最新刊期
卷 48 ， 期 7 ， 2020
- 摘要：To improve the distribution performance of multiobjective particle swarm optimization algorithm, an adaptive multiobjective particle swarm optimization algorithm, based on the decomposed archive, named AMOPSO-DA, is developed in this paper. First,an external archive update strategy, based on the spatial distribution information of optimal solutions, is designed to improve the searching ability of AMOPSO-DA. Second, an adaptive flying parameter adjustment strategy, based on the evolutionary direction information of each particle, is proposed to balance the exploration ability and the exploitation ability. Finally, this proposed AMOPSO-DA is applied to some multiobjective optimization problems. The experiment results demonstrate that AMOPSO-DA can obtain well-distributed optimal solutions.关键词：multiobjective particle swarm optimization algorithm;decomposition approach;external archive;distribution;adaptive203|596|5更新时间：2025-07-08
- 摘要：Attention-based model is a popular model in speech recognition, however it has a disadvantage that the attention-based model may produce abnormal scores. To solve this problem, this paper first proposes a forward attention model, which adopts normal attention score at the previous moment to smooth the abnormal score at the current moment. Then, the model is optimized to add constraint factors to the attention score at the previous moment to achieve the purpose of adaptive smoothing of the above abnormal scores. Then, a multi-scale forward attention model is proposed on the above model. This model introduces a multi-scale method to model the speech primitives of different levels, and then fuses the target vectors of different levels to solve the outliers of attention score. In the experiment, SwitchBoard is adopted as the training set and Hub5'00 as the test set. Compared with the baseline system, the Word Error Rate (WER) of the proposed system decreased by 14.28% relatively.关键词：forward attention mechanism;adaptive smoothing;multi-scale;speech recognition116|262|5更新时间：2025-07-08
- 摘要：To construct a video action recognition model with 2D neural network speed while maintaining the performance of 3D neural network, the 3D multi-branch aggregation lightweight network action recognition algorithm is proposed. Firstly, the neural network is divided into multiple branches by using grouped convolution. Secondly, to promote the information exchange between branches, a multiplexer module with information aggregation function is added. Finally, the adaptive attention mechanism is introduced to redirect channel and spatio-temporal information. Experiments show that, the computational cost of the algorithm on the UCF101 dataset is 11.5GFlops, and the accuracy is 96.2%; the computational cost on the HMDB51 dataset is 11.5GFlops, and the accuracy is 74.7%. Compared with other action recognition algorithms, it improves the efficiency of the video recognition network and reflects certain recognition speed and accuracy advantages.关键词：deep learning;neural network;action recognition67|180|6更新时间：2025-07-08
- 摘要：For the diversity of group behavior characteristics in complex scenes and the problem of difficult interaction modeling, this paper proposes a new two-layered network architecture. The first layer of network combines a pseudo 3D residual network with a graph convolution network to capture the interaction characteristics. The second layer of network, uses the pseudo 3D residual network to capture the group global scene spatio-temporal characteristics. Based on the complementary role of the above features, their group behavior decisions are fused with a weight adaptive adjustment algorithm, which adaptively calculates importance weights for the group behavior categories predicted by the above two channels, and realizes decision fusion of the different prediction results. The method has achieved 91.4% and 97.9% average recognition accuracy on CAD and CAE respectively.关键词：Group behavior recognition;interaction modeling;adaptive decision fusion66|141|4更新时间：2025-07-08
- 摘要：In order to improve the performance of speech enhancement networks by making full use of noisy speech features, based on the correlation of noisy speech in time and frequency, by combining the local feature extraction ability of convolutional neural networks and the long-term dependence modeling ability of gated recurrent unit, a convolutional gated recurrent network suitable for speech enhancement is designed in this paper. This network uses a convolutional network structure instead of a fully connected network structure to improve the feature calculation process in the gated recurrent unit, thereby can better retain the time-frequency structure in the noisy speech features. The experimental results show that compared with other speech enhancement networks, the proposed network has obvious advantages in speech component retention and noise component suppression, and the enhanced speech has better speech quality and intelligibility.关键词：speech enhancement;deep neural network;gated recurrent unit;convolutional neural network111|130|5更新时间：2025-07-08
