Search

Given the existing video summarization algorithms have too much redundancy in video semantics, and destroy the maximum continuity between video frames, a smooth video summarization generation method is designed in this paper. This method is based on the video keyframe selection, which is mainly divided into three parts. In the first part, this method completes the de-redundancy operation based on the similarity calculated by the essential characteristics of the video frame. The second part completes the shot segmentation processing using the capsule network. The third part can be divided into two sub-parts, key frames selection and filling. The former of this part is selects key-frames based on video semantic, and the latter is smooths summarization based on frame-filling. Experimental results show that the video summarization algorithms designed in this paper can outperform several existing algorithms and achieve excellent results in terms of video content integrity, semantic accuracy and so on.

Massive MIMO has become one of the core technologies of 5G mobile communication systems. In this paper, fully complementary sequence precoding is designed to reduce the equivalent channel dimension, thereby greatly reducing the pilot overhead. In the beginning, three design criteria are proposed. Based on the design criteria, this paper uses Gray's complementary sequences to design the precoding matrix and finally optimize it into an omnidirectional precoding matrix with better performance.

With the increasing scale of power grid and the continuous increase of distributed energy, the power system is becoming more and more complicated in structure and environment, and is presenting the typical characteristics of big data, such as multi-source, heterogeneous and large amount. The analysis and processing methods of the existing power problems are obviously insufficient to cope with this change, and the study of new technologies and methods is imperative. The main parts of this paper includes three parts: the research on the fusion technology of power information system log data, the research on log data analysis and processing technology, and the research on the key technology of power information system health evaluation model construction. The first two parts are to provide data and quantitative support for the construction of power information system health evaluation model. Through the organic combination of the three parts, the paper constructs a health evaluation model of power information system.

The connection relationship can reflect the hidden information in the network to a certain extent. Through the recognition of the connection relationship, it can infer its hierarchical position in the communication network, and then find the key nodes in the network and analyze the hidden information in the network. To overcome the obstacles of analyzing communication information from frequency hopping signal cracking, a method for recognition of connection relationships based on deep neural networks is proposed by this paper. Automatically learn features directly from simple spectrum monitoring data without the need to manually extract expert features. This paper implements the solution to the problem of imbalanced data categories of self-built data sets; proposes two design schemes based on different design goals and analyzes the performance of these two design schemes; attempts to use Convolutional Neural Networks (CNN) with different input lengths to realize the recognition, it is found that the input length affects the recognition result; the output of the neural network is corrected by the physical meaning so that the recognition result is better. The experimental results show that the CNN models are suitable for connection relationship recognition.

Aiming at the problem that Mark XIIA Mode 5 burst communication system has low recognition rate under interference environment, this paper constructs a strategybased anti-interference intelligent decision engine, which is applied in this system to form an adaptive burst communication system. The paper gives the structure of adaptive burst communication system, designs anti-jamming strategy and effectiveness evaluation objective function, and studies the strategy-based interrogation/response process decision algorithm. The simulation results show that compared with the traditional Mark XIIA Mode 5 system, the adaptive burst communication system using the strategy-based decision can not only avoid or suppress interferences, but also effectively improve the system's power efficiency and transmission reliability in different interference environments.

This paper studies the relay selection in wireless cooperative communication systems. Since the high mobility of Unmanned Aerial Vehicles (UAVs) will lead to the change of link state, a relay selection algorithm based on position prediction is proposed in this paper to better study this problem. First, the Unscented Kalman Filter (UKF) algorithm is used to predict the real-time location information of each UAV, and based on the location information, it is determined whether the current UAV relay node needs to be replaced. The real-time location information of each UAV is predicted by UKF algorithm, which is the basis for determining whether the current UAV relay node needs to be replaced. Then, the outage probability of each standby UAV relay node is calculated, and the standby UAV relay node is sorted according to the link outage probability. Finally, the greedy algorithm is used to find the node with the smallest interruption probability among the backup UAV relay nodes as the next UAV relay node. The simulation results show that in terms of position prediction, the UKF algorithm is more accurate than the Kalman Filtering (KF) algorithm, and the error is smaller. In terms of interruption probability, the relay selection algorithm proposed in this paper reduces the interruption probability of the link, and the interruption probability of the link is more stable.

In this paper, a sorted phase rotation assisted spatial modulation (SM) is developed to optimize the spatial constellation of the received signals in SM multiple input single output (MISO) system. The proposed scheme improves the conventional phase alignment based SM by sorting the Euclidean norm of the channel, and utilizing the sorting results to adjust the phase of each transmit antenna to further optimize the minimum Euclidean distance (MED) of the received constellation. By adopting a phase rotation criterion that is based on the sorted Euclidean distance at the transmitter, the propsed scheme, termed as SM-SPR is presented in this paper. An upper bound of the bit error rate (BER) for the proposed SM-SPR with MISO configuration is also derived in this paper. Simulation results demonstrate that the proposed SM-SPR scheme achieves significant performance gain than conventional SM system. Furthermore, it outperforms the conventional phase alignment assisted SM with improved MED of the constellation.

