Search

Sleep Apnea (SA) is a prevalent sleep disorder with multifaceted etiologies that can have severe consequences for patients. Diagnosing SA traditionally relies on the in-laboratory polysomnogram (PSG), which records various human physiological activities overnight. SA diagnosis involves manual scoring by qualified physicians. Traditional machine learning methods for SA detection depend on hand-crafted features, making feature selection pivotal for downstream classification tasks. In recent years, deep learning has gained popularity in SA detection due to its capability for automatic feature extraction and superior classification accuracy. This study introduces a Deep Attention Network with Multi-Temporal Information Fusion (DAN-MTIF) for SA detection using single-lead electrocardiogram (ECG) signals. This framework utilizes three 1D convolutional neural network (CNN) blocks to extract features from R-R intervals and R-peak amplitudes using segments of varying lengths. Recognizing that features derived from different temporal scales vary in their contribution to classification, we integrate a multi-head attention module with a self-attention mechanism to learn the weights for each feature vector. Comprehensive experiments and comparisons between two paradigms of classical machine learning approaches and deep learning approaches are conducted. Our experiment results demonstrate that (1) compared with benchmark methods, the proposed DAN-MTIF exhibits excellent performance with 0.9106 accuracy, 0.9396 precision, 0.8470 sensitivity, 0.9588 specificity, and 0.8909 $F_{1}$ score at per-segment level; (2) DAN-MTIF can effectively extract features with a higher degree of discrimination from ECG segments of multiple timescales than those with a single time scale, ensuring a better SA detection performance; (3) the overall performance of deep learning methods is better than the classical machine learning algorithms, highlighting the superior performance of deep learning approaches for SA detection.

The process of reconstructing underlying cortical and subcortical electrical activities from Electroencephalography (EEG) or Magnetoencephalography (MEG) recordings is called Electrophysiological Source Imaging (ESI). Given the complementarity between EEG and MEG in measuring radial and tangential cortical sources, combined EEG/MEG is considered beneficial in improving the reconstruction performance of ESI algorithms. Traditional algorithms mainly emphasize incorporating predesigned neurophysiological priors to solve the ESI problem. Deep learning frameworks aim to directly learn the mapping from scalp EEG/MEG measurements to the underlying brain source activities in a data-driven manner, demonstrating superior performance compared to traditional methods. However, most of the existing deep learning approaches for the ESI problem are performed on a single modality of EEG or MEG, meaning the complementarity of these two modalities has not been fully utilized. How to fuse the EEG and MEG in a more principled manner under the deep learning paradigm remains a challenging question. This study develops a Multi-Modal Deep Fusion (MMDF) framework using Attention Neural Networks (ANN) to fully leverage the complementary information between EEG and MEG for solving the ESI inverse problem, which is termed as MMDF-ANN. Specifically, our proposed brain source imaging approach consists of four phases, including feature extraction, weight generation, deep feature fusion, and source mapping. Our experimental results on both synthetic dataset and real dataset demonstrated that using a fusion of EEG and MEG can significantly improve the source localization accuracy compared to using a single-modality of EEG or MEG. Compared to the benchmark algorithms, MMDF-ANN demonstrated good stability when reconstructing sources with extended activation areas and situations of EEG/MEG measurements with a low signal-to-noise ratio.

The presence of natural ambient noise interferes with the system for locating and identifying underwater targets. This paper suggests that a Dual-Path Transformation Network (DPTN) reduces ambient noise in underwater acoustic signals. First, the input acoustic signals’ higher-order non-linear features are extracted using a multi-scale convolutional encoder neural network. Second, sub-vectors with the same length are created according to the time dimension from the higher-order non-linear features. The sub-vectors are stitched together to form a three-dimensional tensor. Third, a neural network transformer based on the feed-forward network is constructed. Further, to capture long-term series features and separate the target signal from the noisy signals, the three-dimensional tensor is used as the input of the transformer-based masking network. Finally, overlap-add and transpose are used to obtain discernible target signals. The experimental results verify the effectiveness of the proposed underwater acoustic signal denoising algorithm and demonstrate that the proposed DPRN method can obtain higher output signal-to-noise ratio (SNR) and the scale-invariant signal-to-noise ratio (SI-SNR) compared with the other classical algorithms.

Tuberculosis is the leading cause of death due to a single infection prior to the COVID-19 pandemic. Screening of tuberculosis patients in a large population is of paramount importance for disease treatment and control, especially with an economical, accurate, and easy-to-operate method. Based on cough sounds, we proposed DMRNet to distinguish between patients with tuberculosis, other respiratory diseases, and healthy individuals. DMRNet comprises four convolutional blocks and six identification blocks and incorporates dynamic convolution into the first three convolutional blocks to promote feature extraction. After the second and third dynamic convolutions, a polarized self-attention mechanism was added to reduce the information loss caused by the dimensionality reduction. Finally, a multihead self-attention layer is added to the fourth convolutional block and the last three identification blocks to enhance the aggregation of global information. Using a dataset with 1323 cough sound fragments, the results achieved the accuracy, sensitivity, and specificity of tuberculosis screening were 94.32%, 97.73%, and 99.43%, respectively. Compared with reported studies, the proposed model demonstrated better accuracy and reliability. Cough sound-based DMRNet analysis is a promising method for tuberculosis screening, especially in densely populated areas. Owing to its convenience, low equipment requirements, and low cost, it is expected to become an effective tool for community tuberculosis screening.

Video super-resolution technology enhances the display quality of videos by obtaining high-resolution videos from low-resolution videos. Unlike single-image super-resolution, utilizing information between adjacent video frames is crucial in video super-resolution. To improve the performance of video super-resolution reconstruction, a model combining deformable 3D convolution and iterative back projection is proposed to fully exploit the temporal-spatial correlation of video frames. The model takes multiple consecutive video frames as input and outputs the super-resolution reconstruction of the middle frame, including three modules: multi-scale feature extraction, feature fusion, and high-resolution reconstruction. Firstly, multi-scale 3D convolution is used for preliminary feature extraction. Then, deformable 3D convolution and iterative back projection are combined for feature fusion. Finally, multiple residual dense blocks and sub-pixel convolution are used for high-resolution reconstruction, and global residual connections are utilized to obtain the reconstructed high-resolution video. Experimental results on the Vid4 dataset demonstrate that compared to existing methods, this method can effectively improve the peak signal-to-noise ratio and structural similarity performance and achieve better visual effects with 4x super-resolution magnification.