- 摘要：To solve the problem of slow speed and low accuracy in the surface defect detection of hot rolled strips, an improved YOLOv3 algorithm is proposed. Firstly, the weighting K-means clustering algorithm is put forward to optimize priors anchor's parameters, which can improve the match between priors anchor and feature map. Secondly, the improved network structure of the YOLOv3 algorithm is proposed to improve the detection accuracy, whose shallow features and deep features are combined to form the new large-scale inspection layer. The experiments are carried out on the NEU-DET dataset, the results show that the average accuracy of the improved YOLOv3 algorithm is 80%, which is 11% higher than that of the original algorithm; the detection speed is 50fps, which is faster than other strip surface defect detection algorithms based on deep learning.关键词：object detection;strip steel surface defect;YOLOv3;weighting K-means221|235|32更新时间：2025-07-08
- 摘要：To obtain a better shadow removal result on images with complex illumination and texture, we proposed a novel approach based on generative adversarial networks. Firstly, shadow mask is generated by the shadow detection sub-net from input shadow image. Based on this detection result, we proposed an illumination sensitive multi-scale image decomposition method to extract the texture information with less or no illumination information loss. Secondly, shadow matte is generated by the matte generation sub-net for the low scale shadow image to remove shadows in it. Thirdly, the shadow boundary are naturally recovered by the boundary completion sub-net. Finally, the shadow removal result is obtained using a detail recovering method guided by adaptive attenuation factor. Experimental results show that the proposed method can improve the removal performance effectively.关键词：image processing;shadow removal;generative adversarial networks;illumination sensitive;multi-scale decomposition124|113|4更新时间：2025-07-08
- 摘要：In order to speed up the generation of test data that covers the target path, the paper makes good use of the balance of individual traversing program to adjust the evolutionary process of generating test data. First, after the individuals run the program, the number of individuals crossing the true and false branches of each branch node is counted. Then, the program balance is designed and calculated. Finally, the influence on the program balance of each individual is calculated. The individual with high influence has a bigger fitness value to have greater chance to participate in subsequent evolution. The proposed method effectively improves the efficiency of test data generation. The experiment results of benchmark programs and industrial cases show that our methods have superiority in running time and success rate of test data generation when compared with similar methods.关键词：software testing;path coverage;genetic algorithm;test data;program balance83|128|5更新时间：2025-07-08
- 摘要：To solve the contradiction between contrast enhancement and noise suppression in image enhancement, an image enhancement algorithm based on dual domain decomposition (IEDD) is proposed. The principles and methods including spatial domain decomposition, layered spatial images enhancement and transform domain noise reduction, and layered images synthesis are described in detail. Firstly, the image is decomposed into a base layer and a detail layer by a Gaussian filter, that decouples the contrast enhancement and noise reduction. Next, in order to realize the enhancement of base layer and noise reduction of detail layers synchronously, the single-scale Retinex with correction function and the nonsubsampled shearlet denoising algorithm with hard threshold shrinkage are implemented. Finally, to ensure the uniform color and the outstanding detail of the composite image, layered image fusion, gray value extension and differential operator detail enhancement are implemented, realizing the grayscale extension and detail enhancement of the composite image. Experiments show that the performance of the proposed algorithm is better than other nine algorithms in improving image contrast and suppressing noise.关键词：image enhancement;spatial decomposition;Gaussian filter;single scale Retinex;non-subsampled shearlet transform97|143|3更新时间：2025-07-08
- 摘要：Traditional morphology-based neuronal classification approaches largely rely on the feature extraction and selection techniques of neuronal spatial structures, a lot of useful information for neuronal classification may be lost. Using the adaptive projection algorithm to convert the three-dimensional neuron data without feature extraction, this paper proposes a neuronal morphology classification approach based on deep learning networks. The three-dimensional voxel reconstruction is used for the original neuron data, and the two-dimensional neuron data is generated through adaptive projection process. Then, the deep learning model of double convolutional gated recurrent neural networks is established to classify neurons. The proposed approach is successfully applied to three neuronal classification datasets, the experiment results show that the proposed method has higher classification accuracy and flexibility than the neuronal classification methods based on feature extraction.关键词：neuronal classification;adaptive projection;convolution neural network;gated recurrent unit143|228|2更新时间：2025-07-08