Mark XIIA identification foe or friend (IFF) system is the main battle equipment that used by the U.S. and NATO forces. Mode 5 is the latest development mode of the system with a stronger anti-interference ability than the previous mode. However, it is still limited and difficult to transmit reliably in the interference environment of linear frequency modulation (LFM). Therefore, this paper proposes a LFM interference cancellation algorithm based on multiple discrete polynomial phase transform and weighted combination (MDPT-WC) for Mark XIIA Mode 5 system, which can reconstruct LFM interference and eliminate it effectively. The simulation results demonstrate that the LFM cancellation algorithm proposed in this paper can improve the transmission accuracy of Mode 5 system in a wide range of interference to signal ratio (ISR), so the proposed algorithm is valuable in application.

In deep space communication systems, phase noise is a non-negligible factor which affects the bit error rate (BER) performance for the low data-rate receiver. According to the power spectral density of phase noise given in consultative committee for space data systems (CCSDS) Proximity-1 protocol, BER performance degradation is more than 1 dB for 8 bps BPSK communication systems. This paper focuses on digital phase-locked loop (PLL) in the receiver affected by the phase noise. The variance of tracking error of digital PLL is deduced, the BER performance is analyzed, and the parameters of digital PLL are optimized to reduce the performance degradation in this paper.

Spatial diversity can be used over satellite-to-ground optical links to mitigate turbulence induced fading. In this paper, bit-error rate (BER) performances of satellite-to-ground downlink optical communication are evaluated for fiber-based coherent detection receiver system with spatial diversity. The performance analysis is carried out with M-ary phase shift keying (M-PSK) and M-ary quadrature amplitude modulation (M-QAM) schemes. Maximum ratio combining (MRC), equal gain combining (EGC), and selection combining (SC) are evaluated as diversity combining methods. We consider Gamma-Gamma distributed model for the signal irradiance and non-central chi-square distribution for the random fluctuation of fiber-coupling efficiency. BER performances of fiber-based coherent detection receiver system with downlink spatial diversity are analyzed for 8PSK and 8QAM modulation schemes and compared with receiver system with single aperture under the same conditions. BER performances of fiber-based receiver systems with spatial diversity are also studied for MPSK and MQAM with different constellation M. It is observed that 8PSK and 8QAM modulated fiber-based receiver systems with spatial diversity can significantly improve the BER performance with EGC and MRC receiver system. For SC diversity, additional number of apertures does not mean better communication performance under the case of low zenith angles. The research work of this paper has laid a analytical foundation for the application of satellite-to-ground downlink multiple apertures receiver system with fiber-based coherent detection MPSK and MQAM modulations.

This paper proposes a matrix operation method for modeling the three-phase transformer by phase-coordinates. Based on decoupling theory, the 12x12 dimension primitive admittance matrix is obtained at first employing the coupling configuration of the windings. Under the condition of asymmetric magnetic circuits, according to the boundary conditions for transformer connections, the transformers in different connections enable to be modeling by the matrix operation method from the primitive admittance matrix. Another purpose of this paper is to explain the differences of the phase-coordinates and the positive sequence parameters in the impedances of the transformers. The numerical testing results in IEEE-4 system show that the proposed method is valid and efficient.

In recent years, with the development and wide application of unmanned aerial vehicle (UAV) technology, the key technology of unmanned aerial vehicle network has become a new research focus. Compared with traditional networks, UAV networks have the characteristics of dynamic topology and limited energy. These characteristics have brought new challenges to the design and implementation of UAV communication systems. We need to redesign the organizational structure of the UAV network to reduce network power consumption, extend network operation time, and improve network stability. In this paper, we have studied the clustering of UAV networks and designed a clustering strategy. First, calculate the best number of clusters in the current network based on bandwidth balance, then use the K-means algorithm to quickly cluster the entire network, and finally implement a cluster head selection algorithm based on Deep Q-learning (DQN) in each cluster . We simulated the clustering strategy and analyzed the average link retention time, average cluster head retention time, network energy consumption, etc., which proved the efficiency and correctness of the algorithm proposed in this paper.