Ion electric propulsion systems have been widely used in satellite station keeping, attitude control, and orbit transfer due to their advantages of high specific impulse, good flexibility, and high reliability. The electric breakdown might happen at grids or across isolators due to the disturbing effects of the power source or environment. The electric breakdown can induce spurious effects in the power processing unit (PPU) circuit from the return currents and voltage drops, which could be several orders of magnitude higher than the impulses at normal working conditions. When the electric breakdown happens, it might result in serious electromagnetic interference and damage to the electric and electronic equipment in the spacecraft. This paper provides a comprehensive summary of the various factors that can cause electric breakdown in an ion electric propulsion system, such as environmental factors, manufacturing considerations, and operational conditions of the thruster. Furthermore, a comprehensive analysis of the impact of electric breakdown on the ion thruster, electric propulsion system, and spacecraft mission is explored. The effectiveness of different grounding configurations in safeguarding against ion thruster electric breakdown has also been evaluated. Based on the analysis and discussion, the control and protection strategies are proposed for the ion electric propulsion system electric breakdown problems.

The closed-form solution of each rotation joint angle is the most direct and convenient method for humanoid robot arms to imitate the real-time motion of the human arm. Occasionally, the human structure differs from that of a robot, and sensor information is inadequate. Therefore, it is difficult to obtain a closed-form solution for the inverse kinematics (IK) of a robot when using the Denavit–Hartenberg (D–H) method or screw theory alone. This study proposes combining the screw theory and the D–H method to solve the closed-form solution of the IK of a robotic arm with two unique structures: the elbow joint offset and the non-vertical state at the shoulder joint of the robotic arm. The proposed combination method first solves the elbow joint angle using screw theory. Then, the other rotation joint angles are solved using the D–H method. With the proposed method, human arm motion is successfully mapped to the corresponding joint of the robotic arm, and the closed-form solution allows the robotic arm to imitate human arm motion in real–time. Simulation results show that IK can be used to obtain high–precision angles and positions. Experiments show that the robotic arm has an accurate imitation performance, the accuracy of the $z$ -axis of the manipulator is significantly improved by 73.6%, and that of the $x$ -axis and $y$ -axis is also improved by 37.36% and 21.89%, respectively.

The inverse problem of synthetic aperture interferometric radiometers (SAIRs) aims at retrieving the brightness temperature map from the visibility function samples. To efficiently obtain an accurate solution, an accelerated iterative regularization technique is proposed to reconstruct the SAIR brightness temperature map. The acceleration technique modifies the conventional least square function using a negative penalty term to speed up the initial iterations. A series of decreasing coefficients must be tuned to prevent noise amplification in subsequent iterations. Several numerical experiments were performed on one-dimensional (1-D) and (2-D) SAIR systems. The experiment results indicate that the accelerated Landweber method reduces the number of iteration steps by more than 50% and effectively improves the computational speed without reducing the reconstruction accuracy compared to the conventional Landweber method.

Monitoring the depth of unconsciousness during anesthesia is beneficial in both clinical settings and neuroscience investigations to understand brain mechanisms. Electroencephalogram (EEG) has been used as an objective means of characterizing brain altered arousal and/or cognition states induced by anesthetics in real-time. Different general anesthetics affect cerebral electrical activities in different ways. However, the performance of conventional machine learning models on EEG data is unsatisfactory due to the low Signal to Noise Ratio (SNR) in the EEG signals, especially in the office-based anesthesia EEG setting. Deep learning models have been used widely in the field of Brain Computer Interface (BCI) to perform classification and pattern recognition tasks due to their capability of good generalization and handling noises. Compared to other BCI applications, where deep learning has demonstrated encouraging results, the deep learning approach for classifying different brain consciousness states under anesthesia has been much less investigated. In this paper, we propose a new framework based on meta-learning using deep neural networks, named Anes-MetaNet, to classify brain states under anesthetics. The Anes-MetaNet is composed of Convolutional Neural Networks (CNN) to extract power spectrum features, and a time consequence model based on Long Short-Term Memory (LSTM) networks to capture the temporal dependencies, and a meta-learning framework to handle large cross-subject variability. We use a multi-stage training paradigm to improve the performance, which is justified by visualizing the high-level feature mapping. Experiments on the office-based anesthesia EEG dataset demonstrate the effectiveness of our proposed Anes-MetaNet by comparison of existing methods.

The continuous development trend of expressways urgently needs expressway sensing, which is required to be able to perceive unknown risk factors to ensure a safe and stable transportation environment as well as effectively and continuously monitor the safe operation of expressways. Using unconscious payment can greatly improve traffic efficiency and meet the development direction of intelligent transportation. However, charging errors occur when using unconscious payment, primarily because the expressway network fails to realize national networking. Most of the literature focuses on two aspects of sensing, transmission and computing, while research on the overall network architecture is lacking. A ground network cannot solve the network coverage problem in remote areas. Therefore, the basic infrastructure needed to realize expressway sensing includes both aerial and ground nodes, such as cars, roadside equipment, unmanned aerial vehicles, airships, and satellites. This network communication architecture requires effective management of infrastructure and support of various applications. The software-defined network has strong flexibility; thus, in this paper, we propose a software-defined space-ground-vehicle integrated network (SD-SGVIN) architecture that includes three layers: physical facility, logic control, and application. The SD-SGVIN can separate the highway sensing applications from the physical entity. All kinds of applications can acquire, transmit, and process vehicle data through a shared physical infrastructure. We discuss the design, implementation, and advantages of the proposed network structure, hoping to provide a new direction for the research and promotion of expressway sensing and unconscious payment.

All metal sensing based on complementary H-shaped arrays (HSAs) is numerically and experimentally investigated at terahertz frequencies. The symmetry-protected bound states in the continuum (BICs) are activated by adjusting one arm's length of the H-shaped, enabling high-Q quasi-BICs. The Q-factor of quasi-BICs is modulated in terms of structural parameters and metal conductivity. With optimal structural parameters of copper HSAs, the proximity sensing performances are studied as a function of the film thickness and the refractive index. High sensitivity of 151 GHz/RIU and FOM of 18.7 are obtained, respectively, which are high compared with the previous metamaterials with planar substrates. In addition, the HSAs are even sensitive to the ultrathin graphene film. These findings indicate that such HSAs-based sensor has great potential for practical applications in chemical and biomolecular fields.