- 摘要：In order to improve the generation efficiency of multipath coverage test data, a novel method is proposed based on ant colony algorithm (ACO). Firstly, an improved ACO is developed. The importance of an ant to generate test data is considered as a factor for ant state transfer and path mutation. As a result, more ants are guided to traverse small probabilities node and the efficiency of test data generation is improved. Secondly, according to the improved ACO, test data generation of multipath coverage based on single pheromone table and multiple pheromone tables are proposed. In a multiple pheromones table based approach, the pheromone table of each target path is also used to generate test data for other target path, and the test data of multiple paths are generated by running ACO only once. Finally, the effectiveness and complexity of the proposed method are analyzed theoretically. Experimental results show that test data generation based on multi-pheromone tables can effectively generate multipath coverage test data compared with other methods.关键词：test data generation;ant colony algorithm;multiple paths;path coverage;valuableness of ants179|208|8更新时间：2025-07-08
- 摘要：This paper studies the multi-layers embedding problem of survivable virtual networks based on virtual network layer protection. We first establish an integer linear programming (ILP) for the SVNME problem. Then an efficient heuristic algorithm VNP-SVNME is proposed to solve the large-scale problem. Experiments show that the resource mapping cost of the VNP-SVNME algorithm is only 15% higher on average than the optimal solution, but it is better than the state-of-the-art heuristic. In addition, the time complexity is greatly reduced compared to ILP, which can meet the requirements of online virtual network mapping.关键词：virtual network multi-layers mapping;survivability;integer linear programming;heuristic41|82|1更新时间：2025-07-08
- 摘要：Based on the free particle model of quantum system, the Multi-scale Free Particle Optimization Algorithm (MFPOA) is proposed, and the internal mechanism of the algorithm is studied on the basis of the physical model. Through analogy between quantum system and optimization system, the solving process of optimization problem is transformed into the motion process of particles under the microscopic system. The parameter setting of free particle optimization is studied on MATLAB simulation platform, and the differences between the algorithm and similar search mechanism are analyzed. Finally, experiments show that MFPOA is more suitable for solving single-mode functions, and more iterations are needed to solve complex multi-mode functions.关键词：free particle;harmonic oscillator;uniformly distributed sampling;Gaussian sampling;wave function62|118|1更新时间：2025-07-08
- 摘要：To break through the limitations of two-dimensional or three-dimensional Hilbert space, a quantum voting scheme based on d-dimensional three-particle entangled state was proposed, which was designed by using Shamir (t,n) threshold to meet the actual demand for voting. The scheme consists of four entities: the voting management center, the voting group, the scrutineer and the ballot counter, which work together to complete the voting. In this scheme, a d-dimensional quantum entanglement state is used to enhance the applicability, and the single particle is used as the voting carrier to improve the transmission efficiency. The correctness of the proposed scheme was verified by the experimental simulations on the IBM quantum cloud platform. Security analysis shows that the scheme meets the security requirements of voting schemes, and resists intercept-measure-resend attack, entangle-measure attack, and forgery attack. Performance analysis shows that with the increase of the number of participants, the scheme has higher qubit efficiency than other similar schemes.关键词：quantum voting;d-dimensional three-particle entangled state;(t,n) threshold;quantum cloud platform40|67|5更新时间：2025-07-08
- 摘要：The multidimensional hybrid indices optimization problem is a kind of uncertainty multi-objective optimization problems that is difficult to solve. First, we can get relevant weights by optimizing the main parameters of explicit and implicit indices. According to these weights, multidimensional explicit indices can be reduced to an equivalent-interval fitness, and multidimensional implicit indices can be reduced to an equivalent-fuzzy fitness. Equivalent-interval fitness and equivalent-fuzzy fitness can be synthesized to an equivalent-index body. Then, we select advantage individual on the basis of equivalent-index bodies dominant situation according to adaptive reference point and preference area size. Finally, we adopt an implicit-indices estimation strategy with cluster method to realize interactive evolutionary algorithm within the framework of NSGA-II. The proposed algorithm is applied to two optimization problems with hybrid indices, and the results validate its efficiency and generalization.关键词：evolutionary optimization;hybrid indices;genetic algorithms;interaction49|197|3更新时间：2025-07-08