In wireless sensor networks, the signal transmitted by nodes will be affected by the environment. This usually leads to positioning failure or instability. A traditional algorithm is to calculate the distance between nodes by the signal propagation time firstly, and then use the location algorithm based on distance measurement to estimate the coordinates of unknown nodes. However, due to the interference of environmental factors, the method will lead to the accumulation of errors. In order to solve the problems of traditional positioning method and reduce the influence of measurement error, different design schemes are explored in this paper. In this paper, the kernel function is used to improve the positioning accuracy, and the improved method is effective.

Various intelligent analysis applications put forward higher requirements on the image delay performance. Coding format, bit rate and decoding buffer zone setting may influence a camera’s image delay in video surveillance. In this paper, two web cameras of different brands are selected to compare their image delay with different bit rates, coding formats and decoding buffer zone settings. Experimental results show that H.265 has a greater impact on image delay than H.264. The comprehensive computational loads of encoding and decoding need to be considered in analyzing image delay at different video bit rates. The decoding buffer zone setting needs to keep a balance among network jitter, memory and image delay. This paper provides a reference for setting video parameters in different application scenes, so that a better image delay can be achieved to ensure better operation of the system.

With the advancement of communication technology, the interoperability of the power grid operation has improved significantly, but due to its dependence on the communication system, it is extremely vulnerable to network attacks. Among them, the false data injection attack utilizes the loophole of bad data detection in the system and attacks the state estimation system, resulting in frequent occurrence of abnormal data in the system, which brings great harm to the power grid. In view of the fact that false data injection attacks are easy to avoid traditional bad data detection methods, this paper analyzes the different situations of false data injection attacks based on the characteristics of the power grid. Firstly, it proposes to apply the distortion index method to false data injection attack detection. Experiments prove that the detection results are good and can be complementary to traditional detection methods. Then, combined with the traditional normalized residual method, this paper proposes the improved distortion index method based on the distortion index, which is good at detecting abnormal data. The use of improved distortion index method to detect false data injection attacks can make up for the defect of the lack of universality of traditional detection methods, and meet the requirements of anomaly detection efficiency. Finally, based on the MATLAB power simulation test system, experimental simulation is carried out to verify the effectiveness and universality of the proposed method for false data injection attack detection.

Blockchain is being pursued by a growing number of people with its characteristics of openness, transparency, and decentralization. At the same time, how to secure privacy protection in such an open and transparent ledger is an urgent issue to be solved for deep study. Therefore, this paper proposes a protocol based on Secure multi-party computation, which can merge and sign different transaction messages under the anonymous condition by using Pedersen commitment and Schnorr Signature. Through the rationality proof and security analysis, this paper demonstrates the private transaction is safe under the semi-honest model. And its computational cost is less than the equivalent multi-signature model. The research has made some innovative contributions to the privacy computing theory.

Natural language processing (NLP) is a frontier technology in the field of artificial intelligence. Keywords extraction is a key link in NLP and plays an important role in NLP. TFIDF algorithm is considered as the most important invention in information mining. This paper uses the position characteristics of words in the title and full text in the text, and makes weighted improvement on the basis of TFIDF algorithm to improve the accuracy of keyword extraction. In this paper, 400 articles of ACM were used as the training data set, 40 articles as the test set, and accuracy rate, recall rate and F 1 value were used as the evaluation criteria. Experimental data show that this method improves the accuracy of keyword extraction and improves the performance of the original algorithm.

In this paper, based on the analysis of TDOA location algorithm, this paper proposes a mixed Chan_Taylor ranging algorithm based on TDOA, TDOA ranging in the fusion of Taylor algorithm and chan algorithm is proposed, to reduce the positioning error, and ranging experiments are completed on the platform with UWB signal, can be in the process of range due to multipath propagation, environment and equipment power consumption or the error of the interference problems such as optimization, improve the system performance and accuracy. In addition, this paper also discusses the difference in the order of selection when the two algorithms are fused, and it is concluded that it is better to use Chan algorithm first and Taylor algorithm later. Simulation results show that the proposed algorithm has the advantages of improving the positioning accuracy and the stability of the positioning system.

In order to solve the security access problem of power Internet of Things customer side, this paper has designed a miniature communication conversion module which can be deployed in the perceptual layer or marginal layer in power Internet of Things client side. This device can realize such functions as security isolation, information security interaction, access authentication, data/instruction encryption protection, etc. In order to verify the performance of the miniature communication conversion module, this paper has also constructed the testing system for miniature communication conversion module, and tested its security performance. The test results showed that the device can securely isolate the open customer access network from the core network of the power Internet of Things, effectively preventing the core business system from being illegally invaded.