Representativeness error caused by scale transformation (REST) is an intrinsic property of data assimilation, as assimilating new observations likely involves the fusion of multisource and multiscale data. Earlier studies focused on specific cases and failed to obtain a general concept. This study attempts to achieve a further understanding of REST in both theory and practice. Based on scale-related definitions and formulations, the statistical RESTs of observation errors and analysis errors are deduced in stochastic ensemble data assimilation. Experiments based on ensemble Kalman filter are conducted to validate the interpretability of the proposed formulations. A synthetic experiment uses the stochastic Lorenz model as the forecasting operator, and a real-world experiment employs a simple biosphere model as the forecasting operator and uses a series of mixed ground-based and remote sensing soil moisture observations. The results confirm that REST should be proportional to the scale difference when assimilating direct observations and both system states and observations are homogeneous processes. Due to the nonlinearity in modeling, assimilation, and scale transformation, increasing RESTs are found if the scale of the observation is much larger than that of the state space, or multiscale observations are added into the assimilation system. Quantifying REST improves the understanding of uncertainty in data assimilation, but further studies on REST are required in both theory and practice, for example, REST correlates with other errors in forcing, parameters, and models, and introduces an observation operator to assimilate indirect observations.

Complex-valued Magnetic Resonance Imaging (MRI) is widely used in clinical diagnosis. The magnitude images are mainly used for structure visualization, and the phase images reveal tissue properties, such as magnetic susceptibility and fluid flow information. While MRI is slow, compressed sensing (CS) can be used to reconstruct the accelerated acquisitions. Current CS-based MRI reconstruction algorithms mostly focus on magnitude image recovery, with less attention to the recovery of phase images. In this paper, we propose a novel CS algorithm to simultaneously recover the magnitude and phase MR images based on the sparsity of the trigonometric function. The CS method requires a sparse representation of the original images, and it is observed the trigonometric functions of phase images in the Wavelet domain promote sparsity. Therefore, rather than transforming the phase images directly into the Wavelet domain, we calculate the sine and cosine of the phase images whose Wavelet transforms are set as the L1-norm regularization term. The combination of the dual trigonometric functions captures a unique, faithful four-quadrant phase information, which also improves the reconstructed magnitude images through an alternating optimization procedure. Reconstructions of simulated images and in vivo images are studied, and both show the superiority of the proposed method over compared phase recovery algorithms.

In this paper, we investigate the mobile node localization problem in an underwater positioning system using multiple surface beacons, where time synchronization between the surface beacons and underwater mobile nodes is not required. First, we propose a Time-Difference-of-Arrival-based (TDoA-based) node localization method. In the proposed method, as the underwater node receives the positioning messages from more than or equal to three surface beacons belonging to the same grid at one epoch, it calculates the location by itself using the information of received positioning messages and the corresponding arrival times. To deal with the problem that the number of received positioning messages at one epoch by the underwater node is not enough for the location calculation, we present the modified TDoA- based node localization method with buffered messages and the modified TDoA-based node localization method with motion information. The effectiveness of proposed TDoA-based node localization method and two modified methods is validated by simulation results and sea experiment data analytical results. The robustness of the TDoA-based node localization method is improved by two modified methods. Moreover, using the proposed TDoA-based node localization method and two modified methods, the underwater node can localize itself in low energy consumption and passive mode.

Grid-side electrochemical battery energy storage systems (BESS) have been increasingly deployed as a fast and flexible solution to promoting renewable energy resources penetration. However, high investment cost and revenue risk greatly restrict its grid-scale applications. As one of the key factors that affect investment cost, the cycle life of battery heavily depends on its charging/discharging actions during the operation, particularly in the presence of uncertain renewable generation. In this context, it is necessary to consider the operation-dependent cycle life of batteries in optimal BESS sizing, which imposes great challenges to the modeling and solving of the planning problems. In this paper, we propose a novel two-level optimal sizing model for grid-scale BESS, considering its operation under uncertainties induced by volatile wind generation. In the lower level, a long-term chronological operation simulation of BESS is processed with an accurate cycle life model of batteries; in the upper level, marginal economic utility analysis and BESS size reforming are conducted to approach the optimal size of BESS. An iterative algorithm is designed to solve the model effectively. The proposed method is verified on a modified IEEE RTS-24 system and a real provincial power grid of China.

The dynamic behaviors of permanent magnetic synchronous motors considering unbalanced magnetic pull have been extensively investigated. However, the harmonics and cogging effects are ignored and the stator and the rotor are assumed to be smooth cylinders. In this research, an analytical expression of unbalanced magnetic pull considering harmonics and cogging effects is obtained. Based on the model, the motion equations of the rotor system are established. The natural frequency and frequency-response are analyzed by average method, and stability of periodic solution is investigated by Routh-Hurwitz criterion. The analysis indicates that the natural frequency is closely related to stator slot parameters. The frequency-response shows that the amplitude varies between the two static equilibrium points. Besides, when the ratio of slot width to stator slot pitch is close to the peak value, the amplitude tends to be increasing obviously. It is useful for the design of permanent magnet synchronous motors with high stability and low vibration.

The localized faults of bearings often produce a series of periodic impacts. However, how to extract these repetitive transients from the signals with strong noise interference is still a challenging problem. In this paper, a new method called collaborative double difference sparse regularization (CDDSR) is proposed for bearing fault detection. To be specific, the first and second-order difference matrices are integrated into sparse regularization term, and the differential sparsity and denoising effect of fault signal are enhanced by collaboration. According to the time domain impulse characteristics of fault signal, the sparsity of signal itself will also be considered. Based on the majorization-minimization (MM) algorithm, the objective optimization model can be solved quickly. Furthermore, the selection of regularization parameters is deeply studied, and an adaptive parameter selection strategy is given. The performance of CDDSR is verified through simulation analysis and two experimental cases, and the ability of CDDSR to extract fault features is further evaluated by quantitative index. Results demonstrate its superiority in eliminating noise interference and extracting periodic impulses in comparison to other state-of-the-art methods.

The presence of haze could seriously contaminate the observations of optical satellite imagery. Haze not only significantly affects the visual interpretation but also reduces the accuracy of map products. In this article, a semiphysical approach is proposed to reduce the haze effects for Landsat image. The proposed approach is based on the physical model of radiative transfer theory and the presence of dark objects. As the depth map of satellite remotely sensed image is almost a constant value, the coarse transmission map of atmosphere is estimated by the haze thickness, other than the scene depth map. The derived coarse transmission is utilized to correct the color shift induced by airlight. For haze veiled textural information, the guided-filter based approach is adopted to refine the coarse transmission map to restore the textural information. Experiments are conducted upon the images acquired by Landsat at different dates and spatial locations. The visual interpretation upon the dehazed results suggests that the proposed approach could generate visually promising results and preserve spectral properties of land surfaces. Moreover, it also performs favorably against several state-of-the-art deep learning based methods and a classic algorithm. The results of quantitative assessments demonstrate that the developed approach could enhance the information of Landsat scene with minimum spectrum changes. With the visual and quantitative assessments, we can conclude that the proposed approach is a promising solution for Landsat image dehazing. It is expected to improve the data quality of hazy scene and expand the usability of Landsat data.