- 摘要：In this study, we propose an approach to design large depth of field (DOF) millimeter wave (MMW) lens antenna based on Bessel beam. The Bessel beam is generated by using axicon. And the MMW dielectric lens antenna based on Bessel beam or Gaussian beam are designed respectively to satisfy the requirement of imaging. Further, the simulation experiment and analysis are conducted in the 3mm band. The results of simulation indicate that the 3dB beam-width is 3.23mm and the achievable DOF is around 228mm.Compared with the traditional MMW dielectric lens antenna based on Gaussian beam, the achievable DOF can be improved more than five times by MMW dielectric lens antenna based on Bessel beam.关键词：millimeter-wave imaging;depth of field;Bessel beam;security detection53|97|0更新时间：2025-07-08
- 摘要：Due to the particle degeneracy, the particle filter based multi-Bernoulli track-before-detect (TBD) filter has an inaccurate estimate of the multi-target posterior density, which leads to the poor performance of measurement non-coherent integration. In order to solve this issue, the Geodesic particle flow is introduced into the multi-Bernoulli TBD algorithm to improve the estimation of the posterior density. In addition, in the track-merging step, the course information of the target is exploited, which reduces the probability of merging tracks of different targets when they cross. The performance of the proposed algorithm is verified by the simulation results of Swerling 1 fluctuating targets detecting and tracking in Rayleigh clutter.关键词：track-before-detect;multi-Bernoulli filter;particle flow;track merging50|99|4更新时间：2025-07-08
- 摘要：In order to address the issues of scene over-smoothing and motion edge-blurring caused by the existing RGBD scene flow methods under complex scenes, non-rigid movement and motion occlusions, this paper proposes a RGBD scene flow method based on FRFCM (Fast and Robust Fuzzy C-Means) clustering and depth optimization. First, the optical flow information from the consecutive frames is marked as the benchmark and the FRFCM clustering approach is utilized to obtain the initial segmentation of the input image sequences. Second, according to the motion edge information of the depth image, we further optimize the initial segmentation to extract the high-confidence hierarchical motion information. Finally, an energy function of RGBD scene flow based on image segmentation is designed, and the pyramid warping strategy is adopted to compute the scene flow field. We employ the test sets of Middlebury and MPI-Sintel databases to conduct a comparison experiment between the proposed method and the existing RGBD scene flow methods. The experimental results indicate that the proposed method has better accuracy and robustness of scene flow estimation, especially when dealing with the issues of scene over-smoothing and motion edge-blurring.关键词：RGBD scene flow;FRFCM clustering;depth information;over-smoothing;edge-blurring61|88|3更新时间：2025-07-08
- 摘要：It is a challenging task to find the k shortest paths in a time-window network. A node may only be accessible within some specific time windows. In the existing researches, an assume is made that a traveller can pass through an accessible node immediately or wait until the next accessible time window. This paper targets a more general case where a traveller, once arrived at a node, may choose to pass through the node at any discrete times in the time windows of the node. In such a generalized time-window network, the complexity increases significantly, as the size of solution space soars up exponentially. By imitating the natural ripple-spreading phenomenon on a liquid surface, an effective ripple-spreading algorithm (RSA) is proposed for the k shortest paths problem in a generalized time-window network (k-SPPGTW). Besides one-to-one k-SPPGTW, the RSA is also extended to one-to-all k-SPPGTW, where all the k shortest paths from a given source to every other node in the network need to be found. The new method has a theoretically guaranteed optimality. The computational complexity of the RSA is O(k×NATU×NL), where NL is the number of links in the network, and NATU is the average simulated time units for a ripple to travel through a link. The effectiveness and efficiency of the RSA for the k-SPPGTW are demonstrated by some preliminary experimental results.关键词：k shortest paths problem;generalized time-window network;ripple-spreading algorithm37|104|1更新时间：2025-07-08
- 摘要：The layout of the wiring points of the reference potential difference of the compact tension specimen is mostly based on experience, in this paper, a genetic algorithm is proposed to optimize the tension specimen reference potential wiring points layout. The finite element method is used to analyze the potential field of the tensile specimen, the reference potential difference model of the genetic algorithm is constructed, the best reference potential difference value is found, and the optimal wiring point position of the compact tensile specimen is obtained. The results show that the distance between the reference potential difference wiring points has great influence on the reference potential difference, it is close to the current input and the inner part of the upper right end of the tensile specimen are the optimal wiring points position.关键词：compact tensile specimen;genetic algorithm;reference potential difference;optimum wiring point;finite element method;crack monitoring60|12|0更新时间：2025-07-08