For computer-aided diagnosis of medical issues, image classification plays an important role to ensure high prediction accuracy. Convolutional neural network (CNN) is an important aspect of deep learning because it returns a higher performance rate than other traditional machine learning methods and solves end-to-end classification problems Recently, machine learning and artificial intelligence have rapidly advanced in different fields, especially in the health industry. One of the main medical problems in the world today arises from diseases such as brain tumors, breast cancer, and lung cancer. Therefore, this paper aims at employing image classification and medical image segmentation based on deep learning to detect such medical issues on time. In this paper, a 15-layer CNN model is designed to classify and recognize medical images; especially, breast cancer cells, by using relevant CNN classification techniques. The experimental results and supporting data show that the proposed model achieves better performance in terms of higher validation accuracy and training accuracy, as well as lower losses in training and testing, in comparison with other traditional CNN models including AlexNet and VGGNet.

With the rapid development of the network, the network security situation is increasingly severe. Network security situation prediction is widely used to obtain future situation prediction values by analyzing past network security situation sequences. There have been many studies on the specific prediction algorithms, but few studies have analyzed the predictability of the network security situation sequence itself. The central issue of this paper is to explore the inherent correlation of the given historical network security situation sequence, and then analyze the predictability of future network security situation. In this paper, the work is mainly reflected in two aspects. The first is network security situation awareness. We use the optimized KNN algorithm to classify network security situation into different types and the result shows that high accuracy rate can be obtained. This means we can obtain an accurate network security situation sequence. The second is to use the entropy theory to analyze the network security situation sequence, prove that the sequence is highly predictable and calculate the upper and lower prediction accuracy bounds.

Augmented reality technology is widely used in people's life and work, especially in visual aids. How to assist users to quickly find target objects in a complex environment is one of the problems that need to be solved in the development of augmented reality technology. To solve this problem, the paper proposes an augmented reality system based on object detection neural network and a fast method to construct specific object data set. Since it is difficult for the existing object detection network to accurately detect specific objects that are not in the data set, this paper uses the self-built data set to retrain the object detection network to ensure that the object detection network can detect specific objects. Meanwhile, we calibrate the system to ensure that the bounding box in the AR glasses output from the object detection network can accurately select the specific object. This system helps users quickly find the required items from the complex real environment and enhances the interaction efficiency between users and the real world.

Pantograph slide abrasion is one of the key factors that affect urban rail transit vehicles current-collecting quality. In order to measure the abrasion value of pantograph slide in real time, an online-image detecting method was proposed in this paper based on a vehicle-mounted panoramic camera. First, the image preprocessing methods which include enhancement, horizontal correction and morphological processing were performed on pantograph slide in images. Then, a denoising method with wavelet transform and an adaptive threshold method with the Canny operator were used to detect the slide edge. After optimizing the bi-directional edge-growth method based on image gradient, the upper and lower edge of the pantograph slide was extracted by reduplicative detection. Finally, the slide’s thickness and abrasion value could be got from the calculation of edges curve. Verified by online operation videos of TSG18G pantograph, results show that the detection method proposed in this paper can run fast enough to meet the requirement of real-time detection under the premise of boosting the accuracy of monitoring slide wear mutation.

This paper presents a noise channel model to analyses the channel characteristics between legitimate receiver and eavesdropper in quantum noise stream cipher system. The simulation results show that the channel advantage for legitimate receiver over eavesdropper can be formed by combining proper channel codes with quantum noise. Due to the fluctuation of the noise, the eye pattern degradation of the signals increased rapidly with the increasing of modulation order. In addition, the security gap in this paper is obtained by theoretical calculation. And with the increase of number of modulation order, the security gap increases significantly. After the number of modulation order goes beyond 32, the security gap is more than 0, when the number of modulation order goes up to 1024, the security gap is more than 12, which means the security and communication integration security and communication integration can be achieved.

This paper proposes a flexible spectrum access solution for unmanned aerial vehicles (UAVs) communications. The role of each UAV is defined as a secondary user and can effectively detect spectrum holes and opportunistically access the licensed spectrum band without exposing the primary user to harmful interference. We propose and formulate a multi-agent reinforcement learning method to sense the environment through which information about channel occupancy can be detected and used to gain opportunistic spectrum access. We construct a cooperative multi-agent Q-learning algorithm combined with game theory, in which each UAV acts as a learning agent that interacts with the environment through centralized control. Simulation results show that the proposed method of multi-agent learning works effectively, and the final unit channel capacity and the number of frame collisions are compared with the existing algorithms.

In this paper, a threshold selection partial transmit sequence (TS-PTS) method is proposed to reduce the peak-to-average power ratio (PAPR) in orthogonal frequency division multiplexing (OFDM) systems. In proposed TS-PTS method, two orthogonal phase factor sets are firstly introduced to produce two groups of sub-candidate signals. Then, by using the linear combinations of sub-candidate signals from different groups, the OFDM candidate signals are achieved, which avoids a lot of complex multiplications. To further reduce computational complexity, a threshold selection function Tsf(β,N) is employed based on complementary cumulative distribution function (CCDF). Simulation results show that compared with original PTS (OPTS) and the existing PTS method, the proposed TS-PTS method can gain the same PAPR performance with more computational complexity savings.

