Communications and Networking

Image aesthetic assessment is a challenging problem in the field of computer vision. Recently, the input size of images is often limited by the network of aesthetic problems. The methods of cropping, wrapping and padding unify images to the same size, which will destroy the aesthetic quality of the images and affect their aesthetic rating labels. In this paper, we present an end-to-end deep Multi-Task Spatial Pyramid Pooling Fully Convolutional Neural NasNet (MTP-NasNet) method for image aesthetic assessment that can directly manipulate the original size of the image without destroying its beauty. Our method is developed based on Fully Convolutional Network (FCN) and Spatial Pyramid Pooling (SPP). In addition, existing studies regards aesthetic assessment as a two-category task, a distribution predicting task or a style predicting task, but ignore the correlation between these tasks. To address this issue, we adopt the multi-task learning method that fuses two-category task, style task and score distribution task. Moreover, this paper also explores the reference of information such as variance in the score distribution for image reliability. Our experiment results show that our approach has significant performance on the large-scale aesthetic assessment datasets (AVA [1]), and demonstrate the importance of multi-task learning and size preserving. Our study provides a powerful tool for image aesthetic assessment, which can be applied to photography and image optimization field.

This work aims to study the source localization problem using a symmetric array in a near-field environment. To reduce the computational complexity, in this work, two spatial correlation signals are created in which each signal only depends on one parameter of direction of arrival (DOA) or range. In the development process, the each resulting signal still possesses the array spatial structure, and therefore, the atomic norm minimization is utilized to obtain the corresponding solutions. The utilization of atomic norm also allows one to avoid the off-grid problem when the sparse reconstruction concept is employed. The numerical studies demonstrate the proposed method provides a superior performance compared with other approaches.

Understanding the inner working mechanism of deep neural networks (DNNs) is essential and important for researchers to design and improve the performance of DNNs. In this work, the entropy analysis is leveraged to study the neurons activation behavior of the fully connected layers of DNNs. The entropy of the activation patterns of each layer can provide an efficient performance metric for the evaluation of the network model accuracy. The study is conducted based on a well trained network model. The activation patterns of shallow and deep layers of the fully connected layers are analyzed by inputting the images of a single class. It is found that for the well trained deep neural networks model, the entropy of the neuron activation pattern is monotonically reduced with the depth of the layers. That is, the neuron activation patterns become more and more stable with the depth of the fully connected layers. The entropy pattern of the fully connected layers can also provide guidelines as to how many fully connected layers are needed to guarantee the accuracy of the model. The study in this work provides a new perspective on the analysis of DNN, which shows some interesting results.

Recently, owing to graceful performance degradation for various wireless channels, analog visual transmission has attracted considerable attention. The pioneering work about analog visual communication is SoftCast, and many advanced works are all based on the framework of SoftCast. In this paper, we propose a novel analog image communication system called CSCast based block compressive sensing. Firstly, we present the system framework and detailed design of CSCast, which consists of discrete wavelet transform, power scaling, compressive sampling and analog modulation. Furthermore, we discuss how to determine the appropriate value of scaling factor $$\alpha $$ in power allocation, and block size of measurement matrix in compressive sampling. Simulations show that the performance of CSCast better than Softcast in all SNR range, and better than Cactus in high SRN range. In particular, CSCast outperforms over Softcast about 1.72 dB. And CSCast achieves the maximum average PSNR gain 1.8 dB over Cacuts and 2.03 dB over SoftCast when SNR = 25 dB, respectively. In addition, our analyses shows CSCast can save about 75% overhead comparing to SoftCast and Cactus.

Device-to-Device (D2D) communication has been recognized as a promising technology in 5G. Due to its short-range direct communication, D2D improves network capacity and spectral efficiency. However, interference management is more complex for D2D underlaying cellular networks compared with traditional cellular networks. In this paper, we study channel allocation in D2D underlaying cellular networks. A tier-based directed weighted graph coloring algorithm (TDWGCA) is proposed to solve cumulative interference problem. The proposed algorithm is composed of two stages. For the first stage, the tier-based directed weighted graph is constructed to formulate the interference relationship among users. For the second stage, the maximum potential interference based coloring algorithm (MPICA) is proposed to color the graph. Different from the hypergraph previously investigated in channel allocation, our proposed graph reduces the complexity of graph construction significantly. Simulation results show that the proposed algorithm could better eliminate cumulative interference compared with the hypergraph based algorithm and thus the system capacity is improved.

