzzy model. A series of numeric experiments completed for synthetic data and real-world data coming from the machine learning repository provide a useful insight into the effectiveness of the presented development scheme, reveal the impact of some parameters on the performance of the developed model and demonstrate its advantages.

In this paper, an efficient hybrid reliability analysis (HRA) method and a hybrid reliability-based design optimization (HRBDO) approach are proposed for realistic complex engineering structures with random and interval uncertainties. First, the HRBDO model for complex engineering structures is constructed with its objective and performance functions described as the implicit functions of design variables and random and interval parameters. Then, an efficient HRA method based on adaptive step size (ASS-HRA) is put forward to calculate the minimum reliability of the structure's performance function under the influences of both random and interval uncertainties, the computational efficiency and accuracy of which are verified by a benchmark test. Subsequently, an efficient HRBDO approach integrating the proposed ASS-HRA method with the polynomial response surface model (PRSM) is developed for solving the HRBDO problems of complex engineering structures, the effectiveness of which is demonstrated by a numerical example. Finally, the HRBDO of a high-speed press slider demonstrates the efficiency, effectiveness, and versatility of the proposed HRBDO approach based on the ASS-HRA in the design of realistic complex engineering structures.

We propose a weighted Ip minimization method for downlink channel estimation in frequency division duplexing massive multiple-input multiple-output (MIMO) systems. The proposed algorithm involves two stages, in which it first diagnoses the downlink supports by utilizing the channel sparsity in angular domain and angular reciprocity for uplink and downlink channels. In stage two, a weighted Ip minimization algorithm based on the diagnosed supports is used for downlink channel estimation. The diagnosed supports are used for generating the weighting matrix in the weighted Ip minimization. The restricted isometry property (RIP)-based guarantees and upper bound of the recovery error are derived. Our analytical results have the universal forms for the Ip(0 1 minimization, general I1 minimization, joint orthogonal matching pursuit, and simultaneous orthogonal matching pursuit in the medium and high signal-to-noise-rate regions.

We report a rapid and precise spectral determination approach based on visually patterned integrated narrow bandpass filters. The wavelength of light can be determined accurately even by naked eyes in the aid of the device. The experimental results show that both the light with single wavelength and a spectral range can be distinguished rapidly and precisely by naked eyes with spectral resolution higher than 7 nm, which can be even higher than 1 nm with the intended design. It is a very robust and stable way to keep the filtering wavelengths constant and remarkably improves the stability superior to conventional methods. It can be used for high-throughput identification of light source and food safety detection related applications.

Illumination-insensitive image representation is a great challenge in the computer vision field. Illumination variations considerably obstruct the effectiveness of image feature extraction. In this paper, we present a novel and generalized learning framework for illumination-insensitive image representation, which can learn the discriminative features through maximizing the inter-difference and minimizing intra-difference of the images with boosting. Particularly, we enhance the discriminative capacity of illumination-insensitive image representation in three aspects. First, we learn a subset of different synergistic Weber excitation patterns (SWEP) to generate the dominant SWEP (DSWEP) and DSWEP codebook for exploring optimal illumination-insensitive patterns. Second, a compact DSWEP (C-DSWEP) is learned with a boosted set of weight to generate C-DSWEP codebook. Discriminative learning is aimed at robustness and compactness. Third, the discriminative histogram learning model is established for encoding CDSEP to further improve the discriminative ability and reduce redundancy. The extensive experiments on CMUPIE, FERET, Yale B, Yale B ext., LFW, and PhoTex databases have highlighted the superiority and the robustness of our method compared with some other state-of-the-art methods.

Secret sharing is one of the most important cryptographic primitives used for data outsourcing. The threshold-based secret sharing scheme is one of the well-known secret sharing schemes in cryptography. A single secret sharing scheme has low efficiency and multiple secret sharing schemes can greatly improve the efficiency of secret sharing. We present an efficient verifiable threshold multi-secret sharing scheme with different stages in which multiple secrets are shared among participants during one sharing process. In such a way, some qualified subsets of participants can collectively recover these secrets. The verifiable nature means that the participant can verify his/her own share. Our proposed scheme can share multiple secrets and each secret has its own threshold access structure. Moreover, each participant only keeps one share, and however they can reconstruct all secrets in recovery stages. This scheme is a multi-stage-use secret sharing scheme, in which the size of the participant’s share is the same as that of each secret and the amount of information disclosed reaches the optimal lower bound. Compared with the previous verifiable $(t,n)$ -threshold multi-secret sharing scheme, this scheme is of stronger computational security and practicability.