A good, reliable, and efficient communication scheme is essential to the monitoring and management of the operation status of the wind farms. At present, China lacks the communication of 10kV and below voltage level distribution networks, and there is no precedent for communication between distributed wind farms and local power grid communication dispatch centers. This paper proposes a communication scheme suitable for distributed wind farms and local dispatching centers and is practically applied to Huaneng Dingbian Langergou Wind Farms, which is the first exemplary distributed wind farms in China. Through the actual operation of the Langergou Wind Farms in recent years, the reliability of the communication scheme has been verified, which provides a new idea for the communication of distributed wind farms.

In the face of intelligent demand for information search, knowledge search is now the most promising and has been widely used in the current knowledge graph research. However, the traditional search methods based on node label only are difficult to present the semantic relationships between multidimensional nodes due to that the structure information and information in other dimensions is neglected, thus resulting in low semantic relevance and search efficiency for query results. In this paper, in order to improve the semantic relevance and quality of search results, we propose a knowledge searching algorithm based on analysis of multidimensional semantic similarity for knowledge graph systems, which combines both the ontology information and multi-hop neighborhood information together during the search process. The algorithm is designed for a knowledge graph system developed by the State Grid Anhui power distribution network, China. Several experiments are performed and the results show that the proposed algorithm outperforms the recent knowledge searching methods.

A sixteenth-mode substrate integrated waveguide (SIW) bandpass filter with slot line array is proposed in this paper. As the slow wave performance of the slot line array, the circuit size of the proposed bandpass filter is further reduced about 23% comparing with the traditional sixteen-mode one. The simulated and measured results illustrate that the operating frequency is from 1.92 GHz to 2.10 GHz with 9% bandwidth. The return loss at the central frequency is lower than 20 dB while the minimal insertion loss is 0.55 dB.

In this paper, we investigate the outage performance of an incremental decode-and-forward (DF) relaying protocol based simultaneous wireless information and power transfer (SWIPT) network, whose non-linear property is described by a piecewise linear model. Considering a non-linear energy harvesting (EH) model, we propose an optimal power splitting radio to minimize the outage probability of the destination node. On this basis, an incremental relaying protocol is proposed to enhance the system performance, which allows the source to adaptively switch between the direct link (DL) transmission mode and the cooperative relaying (CR) transmission mode via 1-bit feedback from the destination node. The closed-form expressions for the outage probability and system throughput are both derived. Numerical results show that the proposed scheme based on the non-linear energy harvesting (EH) model is able to achieve a lower outage probability and a higher throughput than the existing schemes.

Blind signal separation is a mixed source separation technique for signals that are simultaneously mixed in time, space, and frequency without the complete parameters of the mixing matrix and the source signal[11]. The subject of blind signal separation of linear instantaneous mixed signals is the most studied cases at present. In this filed, a fast independent component analysis algorithm for blind separation, namely the FastICA algorithm, is mostly studied. This algorithm not only has fast convergence speed but also good separation performance, so it is widely used in practice. Through the study of the original FastICA algorithm based on maximizing negative entropy, in order to solve the problem of large calculation amount and large number of iterations in the separation process, this paper proposes an improved FastICA algorithm by combining the joint diagonalization of fourth-order matrix cumulants with the original FastICA algorithm. The simulation results show that the improved algorithm can greatly reduce the number of operation iterations and accelerate the operation speed under the premise of ensuring the separation effect.

Betweenness is a crucial measure to evaluate the significance of the nodes and link in the networks, which is widely used in centrality ordering, failure cascading modeling and path planning. However, we have very little knowledge about the betweenness, which prevents the deep analysis. In this paper, we discovered two betweenness identities, which represent that the node and link betweenness is closely relative to the average length of the shortest paths in the network. Betweenness distribution is a globally parameter and uniquely determines the small-world property of the network.

The intelligent reflecting surface (IRS) has been recently brought up and foreseen to be one of the revolutionary technologies for future communication systems. A significant number of low-cost passive elements (PAs) can effectively reflect incident signals and collaboratively achieve beamforming by controlling phase shifts at PAs. In this paper, we study the passive beamforming problem (PBP) in the IRS-assisted multiple-input single-output (MISO) wireless system with imperfect channel station information (CSI). Considering the effect of imperfect CSI, we propose to jointly solve the non-convex beamforming problem and the CSI denoising problem with specially-designed autoencoder based deep neural networks (DNN). Moreover, instead of purely applying the DNN, we elaborate designs of the activation function, the loss function, and the feature selection. The experimental results demonstrate the advantages of the proposed method: 1) It can provide comparable performance with state-of-art approach while consuming significantly less computational time; 2) It can achieve robust beamforming over different kinds of channel estimation errors; 3) It requires no prior knowledge of the labeled targets (phase shifts), which demonstrates its potential application with CSI obtained from real environments.