Orthogonal frequency division multiplexing (OFDM) system is very sensitive to the phase noise especially in high frequency since the orthogonality between sub-carriers is easily destroyed. It is very important to estimate and compensate the phase noise in the research of 5G systems. The influence of phase noise on OFDM systems is manifested in two aspects: introducing common phase error (CPE) and causing inter-carrier interference (ICI). In this paper, we propose a new joint channel and CPE estimation algorithm to obtain more accurate channel and CPE estimates through iterations. In each iteration, we update the channel and CPE estimates to make them closer to the true value. Besides, the performance improvement brought by the algorithm under the simplified system model is analyzed. Simulation results show that this algorithm has a great impact on improving the accuracy of channel and CPE estimation.

With the development of multimedia technologies, VR services have quickly gained popularity at an accelerating speed. To reduce the high cost of purchasing high-performance VR terminals for end users and to enhance the user experience, recently, the concept of cloud-based VR was proposed which brings the cloud computing technologies to VR services. On-cloud GPU clusters and multi-core servers are expected to be used for simplifying VR terminals at the users’ side. This idea, however, arises several challenges in deploying such cloud-based VR system for practical applications, among which the cloud-to-end latency is mainly concerned. In this paper, we designed a practical solution for bearing cloud-based VR applications. We aim at reducing the cloud-to-end latency to improve the experience of end users. In our system, a frame splitting technique was proposed to fulfill the goal. Specially designed algorithms including reference frame determination and rate control strategies were also included to limit the computational complexity and improve the coding efficiency while obtaining promising user experience. Experimental results showed that the proposed system can significantly reduce the cloud-to-end latency.

A panorama is a wide-angle view picture with high-resolution, usually composed of multiple images, and has a wide range of applications in surveillance and entertainment. This paper presents a end-to-end real-time panoramic face detection video system, which generates panorama video efficiently and effectively with the ability of face detection. We fix the relative position of the camera and use the speeded up robust features (SURF) matching algorithm to calibrate the cameras in the offline stage. In the online stage, we improve the parallel execution speed of image stitching using the latest compute unified device architecture (CUDA) technology. The proposed design fulfils high-quality automatic image stitching algorithm to provide a seamless panoramic image with 6k resolution at 25 fps. We also design a convolutional neural network to build a face detection model suitable for panorama input. The model performs very well especially in small faces and multi-faces, and can maintain the detection speed of 25 fps at high resolution.

Some traditional Wi-Fi indoor passive human detection systems only extract the coarse-grained statistical information such as the variance, which leads to low detection accuracy and poor adaptability. To solve the problem, we propose a new coherence histogram for Wi-Fi indoor passive people detection. In the histogram construction process, the method leverages time continuity relationship between received signal strength (RSS) measurements. The coherence histogram captures not only the occurrence probability of signals but also the time relationship between adjacent measurements. Compared to statistical features, the coherence histogram has more effective fine-grained information. The feature vector consists of coherence histograms is used to train the classifier. To eliminate the position drift problem, the Allen time logic helps to establish the transfer relationship between the sub-areas, we correct the results to improve the location accuracy. Compared with the classic passive human detection technology, the F1-measure is improved by nearly 5%.

The convolutional neural network (CNN) decoder for general convolutional decoding is proposed. The parameters of CNN are determined by the initial state of each input block and the constraint relationship between adjacent bits is extracted by the convolutional layer as the constraint features. Then CNN decoder realizes decoding process through the extracted constraint feature instead of codewords directly. The result shows that, without changing the structure of decoder, the decoding performance of CNN decoder on different convolutional codes is equivalent to Viterbi soft decoding algorithm. Compared with Viterbi decoding, the larger constraint length or the lower SNR, the greater gain can be obtained in CNN decoder. Besides, we consider CNN trained by the two kinds of training sets in order to further investigate the potential and limitations of CNN decoder with respect to decoding performance, analysing the advantages and factors of these two kinds of training sets.

The K-Nearest Neighbor (KNN) algorithm is widely used in practical life because of its simplicity and easy understanding. However, the traditional KNN algorithm has some shortcomings. It only considers the number of samples of different classes in k neighbors, but ignores the distance and location distribution of the unknown sample relative to the k nearest training samples. Moreover, classes imbalance problem is always a challenge faced with the KNN algorithm. To solve the above problems, we propose an improved KNN classification method for classes imbalanced datasets based on local distance mean and centroid (LDMC-KNN) in this paper. In the proposed scheme, different numbers of nearest neighbor training samples are selected from each class, and the unknown sample is classified according to the distance and position of these nearest training samples. Experiments are performed on the UCI datasets. The results show that the proposed algorithm has strong competitiveness and is always far superior to KNN algorithm and its variants.