- 摘要：In order to achieve energy-saving operation of three-phase bridge rectifier, the topology of an energy-saving three-phase bridge zero-current switching rectifier is proposed. The switching devices of the rectifier can achieve zero-current turn off when the auxiliary resonant circuit on the bridge arm of each phase is in the working state. Three-phase bridge rectifier usually uses insulated gate bipolar transistor (IGBT) as switching devices and the realization of zero-current turn off can eliminate the switching loss caused by tail current of IGBT. The working process of the circuit is analyzed. The experimental results on a three-phase 3kW prototype show that the switching devices achieve zero-current soft-switching. Therefore, this topology can achieve the energy-saving operation of three-phase bridge rectifier which uses IGBT as switching devices.关键词：energy-saving;rectifier;auxiliary circuit;zero-current turn off;tail current132|16|0更新时间：2025-07-08
- 摘要：Image super-resolution reconstruction (SR) aims to obtain high-resolution images from one or more low-resolution images. Recently, SR has been developing and widely applied in different fields. This survey retrospects the history of SR technique and provides a comprehensive overview of representative SR methods, with an emphasis on recent deep learning-based approaches. We elaborate the details of various deep learning-based SR methods, including their strengths and weakness, in terms of the deep learning model, architecture, and message pass. Finally, we discuss the possible research directions on SR technique.关键词：image super-resolution reconstruction;deep learning;image processing;methods review167|14|21更新时间：2025-07-08
- 摘要：The current social network has replaced traditional media as an important platform for information exchange. The information in social networks has the advantages of fast dissemination, wide range, and strong immediacy. However, due to the lack of effective supervision means when publishing information, the social network platform has also become a hotbed of rumors. Therefore, the rapid and effective detection of social network rumors is essential for purifying the network environment and maintaining public safety. Firstly, this article explains the definition of rumors, and the problems of current rumors detection and detection process are described. Secondly, different data acquisition methods are introduced and their advantages and disadvantages are analyzed. At the same time, different data annotation methods in rumor detection are compared. Thirdly, according to the development of rumor detection technology, analyze and compare the existing rumors detection methods of artificial, machine learning and deep learning. Fourthly, current mainstream algorithms are empirically evaluated under the same open data set through experiments. Finally, analyze and summarize the challenges faced by current social network rumor detection technology.关键词：social network;rumor detection;cyberspace security108|4|14更新时间：2025-07-08
- 摘要：Residual neural network (ResNet) has witnessed tremendous amount of attention in deep learning research over the last few years and has made great achievements in computer vision. In this paper, the ResNet is summarized in the following aspects: Firstly, the basic structure and working principle of the ResNet are expounded; Secondly, in model development, the eight network models of the ResNet are summarized in time sequence; Thirdly, in structural optimization, the research progress is described from five aspects of ResNet, including convolutional layer, pooling layer, residual unit, fully connected layer and the whole network; Finally, the application of ResNet in medical images processing is mainly discussed from two aspects of image recognition and image segmentation. In this paper, the principles, models, and structures of ResNet are systematically summarized, which has positive significance to the research and development of ResNet.关键词：residual neural network;network structure;medical image118|4|14更新时间：2025-07-08
- 摘要：Based on the wide bandgap semiconductor SiC APD ultraviolet single photon detector, a 1×8 linear array photon counting readout circuit is proposed in this paper. According to the condition of the light intensity and through suitable timing control, the unitary fixed mode or complementary gating detection mode can be selected to implement the functions of sensing and counting for the wide dynamic range ultraviolet photon signals. The readout integrated circuit is fabricated by TSMC 0.18μm CMOS process, the test results show that the readout circuit is capable of single photon detection, and the performance is in agreement with the expected prediction by the simulation results. Finally, the two-dimensional turntable of the micro-motion system is utilized to complete the solar blind single-wavelength photon detection and imaging, so as to implement accurate discrimination and target localization on the multiple independent ultraviolet light sources.关键词：SiC avalanche photodiode(SiC APD);readout integrated circuit (ROIC);complementary gating;ultraviolet single photon detection and imaging69|8|0更新时间：2025-07-08
- Technical support is provided by Beijing Founder electronics co., LTD 京ICP备09064830号-19
京公网安备11010802024621 - It is recommended to read the content of this site in Chrome&IE9+. Please switch to extreme mode in browser 360.
- Cookies We use cookies to help provide and enhance our service and tailor content. By continuing, you agree to the use of cookies.
0