With the rapid development of technology and business models, cloud computing has been widely used as a mature computing paradigm. Cloud computing provides a layered service system with its powerful computing and storage capabilities, and the emergence of 5G and edge computing has made it possible to provide end users with convenient, multi-platform, access to any time and place. Cloud games are the cloud computing applications that have emerged in recent years and are considered as a "killer". However, due to the current game delay control of cloud gaming has not reached an acceptable level, and based on hardware virtualization and cost considerations, cloud gaming related technologies are still in the exploration stage. Based on the above background, we first introduced the background and foundation of cloud games. Then we analyze the advantages of cloud gaming and the current problems. Additionally, this paper attempts to give solutions to related problems based on 5G and edge computing technology, and at the same time summarizes the research results and conclusions.

In order to reduce the pilots overhead and improve the estimation accuracy of doubly selective sky wave channel estimation based on compressed sensing (CS), a new compressed sensing estimation scheme is proposed by exploiting the temporally correlated sparsity of channel coefficient in this paper. Combining with the transmission model of High-frequency (HF) orthogonal frequency division multiplexing (OFDM) system and the sparsity of sky wave channel in transformed domain, the channel estimation problem is formulated under the framework of compressed sensing. On this basis, the temporal correlation of channel coefficient sparsity is theoretically verified by proving that the channel coefficients in several successive OFDM symbols share the common support set. Since the temporal correlation enhances the sparsity of channel coefficients in time dimension, then higher reconstruction accuracy of compressed sensing and the reducuction of pilot overhead can be expected. Simulation results verify the effectiveness of the proposed scheme.

In confronting plenty of trajectory records are generated and stored by GPS-equipped mobile devices making the device memory suffer a heavy storage pressure, it is a vital issue to compress the trajectories. This paper proposes a trajectory compression algorithm Velocity Preserving Trajectory Compression based on Retrace Point Detection (VPTR). In VPTR, the retrace points and the velocity errors are taken to reflect the speeds and directions attached with the points. VPTR first determines whether the points to be retrace points on basis of the changed movement directions, and then the retrace points are extracted from the original trajectories. Especially, the retrace points are put in a buffer, and the sub-trajectories in the buffer are compressed according to the measured velocity errors. Simulations are carried out on the Geolife trajectory dataset, and the simulation results indicate that VPTR can achieve a preferable trade-off among the compression error, compression ratio and running time.

Millimeter-wave (mmWave) massive multiple-input multiple-output (MIMO) system is a key technology of 5G communications and a research focus of 6G communications in the future. It can be widely used in communication, radar, power and transmission systems. In the mmWave MIMO system, hybrid beamforming (HBF) is regarded as a potentially promising technology, because it keeps a good balance between hardware complexity and system performance. In this paper, we propose a new hybrid beamforming scheme for the multi-user MIMO (MU-MIMO) scenario. We first design the analog beamforming by trying to avoid the loss of information to maximize the sum-rate of the communication system. Then based on the idea of successive interference cancellation (SIC), we develop the successive block diagonalization (SBD) method to eliminate interference between two adjacent users, which can avoid the computation caused by the influence between users that are far apart. Unlike the baseband BD technology, the proposed scheme has low complexity. The simulation results are shown to validate the proposed scheme.

The most common and effective attack against SDN is DOS/DDoS attacks from data plane. It takes advantage of the characteristics of communication mechanism between the SDN controllers and switches. This paper analyzes the principle and characteristics of the attacks, and proposes a detection method based on conditional random field (CRF). We firstly select several traffic statistical attributes as classification features, which have strong representativeness to the DoS attacks against SDN. Then on the basis of the CRF theory, we establish a traffic classification model. The classification model can be used for anomaly inference of real-time collected and unlabeled traffic, so as to infer whether there has DoS attack traffic. The experimental results confirm the method’s ability to detect DoS attacks against SDN.