Applying recommendation algorithms in mobile edge caching can further improve the utilization of the caching and relieve the pressure of the backhaul links. The key is to capture accurate user preferences which are usually influenced by the user’s request record and current request. In this paper, we propose a content recommendation algorithm based on both history request record and current interest. The content, user preferences and user’s requests are modeled as vectors from multiple content dimensions. Based on user’s request record, we capture the user preferences vector (Pre-Vector) by using the maximum likelihood estimation. The Pre-Vector accurately reflects user preference but has hysteresis. The user current request vector (Req-Vector) can reflect the user’s current interest but its accuracy is not stable. We propose the preference-based recommendation list and the request-based recommendation list based on the Pre-Vector and the Req-Vector respectively. In order to ensure the accuracy of the recommendation list, the final recommendation list is generated based on the Pre-Vector and the Req-Vector’s cosine similarity. The simulation results show that, the proposed algorithm has improved caching hit rate compared with existing recommendation algorithms.

With the development of the Internet of Things, a large number of connection requirements for sensing and control are generated. However, in wireless positioning, Narrowband Internet of Things (NB-IoT) has poor positioning accuracy which takes the cell-ID positioning method. The further integration of 5G and NB-IoT networks is expected to effectively improve the positioning accuracy of NB-IoT networks. Therefore, the high-precision positioning algorithm for researching converged networks has broad application prospects and academic significance. In order to improve the positioning accuracy of NB-IoT, based on the 5G and NB-IoT heterogeneous positioning framework, we propose to introduce a number of cluster nodes, which have the function of communicating with 5G and NB-IoT networks simultaneously. The signal bandwidth in NB-IoT network is narrow and clock synchronization is difficult to accomplish, so only DOA (Direction of Arrival) and RSSI principles can be considered. In this paper, we firstly use 5G to perform high-precision positioning of cluster nodes according to the principles of TDOA (Time Difference of Arrival). Based on the solution space (x $$ \pm $$ εx, y $$ \pm $$ εy), the NB-IoT terminal is located by the cluster nodes according to the DOA and RSSI fusion method. This method helps reduce the matching time and improve the accuracy of single DOA/RSSI positioning method. Meanwhile, in the case of allowing cluster node errors, higher precision NB-IoT network positioning results can be obtained. Compared to a single NB-IoT network positioning, the final positioning accuracy of NB-IoT terminal can be improved by 80–90%.

Although the Global Positioning System (GPS) can provide more accurate outdoor positioning services, it cannot detect the signals in indoor environments or in densely populated areas. Therefore, indoor positioning service has gradually been paid more attention. Most researchers currently use a nine-axis inertial sensor for indoor positioning. However, when the object is moving fast and frequently, it is obvious that using nine-axis inertial sensor has a large amount of computation. In addition, Kalman filtering algorithm is always cumbersome when data fusion is carried out for inertial sensors. The use of zero-velocity update algorithm (ZVU) to improve double integral can reduce the cumulative error, but the degree is far from enough. This paper mainly completes the following works: Firstly, the six-axis inertial sensor is used for indoor positioning. Then the digital motion processor is used instead of Kalman filter for attitude solution. Lastly, ZVU is optimized. Specifically, in the six-axis inertial sensor, the three-axis accelerometer is used to measure the force of the object, and the three-axis gyroscope is used to detect the current posture of the object. Since the three-axis magnetometer is missing, it is possible to effectively reduce a part of the calculation amount. In addition, the digital motion processor is used instead of the Kalman filter for the attitude solution, which avoids cumbersome filtering and data fusion. Finally, we optimize the ZVU so that the cumulative error is reduced again. The experimental results show that the algorithm proposed in this paper has certain feasibility and practical application value.

The rapidly increasing content fetching requirements pose challenges to the transmission performance of traditional cellular system. Due to the limited transmission performance of cellular links and the caching capabilities of the base stations (BSs), it is highly difficult to achieve the quality of service (QoS) requirements of multi-user content requests. In this paper, a joint user association and content placement algorithm is proposed for cellular device-to-device (D2D) communication network. Assuming that multiple users located in a specific area may have content requests for the same content, a clustering and content placement mechanism is presented in order to achieve efficient content acquisition. A joint clustering and content placement optimization model is formulated to minimize total user service delay, which can be solved by Lagrange partial relaxation, iterative algorithm and Kuhn-Munkres algorithm, and the joint clustering and content placement strategies can be obtained. Finally, the effectiveness of the proposed algorithm is verified by MATLAB simulation.