Simultaneously high quality-factor (Q) and transmission (T) are highly desired in various optical, photonic, and optoelectronic applications such as filters, sensors, photodetectors and lasers. However, a trade-off between high Q and high T exists widely in optical systems thanks to the different physics triggers underneath. Here, as an example, we experimentally demonstrate a Bragg filter composed of niobium pentoxide (Nb$_2$O$_5$) and silica (SiO$_2$) stacks which enable high Q of 183 and high T of 91.3%. Balancing dissipation and radiation rate of the optical system is crucial to the performance of the device, which is validated by modulating the absorption of material (Nb$_2$O$_5$) and the number of stacks. Further, with the principle the tunable Bragg filter is able to work in a similar way at optical wavelengths, i.e., maintaining almost unchanged FWHM (full width at half-maximum) and T values. We believe our work offers an efficient strategy for achieving high Q and T optical systems to meet diversified application requests.

Objective: Chronic diseases have become the most prevalent and costly health conditions in the healthcare industry, deteriorating the quality of life, adversely affecting the work productivity, and costing astounding medical resources. However, few studies have been conducted on the predictive analysis of multiple chronic conditions (MCC) based on the working population. Results: Seven machine learning algorithms are used to support the decision making of healthcare practitioner on the risk of MCC. The models were developed and validated using checkup data from 451,425 working population collected by the healthcare providers. Our result shows that all proposed models achieved satisfactory performance, with the AUC values ranging from 0.826 to 0.850. Among the seven predictive models, the gradient boosting tree model outperformed other models, achieving an AUC of 0.850. Conclusions: Our risk prediction model shows great promise in automating real-time diagnosis, supporting healthcare practitioners to target high-risk individuals efficiently, and helping healthcare practitioners tailor proactive strategies to prevent the onset or delay the progression of the chronic diseases.

Compatibility or competition is a key strategic decision for multihoming software developers. Compatibility is a strategy by which companies set up their software to be more compatible with their competitors’ products, while competition is an opposite strategy by which the companies set up their software without any compatibility. We study the strategic decision of a multihoming software company that competes with two firms on different operating systems, (such as Windows and Mac). By formalizing a game-theoretic model to capture the motivation of a multihoming software developer, we find that the compatibility strategy is mainly incentivized by the trade-off between the spillover effect and the compatibility cost. We then solve for reasonable software prices and compatibility level given the market potential from the software firms’ perspective. Several managerial guidelines are obtained to determine the optimal decision in the software development economy. Interestingly, when both the spillover effect and the compatibility cost are low enough, the multihoming software firm still chooses the competing strategy. Finally, by reporting the numerical analysis, we verify the effectiveness of the theoretical results derived in our model.

Mental workload has been widely estimated based on electroencephalography (EEG) in the frequency domain. However, simple frequency features are not entirely accurate indicators of the cognitive load because surface EEG signals are weak, nonstationary and randomness. We hypothesize that graph methods, which analyse the relationship between each point and other points of the EEG signals, may provide a more precise identification of the mental load. To investigate this hypothesis, we aim to identify the optimum graph features from 14 channel EEG recordings (sampling rate = 128 Hz) in order to detect the high cognitive load related to multitasking. Three graph features: mean degree $\overline {d}$ , clustering coefficient $\overline {c}$ , and degree distribution $p(k)$ , are extracted from 48 subjects EEG records. Each experimental subject conducts two tasks: without tasks and with a simultaneous capacity task, respectively. After the experiment is completed, the feeling of the subject with the cognitive load tags in three types: low load, medium load, and heavy load. The optimal features of these three levels of the subject sensation and two types of cognitive load in different tasks are selected on the basis of statistical analysis. Then all graph features are forwarded into a support vector machine (SVM) and a decision tree to conduct objective scoring classification and a three subjective rating classification, respectively. Based on the present results,channels O2, T8, FC6, F8, and AF4 are considered optimal for a more efficiently estimation of the cognitive load. $\overline {c}$ associated with F8 and T8 during low cognitive load is significantly lower than those associated with high cognitive load (p < 0.001). Using three graph features, the accuracy of identifying two types of mental load is 89.6%. Current findings suggest that the mental workload associated with multi-tasks can be accurately assessed using the graph approaches to EEG data.

Aligning and balancing the marginal and conditional feature distributions are two critical procedures for unsupervised domain adaptation (UDA) problems. However, existing methods usually consider the former while ignoring the latter. To improve the problems of instability and imbalance, we propose the Adaptative Joint Distribution Adaptation Network (AJDAN) by analyzing the multi-modal interactions between the two types of distributions and adding a self-learning network to simultaneously balance them. Furthermore, we give higher weights to samples that are far away from the domain boundary (easy-to-classify samples) using Strong Binary Cross-Entropy (SBCE). The strong alignment strategy is adjustable and allows the network to better train easy-to-classify samples than traditional Binary Cross-Entropy (BCE) in various scenarios. The experiment shows that AJDAN with SBCE (AJDAN+S) provides an average of 68.3% accuracy on the Office-Home dataset, and 89.1% accuracy on the Office31 dataset, showing its superiority by 2~3 percent above the existing state-of-the-art methods.

Softmax is widely used in neural networks for multiclass classification, gate structure, and attention mechanisms. The statistical assumption that the input is normally distributed supports the gradient stability of softmax. However, when used in attention mechanisms such as transformers, because the correlation scores between embeddings are often not normally distributed, the gradient vanishing problem appears, and we prove this point through experimental confirmation. In this work, we suggest replacing the exponential function with periodic functions, and delve into some potential periodic alternatives of Softmax from the viewpoint of value and gradient. Through experiments on a simply designed demo referenced to LeViT, our method was proven to be able to alleviate the gradient problem and yield substantial improvements compared to Softmax and its variants. Further, we analyze the impact of pre-normalization for Softmax and our methods through mathematics and experiments. Finally, we increase the depth of the demo and prove the applicability of our method to deep structures. The code are available at https://github.com/slwang9353/Period-alternatives-of-Softmax.