In order to improve the reliability of low-earth-orbit (LEO) satellite communications, in this paper, we propose a forwarding protocol for cooperative LEO satellites communication system, called threshold-and-forward protocol. Note that multi-satellites can receive and forward the user's signal. After one satellite receives the signal, it compares the power of the signal with a given threshold value. If the signal is higher than the threshold value, the signal is limited to a given value. Otherwise, it is linearly processed and forwarded to the receiver. Receiver combines the signal from different satellites and tries to decode it. The P.D.F. of signal with different forwarding protocols are analyzed. Simulation results show that the performance of the proposed forwarding protocol outperforms the existing forwarding protocols such as amplify-and-forward, demodulate-and-forward and estimate-and-forward in the multi-satellites cooperative scenario.

Edge computing has emerged as a prospective paradigm to meet ever-increasing service demands in wide ranges of Internet of Things (IoT) applications. However, massive amount of application data and real-time response are not orchestrated smoothly in this pattern, especially in battery-powered and mission-critical sensor networks. In order to cope with the challenges, this paper proposes a new cooperative resource allocation algorithm which couples reinforcement learning networks and prediction neural networks for accurate mobile targets tracking. Specifically, a hierarchical structure that performs collaborative computing is designed for alleviating computing pressure of front-end devices which are supported by edge servers. The optimization objective leading to the minimal system cost is formulated which is constrained by the system response latency and the tracking accuracy. Simulation results demonstrate that the proposed algorithm outperforms the greedy scheme 24.5% in terms of the energy cost. It also obtains significant 23.1% enhancement in tracking accuracy compared with the non-cooperative scheme.

At present, with the development of the Internet of Things (IoT) and the Internet of Everything (IoE), Mobile edge computing (MEC) is proposed to provide universal and flexible computing services at the edge of the wireless access network. To use the services provided by the MEC, how to make efficient and reasonable offloading decisions is very important. In this paper, we study the problem of interference-aware multi-user computation offloading in the MEC. For this problem, we formulate a multi-user computation offloading game model and analyze the existence of nash equilibrium. Then, we design the computation offloading algorithm (e.g., Nash-Q-learning) based on nash equilibrium and reinforcement learning to minimize system overhead. Furthermore, in order to avoid Nash-Q-learning suffering from dimensional disaster, we get a deep reinforcement learning algorithm (e.g., Nash-DQN) by adding the neural networks to Nash-Q-learning. The performance of the proposed algorithms has been compared with the other algorithms by simulation. The simulation results show that the performance of the proposed algorithms is superior to the other multi-user computation offloading algorithms.

In this paper, the massive random access scenario that a part of users transmit both common and user-specific information and the others transmit only common information is considered, and a mixed massive random access scheme is proposed. In the proposed scheme, the common information is transmitted by unsourced random access (URA) technique, and the user-specific information by index-modulated (IM)-aided sourced random access (SRA). Specifically, the concatenated scheme consisting of inner compressed sensing codebook and outer parity-check code is employed in URA, and IM-aided non-coherent grant-free non-orthogonal multiple access scheme is employed in SRA. The URA and SRA signals are superposed at the transmitter of users who transmit both common and user-specific information. At the receiver, first, the covariance-based non-Bayesian multiuser detection is employed to estimate the indices of active codewords related with common information in each block, and both user activity and transmit data related with user-specific information. Further, the estimated indices of active codewords in different blocks are stitched based on the outer-code parity to recover the common information. Finally, simulation results show that the proposed mixed scheme can work well in the system without channel state information including large- and small-scaling fading coefficients at the receiver.

PIN is widely used as an authentication method on the mobile phone in various public places. However a malicious person nearby may peek into the screen and remember the password. This kind of shoulder surfing attack is more difficult to prevent in modern society. A stronger adversary may use his camera or video surveillance system to record the entry process. In this paper, we present three approaches to solve this problem. Those approaches utilize the phone’s vibration as the covert channel to deliver challenge- response secretly from camera. The experimental results show that those approaches can effectively prevent the shoulder surfing attack, even if the password entry process is completely exposed to the adversary's camera.

In this paper, a spectral efficiently adaptive modulation DFT-spread polar coordinate-based orthogonal frequency division multiplexing (P-OFDM) scheme is proposed and demonstrated in a hybrid fiber-visible laser light communication (fiber-VLLC) system. Adaptive modulation scheme and DFT-spread is used to reduce the PAPR and the impact of imbalanced SNR distribution. In contrast, the polar coordinate-based transform is used to obtain real and non-negative symbol to realize full subcarrier use. After 10km SSMF and 5m FSO transmission, the simulation results show that the proposed adaptive 128/64QAM DFT-spread P-OFDM scheme improves both the BER and spectral efficiency in VLLC system.