As there is an increasing trend to deploy geographically distributed (geo-distributed) cloud datacenters (DCs), the scheduling of data-oriented tasks in such cloud DC systems becomes an appealing research topic. Specifically, it is challenging to achieve the distributed online scheduling that can handle the tasks' acceptance, data-transfers, and processing jointly and efficiently. In this paper, by considering the store-and-forward and anycast schemes, we formulate an optimization problem to maximize the time-average profit from serving data-oriented tasks in a cloud DC system and then leverage the Lyapunov optimization techniques to propose an efficient scheduling algorithm, i.e., GlobalAny. We also extend the proposed algorithm by designing a data-transfer acceleration scheme to reduce the data-transfer latency. Extensive simulations verify that our algorithms can maximize the time-average profit in a distributed online manner. The results also indicate that GlobalAny and GlobalAny_Ext (i.e., GlobalAny with data-transfer acceleration) outperform several existing algorithms in terms of both time-average profit and computation time.

The text presented in videos contains important information for content analysis, indexing, and retrieval of videos. The key technique for extracting this information is to find, verify, and recognize video text in various languages and fonts against complex backgrounds. In this paper, we propose a novel method that combines a corner response feature map and transferred deep convolutional neural networks for detecting and recognizing video text. First, we use a corner response feature map to detect candidate text regions with a high recall. Next, we partition the candidate text regions into candidate text lines by projection analysis using two alternative methods. We then construct classification networks transferred from VGG16, ResNet50, and InceptionV3 to eliminate false positives. Finally, we develop a novel fuzzy c-means clustering-based separation algorithm to obtain a clean text layer from complex backgrounds so that the text is correctly recognized by commercial optical character recognition software. The proposed method is robust and has good performance on video text detection and recognition, which was evaluated on three publicly available test data sets and on the high-resolution test data set we constructed.

A new partitioned stator hybrid excitation (HE) doubly salient machine with permanent magnets (PMs) sandwiched between adjacent inner stator tooth tips and with armature and field windings housed in outer and inner stators, respectively, is proposed in this paper. In comparison of the conventional HE machines with single stator, the space utilization is significantly boosted. So more windings and PMs can be employed to obtain increased torque density and improved flux regulation capability. In order to investigate the effect of inner stator yoke, which contributes significant influence to electromagnetic performances, the equivalent magnetic circuit method is employed. Moreover, the multi-level design optimization method is adopted to optimize the key parameters efficiently. Notably, the rated operating point with excitations of PMs and field current of 1 A is defined according to its efficiency. Then, the electromagnetic performances at different operating points are analyzed and compared based on the 2-D finite-element analysis. The results reveal that the output torque is increased at new rated condition due to the flux-enhancing current, and the flux-weakening capability is also improved, so that a wider constant power speed range can be achieved because of the defined rated operating point.

In this paper, a simple self-homodyne detection scheme is proposed and experimentally demonstrated for applications in optical access networks. Unmodulated pilot subframes are periodically inserted and interleaved with data subframes to form an orthogonal frequency-division multiplexing (OFDM) signal. Owing to the coherence between the embedded pilot and data subframes, a Mach-Zehnder delay interferometer with a free spectral range equal to the frequency of the subframe is deployed for the self-homodyne detection of the OFDM signal. It saves the local oscillator and optical hybrid, which are usually used in a conventional coherent receiver, thus reducing the hardware complexity and implementation cost. Meanwhile, the digital signal processing (DSP) for the self-homodyne detection is free from carrier-frequency-offset compensation and carrier-phase estimation, reducing the complexity, power consumption, and latency of the system. It also allows the use of a low-cost laser source as the source for the downstream signal. The proposed scheme provides a cost-effective and energy-efficient downstream solution for optical access networks, owing to the hardware saving and complexity reduction in DSP. A 10-Gb/s OFDM downstream transmission over a 20-km standard single-mode fiber is experimentally demonstrated with an error-free operation, using both a 10-MHz distributed feedback laser and a 100-kHz external cavity laser as downstream laser sources.

Wireless sensor networks (WSNs) are currently being used for monitoring and control in smart grids. To ensure the quality of service (QoS) requirements of smart grid applications, WSNs need to provide specific reliability guarantees. Real-time link quality estimation (LQE) is essential for improving the reliability of WSN protocols. However, many state-of-the-art LQE methods produce numerical estimates that are suitable neither for describing the dynamic random features of radio links nor for determining whether the reliability satisfies the requirements of smart grid communication standards. This paper proposes a wavelet-neural-network-based LQE (WNN-LQE) algorithm that closes the gap between the QoS requirements of smart grids and the features of radio links by estimating the probability-guaranteed limits on the packet reception ratio (PRR). In our algorithm, the signal-to-noise ratio (SNR) is used as the link quality metric. The SNR is approximately decomposed into two components: a time-varying nonlinear part and a non-stationary random part. Each component is separately processed before it is input into the WNN model. The probability-guaranteed limits on the SNR are obtained from the WNN-LQE algorithm and are then transformed into estimated limits on the PRR via the mapping function between the SNR and PRR. Comparative experimental results are presented to demonstrate the validity and effectiveness of the proposed LQE algorithm.