In this paper, a mutual coupling reduction technique using filtering structures between two antennas resonating in adjacent frequency bands is proposed. Two patch antennas resonating at low band (4.8-5.0 GHz, part of 5G Band N79) and high band (5.15-5.35 GHz, part of IEEE 802.11 ax 5 GHz Band), which are close to each other in frequency spectrum, are used as an illustrative example. The decoupling structure consists of two open-loop resonators with filtering function coupled to the feed line on its edges. By loading the filtering structures, the isolation between the two antennas is improved from poorer than 15 dB to more than 25 dB when the reflection coefficients are still satisfactory in their respective operating frequency bands, showing significant improvements compared to the reference antennas where no filtering structures are introduced. The proposed method can be easily applied to either mobile terminal or 5G CPE (Customer Premise Equipment) devices where both 5G and Wi-Fi are installed as well as many other antennas with similar application scenarios.

Non-thermal plasma is an alternative technology of exploiting biomass for efficient production of value added energy carriers. In this study, plasma catalytic liquefaction (PCL) system is successfully developed to achieve rapid and efficient liquefaction of sawdust in the presence of co-solvent polyethylene glycol 200 (PEG 200) and glycerin as well as the acid catalyst H2SO4. The influences of different operating factors such as the catalyst concentration, reaction time, molar ratio of PEG 200 and glycerin and the power supply are investigated in terms of the liquefaction yield and solution temperature, while the roles of the applied voltage and the duty cycle of power supply in the PCL process are discussed in detail. Experimental results show that the sawdust could be completely liquefied into crude bio-oil with a high heating value (HHV) of 19.18 MJ/kg in the mixture containing sawdust and solvent in a ratio of 1/7, and 1 wt.% catalyst H2SO4. Increasing the applied voltage and duty cycle are more efficient giving higher yield of liquid products. Based on the analysis of the processing conditions, suggested that the enhanced performance of the PCL process is linked to the superiority of heating that induced by electric field and the presence of quantities reactive species during the plasma discharge. The physiochemical characteristics of the biomass sample and the liquefied products are interpreted by using elemental analysis, thermogravimetric analysis, Fourier transform infrared spectroscopy (FTIR) and the identities and concentration of the liquid products are determined by gas chromatography-mass spectrometry (GC-MS).

Steganography is to hide secret information in a normal cover, so that the secret information cannot be detected. With the rapid development of steganography, it's more and more difficult to detect. Steganalysis is the counter of steganography. In order to improve the detection effect, more complex high-dimensional features are proposed for steganalysis. However, this also creates huge redundancy features, which in turn consume generous time. Feature selection is a technique that can effectively remove redundant features. In this paper, we propose a new blind image steganalysis algorithm to distinguish stego images from cover images using a nature-inspired feature selection method based on the binary bat algorithm(BBA). Meanwhile, SPAM and several classifiers have been used to improve the detection effect. Furthermore, we select the ideal feature subset using BBA from the original features and use the selected feature subset to train the several classifiers. The experimental results demonstrate that our proposed method can improve the detection effect and reduces the redundant features.

Event Extraction plays an important role in information-extraction to understand the world. Event extraction could be split into two subtasks: one is event trigger extraction, the other is event arguments extraction. However, the F1-Score of event arguments extraction is much lower than that of event trigger extraction, i.e. in the most recent work, event trigger extraction achieves 80.7%, while event arguments extraction achieves only 58%. In pipelined structures, the difficulty of event arguments extraction lies in its lack of classification feature, and the much higher computation consumption. In this work, we proposed a novel Event Extraction approach based on multi-layer Dilate Gated Convolutional Neural Network (EE-DGCNN) which has fewer parameters. In addition, enhanced local information is incorporated into word features, to assign event arguments roles for triggers predicted by the first subtask. The numerical experiments demonstrated significant performance improvement beyond state-of-art event extraction approaches on real-world datasets. Further analysis of extraction procedure is presented, as well as experiments are conducted to analyze impact factors related to the performance improvement.

Multi-block linear constrained separable convex minimizations are ubiquitous and have been drawing increasing attention in recent researches. The alternating direction method of multipliers (ADMM) has been well studied and used in the literature for the two-block case. However, the direct extension of the ADMM to the multi-block case is not necessarily convergent. ADMM with Gaussian Back Substitution and ADMM with Prox-Parallel Splitting are two useful schemes to deal with the multi-block situation. Nevertheless, only a sublinear convergence rate was given in previous studies. In this paper, we prove the linear convergence rate of these two schemes under some assumptions. The proofs mainly depend on the variational inequalities.

In recent years, long term evolution for railway (LTE-R) has been a promising technology to meet the growing demand for railway wireless communication. To realize the active maintenance of LTE-R base station, it is of great significance to precisely predict the communication quality (CQ) of LTE-R base station. Given that the existing LTE CQ prediction methods can not support the active maintenance of LTE-R base station. Furthermore, the LTE-R base station has its unique characteristics in time relationship and regional impact, one of the most challenging problems is to effectively integrate the temporal and spatial information to improve the effect of CQ prediction. To solve the above problems, we choose daily evolved radio access bearer (E-RAB) abnormal release ratio as the CQ indicator, and propose a new deep learning-based CQ prediction approach for LTE-R. Considering the influence of adjacent base stations, this method conducts temporal-spatial collaborative prediction on multivariate time series collected from the CQ data of these stations. First, to eliminate the negative effect of redundant variables, a new variable filter method based on max-relevance, and min-redundancy (MRMR) criterion and binary particle swarm optimization (BPSO) is proposed to select a variable set from the CQ data of related base stations. Second, a new recurrent convolutional neural network (RCNN) model with a self-attention mechanism is proposed to extract temporal-spatial features from the selected variable set. With these features, we build a collaborative prediction model for CQ prediction. Experimental results on real-world LTE-R CQ datasets demonstrate the superiority of the proposed method in CQ prediction.

Conventional optimization-driven secure relay selection relies on maximization algorithm and accurate channel state information (CSI) of both legitimate and eavesdropper channels. Particularly, estimating and collecting accurate eavesdropper CSI is a difficult task. In this paper, we exploit the benefits of machine learning in solving secure relay selection problem from a data-driven perspective. We convert secure relay selection to a multiclass-classification problem and solve it by a decision-tree-based scheme, which is composed of three phases - preparing training data, building decision tree and predicting relay selection. To meet decision tree's requirement that input features must take discrete values, a feature extraction method is proposed to generate discrete input by quantizing the accurate CSI of legitimate and eavesdropper channels. By this means, the decision-tree-based relay selection only requires quantized CSI feedback which takes substantially fewer bits in predicting phase. For the purpose of optimizing quantization parameters and enhancing decision tree prediction, we further derive three splitting criteria, i.e. information gain, information gain ratio and Gini index. Simulation results show that if the quantization parameters are set properly, the proposed decision-tree-based scheme can achieve satisfactory performance in terms of average secrecy rate while reducing computational complexity and feedback amount.