The broadcast characteristic of wireless network makes the secure access of device face various threat. Among them, the mimicry attack is a kind of fraud attack in which the malicious users imitating the MAC (Media Access Control) address or other identification credentials of the legitimate user. To cope with it, this paper proposes a mimicry attack detection method based on the received signal strengths (RSSs) of multiple receivers. Different from the MAC address and other identification credentials, RSS is difficult to forge as it is related to the physical location of the transmitter. By analyzing the RSS vector of the target device, the system can detect the identity change of the access device without relying on the content of the packet, and thereby detect the mimicry attack. The experiment results on the implemented ZigBee platform show that the farther the distance between malicious devices and legitimate devices, the slower the movement speed transmitter, the higher the recognition rate of the proposed method. When four ZigBee devices are used as receivers, and the distance between the malicious device and the legitimate device is more than 5 meters, the detection rate of the proposed method is more than 92%. This method can be used as a powerful supplement to the traditional authentication system based on MAC address to further enhance the security of wireless communication.

In recent years, coflow size-based scheduling has become a hotspot in data center networks since it can effectively optimize the coflow completion time (CCT). However, it is a challenging work to predict the coflow size accurately and rapidly with limited information. This paper proposes a Broad Learning System based coflow size inferring model, which employs Multi -RELIEF method to extract features and trains BLS model on Facebook logs with incomplete information. Experiment results show that the accuracy and sensitivity of BLS-based inferring model are higher than the ELM and XGBoost methods by 31.6% when we use only half of coflow information, and the prediction time remains below 6ms that meets the online processing requirement of most data center networks. Besides, the simulated scheduling results indicate that the CCT of the BLS-based coflow scheduler outperforms the FIFO, Aalo and MCS schedulers at most time.

This paper describes a 12-bit 100KS/s cyclic analog-to-digital convertor (ADC) for X-Ray detector readout circuit implemented in SMIC 0.13μm CMOS process. The ADC adopted operational amplifier sharing technology to reduce the number of capacitance and operational amplifier. At different cycle stages, the gradually decreasing tail current of the op amp reduces the power consumption of the cyclic ADC. With the proposed clock, the kickback noise of dynamic comparator has no influence on sampling signal. The simulation result shows that the signal-to-noise-plus-distortion (SNDR) is 70.4dB, the spurious free dynamic rang (SFDR) is 84.2dB, the total harmonic distortion (THD) is -82.2dB, the effective number of bits (ENOB) is 11.39bit, and the power consumption of the ADC is 112μW.

The task of semantic segmentation is to correctly classify every pixel of one image. Benefit from the full convolutional neural network (FCN), the image segmentation task has step into a new stage. Since Google has shown its exploration of semantic segmentation, and proposes EncoderDecoder algorithm with Atrous Separable Convolution (Deeplab_v3_plus) method for enhancing the performance of image segmentation. Following the previous work, we will lose a great deal of detail information in the process of downsampling. Therefore, we propose to supplement the compressed channel information into the different levels feature map to compensate the details lost in the process of down sampling. This paper introduces how do we compress channel information and channel information fusion. We present a Channel Compression-Encoder-Decoder with Atrous Separable Convolution Net (CC-Deeplab_v3_plus), Eventually our CC-Deeplab_v3_plus relative to Deeplab_v3_plus in the PASCAL VOC2012 and own pest datasets get 1% Mean Intersection Over Union (MIUO) boost.

This paper focuses on the task planning problem for unmanned aerial vehicle (UAV) swarm. Considering the diverse requirements of different tasks, the variety of the execution capabilities of UAVs, and the effect of task execution sequence, the multi-type task allocation (MTTA) model is proposed with the objective of maximizing the global task execution reward, subject to the maximum flight length and limited operation ability of different UAVs, as well as the resource requirements and execution time window of different tasks. Moreover, the auction mechanism-based task allocation (AMTA) algorithm is proposed to allocate the conflict-free task sequences for UAVs. Simulation results demonstrate that the proposed AMTA algorithm outperforms the reference algorithm on performance in terms of total reward, path length, completion time, and energy consumption.

The continuous development of wireless data service has brought great pressure to the limited spectrum resources. Moreover, increasing the utilization rate of authorized frequency bands has its upper limit to the increase of capacity and transmission rate, and it is expensive to solve the situation of information explosion in time. In order to overcome such challenges brought by explosive data traffic demand and scarce frequency spectrum resources, the deployment of Long Term Evolution (LTE) to the unlicensed frequency bands is regarded as an effective way for largely improving the spectrum efficiency. The inefficient utilization of the spectrum leads to the idle space in the channel known as white space. And the white space of WiFi offers LTE considerable spectrum opportunity at no cost. Based on Multi Armed Bandit (MAB) algorithm for white space time estimation, this paper proposes an LTE scheduling algorithm based on knapsack algorithm. The results show that using this method can efficiently improve the spectrum resource efficiency, enlarge the transmission rate and thus expand the overall capacity of the system.