Face sketch synthesis for photos is an applied research topic and it is critical for criminal investigation. However, sketch synthesis remains some challenges because of the blur and artifacts in the generated face sketches. To mitigate these problems in face sketch synthesis, we propose a fast Generative Adversarial Network with fast Multi-scale feature channel Attention, namely MAGAN. In the generator network, multi-scale features are extracted by proposed multi-scale feature extraction to produce detailed sketches. Then, a channel attention mechanism is applied to emphasize the significance of important feature channels, further enhancing the synthesized sketches. Besides, the loss of patch-wise high-layer features from the VGG-19 network is applied to supervise the generator to synthesize more realistic sketches. To accelerate the training process, the features from the pooling layers are adopted to calculate the pseudo sketch feature loss. The experimental results demonstrate that our MAGAN can achieve better performance in both visual evaluations and quantitative evaluations (in terms of feature similarity and learned perceptual image patch similarity), compared with the state-of-the-art methods.

Modern cyber-physical power systems are vulnerable to cyber attacks. Given that cyber and physical networks are coupled tightly, attacks in the cyber layer can penetrate the physical layer, causing the outage of transmission lines and other physical equipment, thus changing the topology of the grid. In some extreme scenarios, the topological change will disrupt the emergency response of the grid, eventually causing cascading outages along with a blackout. Therefore, as the defender, the operator of a cyber-physical power system should identify critical cyber attacks. In this paper, patterns of sequential cyber topological attacks are analyzed. Firstly, a coordinated attack process is established, including mechanism and probability analysis considering the different timescales. Secondly, the concept of patterns is defined as minimal attack sequences aimed at causing blackouts. Furthermore, the representativeness of patterns is illustrated, which can significantly reduce the storage of risky attack sequences. Thirdly, to address the problem that the identification of patterns is computationally intensive, a search strategy that selects the next attack target dynamically and increases the search depth gradually is proposed to avoid unnecessary search trials. Lastly, tests are carried out on the IEEE 39-node system using the AC power flow model, which validates the representativeness of patterns and the performance of the proposed search strategy.

Using the same algorithm and hyperparameter configurations, deep reinforcement learning (DRL) will derive drastically different results from multiple experimental trials, and most of these results are unsatisfactory. Because of the instability of the results, researchers have to perform many trials to confirm an algorithm or a set of hyperparameters in DRL. In this article, we present the policy return method, which is a new design for reducing the number of trials when training a DRL model. This method allows the learned policy to return to a previous state when it becomes divergent or stagnant at any stage of training. When returning, a certain percentage of stochastic data is added to the weights of the neural networks to prevent a repeated decline. Extensive experiments on challenging tasks and various target scores demonstrate that the policy return method can bring about a 10% to 40% reduction in the required number of trials compared with that of the corresponding original algorithm, and a 10% to 30% reduction compared with the state-of-the-art algorithms.

Holiday tourism flow is a significant indicator to evaluate the development of tourism. The exploration of the rule of tourism flow between cities can not only provide reasonable suggestions for stimulating demand, promoting consumption and economic development, but also make crucial significance for the management of tourism destinations and the optimization of spatial structure of tourism flow. Based on Weibo check-in data, this paper, by using social network analysis, studies the spatial distribution and network structure characteristics of tourism flow in 50 major cities in China during the National Day holiday in 2018. The results show that: 1) the tourism flow connection of the main cities in China represents a diamond shaped spatial structure with “Beijing-Shanghai-Guangzhou-Chengdu” as the core. There exists spatial heterogeneity in different levels of tourism flow intensity and proximity and selectivity in tourism links between cities; 2) the intensity of tourism connection between cities in China is clearly divided into different levels. On the basis of index of degree of centrality, Beijing and Shanghai are far higher than other cities;3) there are obvious differences between core nodes and edge nodes, with the core nodes often composed of cities with high economic development or rich tourism resources. Although the number is small, it plays a significant role in driving the edge cities; 4) the urban tourism flow network is relatively stable, but most cities have relatively weak tourism links and more small-scale tourism flows. In the division of cohesive subgroups, the fifth and sixth subgroups are not only the main tourist sources but also the main destinations. Whether it is the internal connection of subgroups or the connection with other subgroups, tourism flow has a very high density of connection.

In this paper, we propose and experimentally demonstrate a novel method for simultaneously measuring mode-dependent loss (MDL) and differential modal group delay (DMGD) in few mode fibers (FMFs). This method is based on analyzing the Rayleigh backscattering amplitudes obtained with a special nondestructive measurement system. With this method, the MDL and DMGD in about 9.8 km 3-mode fiber are successfully measured, which are 0.186 dB and 3.581 ps/m between LP01 and LP11a and 0.157 dB and 3.604 ps/m between LP01 and LP11b, respectively. Then, the experimental results of DMGD and MDL with different methods are compared, which show that the values of DMGD and MDL obtained by the proposed method agree well with that using the time-of-flight (TOF) and conventional transmission methods, respectively. In addition, it is also demonstrated that the proposed scheme can be scalable to measure FMFs with a large number of modes.

We propose a point-by-point inscription method for fiber Bragg grating (FBG) fabrication using a femtosecond pulsed laser focused through a fiber polyimide coating, realizing a 2-μm-wavelength-band FBG. A femtosecond laser acts as the inscription laser, focused by a 63× oil lens. A first-order FBG was fabricated with a grating interval of 700 nm. The 15-μW pulsed laser was focused directly on the fiber core through the polyimide coating and cladding. A first-order FBG for 2004.8-nm light with more than 56% reflectances was fabricated. High wavelength sensitivity and linearity of the FBG were demonstrated by temperature and strain-sensing testing. The first-order FBG of 2004.8 and 2000.2 nm resonant wavelengths could be realized simultaneously through parallel inscription method in 3000 μm long grating area.

In this paper, an extremely compact two-element linear polarized multiple-input-multiple-output (MIMO) antenna array with a metasurface superstrate is proposed. Instead of using periodic square split-ring resonators which occupy larger space and which are incident angle variant, double-layer short wire is utilized as the unit cell of the metasurface. The metasurface is compact in size and effective in decoupling two nearby Bowtie antennas strongly coupled in the H-plane with the spacing of only 0.27 wavelength. After decoupling, the isolation between the two antennas has been improved from around 10 dB to more than 25 dB within the band of 2300 to 2690 MHz while their reflection remained below −15 dB. Moreover, the radiation pattern after adding the metasurface superstrate is well maintained with total efficiency improvement by about 10%, and the envelope correlation coefficient between the two antennas is reduced from 0.35 to below 0.12 within the whole band of interest. The proposed method can find plenty of applications in MIMO and 5G communication systems.

Compressed Sensing (CS) attempts to acquire and reconstruct a sparse signal from a sampling much below the Nyquist rate. In this paper, we proposed novel CS algorithms for reconstructing under-sampled and compressed electrocardiogram (ECG) signal. In the proposed CS-ECG scheme, the ECG signal was first sub-sampled randomly and mapped onto a two-dimensional (2D) space by using Cut and Align (CAB), for the purpose of promoting sparsity. A nonlinear optimization model was then used to reconstruct the 2D signal. In the compression scheme, the ECG signal was mapped into the frequency domain, and the compression was achieved by a series of multiplying and accumulating between the original ECG and a Gaussian random matrix. For the reconstruction, two matching pursuits (MP) methods and two blocks sparse Bayesian learning (BSBL) methods were implemented and evaluated by the percentage root-mean-square difference (PRD). Based on the test with real ECG data, it was found that the proposed CS scheme was capable of faithfully reconstructing ECG signals with only 30% acquisition.

As millimeter-wave technology becomes more and more mature, directional modulation (DM) based on large uniform linear arrays, where the adjacent antennas spaced half-wavelength of the frequency of interest, can be realized to ensure the security of millimeter-wave systems. However, for small inter-element spacings, the mutual coupling effects are severe. Therefore, a DM technique using linear sparse arrays (LSAs) is presented to improve spatial resolution and mitigate the effects of mutual coupling. The main idea for DM signal synthesis is to modulate the radiation pattern at the symbol rate by randomly selecting a sparse rate, i.e., a filling fraction, from the given interval [αmin, αmax] and an array formation, i.e., an antenna subset, from the codebook P. This results in a directional radiation pattern that projects a standard constellation in the desired direction and distorted randomized constellations in other directions, which offers security for wireless communication. Numerical comparisons of the symbol error rate performance of the proposed DM technique against conventional array transmission and antenna subset modulation are presented to highlight the advantages of the DM technique for LSAs.

The in-cessation trend of railways ramification calls for railway sensing on an urgent basis. Railway sense is required to keep suspecting potential danger in a large scope and provide a safe transportation environment. The fundamental infrastructure to realize railway sensing comprises of space and terrestrial integrated network (STIN) nodes, such as high-speed railway, trackside equipment, unmanned aerial vehicle, airship, and remote sensing satellite. This architecture needs to support diverse applications flexibly and ensure efficient infrastructure management. Inspired by the philosophy of software-defined network, which attempts to give more flexibility to networks, we propose a software-defined sensing and integrated architecture for such a network. We decouple the railway sensing application from the physical infrastructure. Besides, we designed centralized controllers to manage physical facilities and supply APIs of data processing, including acquisition, transmission, computation, and storage. Various applications can share common infrastructure with such properties, and each of these applications can customize its data acquisition, transmission, and computing by requesting APIs of controllers. This paper includes discussions of the design details of the proposed architecture and benefits. We also proposed some open problems and the potential solutions to solve them, which we hope to provide railway sensing research in a new direction.

The Map-Join-Reduce programming model effectively supports the join operation among different heterogeneous data sets by adding the Join module and processes the multiway joining directly. In this paper, we propose a rigorous description of the Map-Join-Reduce that performs as an executable specification. First, this paper briefly introduces the differences between the Map-Join-Reduce and the MapReduce. Then, we use Haskell to specify each module of the Map-Join-Reduce programming model and analyze the structure and function of each module. Finally, we test the specification by analyzing an example of the mall sales records. The executable specification contributes to helping the developers to unscramble the relationship between the MapReduce and the Map-Join-Reduce, which may serve as a basis for further development of the theory of related programming model design. Furthermore, the most important function of an executable specification is guaranteeing the target informal or semi-formal model with interesting properties. This paper is a forward step to prepare for verifying related properties and, even, providing verified prototypes.

This paper proposes a learning-based image interpolation method based on weighted direct nonlinear regression. It attempts to learn the nonlinear relationship between the low-resolution patches and their corresponding high-resolution patches by using an external database of natural images. In the training phase, without being initialized to the same size as the high-resolution patches using bicubic interpolation, low-resolution patches are directly used for training, which will reduce the number of parameters. Dictionary learning combined with nearest neighbor searching based on the normalized correlation coefficient in the entire training set is proposed as a soft classification method. Afterward, a single hidden-layer feed-forward network with random input weights is adopted to learn the direct nonlinear mapping for the regression in each class. In the interpolation phase, the strategy of weighting multiple estimations is applied to enhance the interpolation performance. Furthermore, a refinement process is proposed to improve the interpolation performance with sharper edges and richer details. Extensive experimental results demonstrate that our proposed method is not only effective and efficient but also outperforms many state-of-the-art methods.

Person re-identification has gradually become a popular research topic in many fields such as security, criminal investigation, and video analysis. This paper aims to learn a discriminative and robust spatial–temporal representation for video-based person re-identification by a two-stage attribute-constraint network (TSAC-Net). The knowledge of pedestrian attributes can help re-identification tasks because it contains high-level information and is robust to visual variations. In this paper, we manually annotate three video-based person re-identification datasets with four static appearance attributes and one dynamic appearance attribute. Each attribute is regarded as a constraint that is added to the deep network. In the first stage of the TSAC-Net, we solve the re-identification problem as a classification issue and adopt a multi-attribute classification loss to train the CNN model. In the second stage, two LSTM networks are trained under the constraint of identities and dynamic appearance attributes. Therefore, the two-stage network provides a spatial–temporal feature extractor for pedestrians in video sequences. In the testing phase, a spatial–temporal representation can be obtained by inputting a sequence of images to the proposed TSAC-Net. We demonstrate the performance improvement gained with the use of attributes on several challenging person re-identification datasets (PRID2011, iLIDS-VID, MARS, and VIPeR). Moreover, the extensive experiments show that our approach achieves state-of-the-art results on three video-based benchmark datasets.

Phishing attack is now a big threat to people's daily life and networking environment. Through disguising illegal URLs as legitimate ones, attackers can induce users to visit the phishing URLs to get private information and other benefits. Effective methods of detecting the phishing websites are urgently needed to alleviate the threats posed by the phishing attacks. As the active learning capability from massive data sets, the neural network is widely used to detect the phishing attacks. However, in the stage of training data sets, many useless and small influence features will trap the neural network model into the problem of over-fitting. This problem usually causes the trained model that cannot effectively detect phishing websites. In order to alleviate this problem, this paper proposes OFS-NN, an effective phishing websites detection model based on the optimal feature selection method and neural network. In the proposed OFS-NN, a new index, feature validity value (FVV), is first introduced to evaluate the impact of sensitive features on the phishing websites detection. Then, based on the new FVV index, an algorithm is designed to select the optimal features from the phishing websites. This algorithm is able to alleviate the over-fitting problem of the underlying neural network to a large extent. The selected optimal features are used to train the underlying neural network, and finally, an optimal classifier is constructed to detect the phishing websites. The experimental results show that the OFS-NN model is accurate and stable in detecting many types of phishing websites.

Recently, person re-identification has become one of the research hotspots in the field of computer vision and has received extensive attention in the academic community. Inspired by the part-based research of image ReID, this paper presents a novel feature learning and extraction framework for video-based person re-identification, namely, the extended global-local representation learning network (E-GLRN). Given a video sequence of a pedestrian, the holistic and local features are simultaneously extracted using the E-GLRN network. Specifically, for the global feature learning, we adopt the channel attention convolutional neural network (CNN) and the bidirectional long short-term memory (Bi-LSTM) networks, which are responsible for introducing a CNN-LSTM module to learn the features of consecutive frames. The local feature learning module relies on the key local information extraction, which is based on the Bi-LSTM networks. In order to obtain the local feature more effectively, our work defines a concept of “the main image group” by selecting three representative frames. The local feature representation of a video is obtained by exploiting the spatial contextual and appearance information of this group. The local and global features extracted in this paper are complementary and further combined into a discriminative and robust feature representation of the video sequence. Extensive experiments are conducted on three video-based ReID datasets, including the iLIDS-VID, PRID2011 and MARS datasets. The experimental results demonstrate that the proposed method outperforms state-of-the-art video-based re-identification approaches.

Identification of circRNA-disease associations provides insight into the mechanism that circRNAs cause diseases. Wet experimental identification of circRNA-disease associations is time-consuming and labor-intensive, and thus developing computational methods for the circRNA-disease association prediction is an urgent task. In this paper, we propose a linear neighborhood label propagation method to predict circRNA-disease associations, named CD-LNLP. First, CD-LNLP uses association profiles based on known associations to calculate circRNA-circRNA similarities and disease-disease similarities. Next, CD-LNLP implements the label propagation based on the circRNA-circRNA similarity-based graph and the disease-disease similarity-based graph respectively to predict circRNA-disease associations. Finally, we combine the outputs from circRNA-circRNA similarity-based graph model and disease-disease similarity-based graph model to produce the results. In the experiments, CD-LNLP achieves impressive performance with the AUPR score of 0.4487 and the AUC score of 0.9007 and outperforms outstanding baseline methods (collaborative filter method, KATZ, nonnegative matrix factorization method, and resource allocation method) and the state-of-the-art method MRLDC. The case studies show that CD-LNLP identifies novel circRNA-disease associations, which are validated by up-to-date circRNA-disease databases and literature respectively. In conclusion, CD-LNLP is a promising method for predicting circRNA-disease associations.

An ultra-high precision magnetic field measurement is of great scientific and economic significance. An atomic magnetometer that operates in a spin-exchange relaxation-free (SERF) regime has superior sensitivity and, thus, is of great significance for the ultra-high precision magnetic field measurement. In order to reduce the noise of a SERF atomic magnetometer and further improve its sensitivity, a method for noise reduction based on the multigene genetic programming (MGGP) is presented. Different from the existing methods, in this method, the model of magnetometer noise is established based on the experimental data. Namely, the noise of the SERF atomic magnetometer is first modeled by the MGGP algorithm and then reduced by the obtained model. Besides, in this way, the sensitivity is improved. The experimental results indicate that our MGGP model can adequately reflect the characteristics of the SERF magnetometer noise. Moreover, after applying the proposed method, the sensitivity of the SERF magnetometer is improved about 13 times at 1 Hz, and there is also a significant sensitivity increase at the frequencies less than 10 Hz. Therefore, the proposed method can effectively reduce the noise and improve the sensitivity of the SERF magnetometer in the low-frequency band.

The prognostic and health management (PHM) of lithium-ion batteries has received increasing attention in recent years. The remaining useful life (RUL) prediction and state of health (SOH) monitoring are two important parts in PHM of the lithium-ion battery. Nowadays, the development of signal processing technology and neural network technology introduces new data-driven methods to RUL prediction and SOH monitoring of the lithium-ion battery. This paper presents a neural-network-based method that combines long short-term memory (LSTM) network with particle swarm optimization and attention mechanism for RUL prediction and SOH monitoring of the lithium-ion battery. Before predicting RUL of the lithium-ion battery, the Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) is utilized for the raw data denoising, which can improve the accuracy of prediction. A real-life cycle dataset of lithium-ion batteries from NASA is used to evaluate the proposed method, and the experiment results show that when compared with traditional methods, the proposed method has higher accuracy.

Directional modulation (DM) technique based on artificial noise (AN) enables the physical layer security (PLS) for wireless communications in free space. However, on one hand, the introduction of AN results in less power efficient for a DM system; on the other hand, the AN-aided DM technique is strongly dependent on the channel state information (CSI) of transceivers, which greatly limits its practical application. What is more, when the eavesdropper (EVE) is slowly approaching the legitimate user (LU) or the EVE and the LU are in the same direction, the transmission security cannot be guaranteed. To circumvent these problems, DM technique with multiple parameters weighted-type fractional Fourier transform (MP-WFRFT) and chaotic scrambling (CS) aided is proposed in this paper to realize the power-efficient and security-enhanced wireless transmissions. Then, the symbol error rate (SER), secrecy rate, robustness, and anti-interception performance of the proposed method are analyzed and simulated. The simulation results are presented to verify that the proposed method is more power-efficient than the traditional AN-aided DM schemes, and the PLS is guaranteed due to the proposed approach, albeit knowing the intended direction for an EVE.