With the rapid development of emerging smart homes applications, the home security systems based on passive detection without carrying any devices has been increasing attention in recent years. Through-The-Wall (TTW) detection is a great challenge since through-the-wall signal can be severely attenuated, and some of the existing TTW-based detection techniques require special equipment or have strict restrictions on placement of devices. Due to the near-ubiquitous wireless coverage, WiFi based passively human detection technique becomes a good solution. In this paper, we propose a robust scheme for device-free Through-the-wall Human Detection (T-HuDe) in TTW with Channel State Information (CSI), which can provide more fine-grained movement information. Especially, T-HuDe utilizes motion information on WiFi signal and uses statistical information of motion characteristics as parameters. To evaluate T-HuDe performance, we prototype it in different environments with commodity devices, and the test results show that human activity detection rate and human absence detection rate of T-HuDe are both above 93% in most detection areas.

This paper considers a suspicious communication network with multiple suspicious source-destination nodes and multiple wireless powered legitimate eavesdroppers, where the legitimate eavesdroppers are assumed to be collusive or non-collusive. A minimum harvested energy constraint is applied at each eavesdropper such that each eavesdropper must harvest a minimum required energy. The legitimate eavesdropping in such a scenario is investigated and our aim is to maximize the average successful eavesdropping probability by optimizing the power splitting ratio at each eavesdropper under the minimum harvested energy constraint. The optimal algorithm is proposed to solve the optimization problem for both collusive eavesdroppers and non-collusive eavesdroppers. Simulation results show that the proposed algorithm achieves the upper bound of the successful eavesdropping probability when the energy harvesting efficiency is large, the required minimum harvested energy is small, or the transmit power of the suspicious source node is high.

The emergence of the Internet of Things (IoT) has promoted the interconnection of all things. And the access control of devices and accurate service promotion are inseparable from the acquisition of location information. We propose WiHlo, a passive localization system based on WiFi Channel State Information (CSI). WiHlo directly estimates the human location by refining the angle-of-arrival (AoA) of the subtle human reflection. WiHlo divides the received signals into static path components and dynamic path components, and uses phase offsets compensation and direct wave suppression algorithms to separate out the dynamic path signals. By combining the measured AoAs and time-of-arrivals (ToAs) with Gaussian mean clustering and probability analysis, WiHlo identifies the human reflection path from the dynamic paths. Our implementation and evaluation on commodity WiFi devices demonstrate WiHlo outperforms the state-of-the-art AoA estimation system in actual indoor environment.

Given a piece of acoustic musical signal, various automatic music transcription (AMT) processing methods have been proposed to generate the corresponding music notations without human intervention. However, the existing AMT methods based on signal processing or machine learning cannot perfectly restore the original music signal and have significant distortion. In this paper, we propose a novel processing method which integrates various AMT methods so as to achieve better performance on music transcription. This integrated method is based on the entropic regularized Wasserstein Barycenter algorithm to speed up the computation of the Wasserstein distance and minimize the distance between two discrete distributions. Moreover, we introduce the proportional transportation distance (PTD) to evaluate the performance of different methods. Experimental results show that the precision and accuracy of the proposed method increase by approximately 48% and 67% respectively compared with the existing methods.

Non-coherent ultra-wideband (UWB) systems have attracted great attention due to their low complexity, and without the need of channel estimation. In order to improve the transmission reliability, polar codes were recently introduced into non-coherent UWB systems because of their capability of approaching the Shannon channel capacity, and their low complexity in both coding and decoding. In the case of polar codes with medium and short length, the bit error rate (BER) performance of coded incoherent UWB systems is limited to incompletely channel polarization, poor Hamming distance and the sensitivity of successive cancellation (SC) decoding resulting in error propagation. In order to improve the performance of coded systems using polar codes with medium and short length, Spinal-Polar codes were recently presented, in which inner codes and outer codes are complementary, and the outer codes have good pseudo-random characteristics and error correction performance in the case of short length. Therefore, in this paper, the interleaved Spinal-Polar codes are introduced into the non-coherent UWB systems. Simulation results show that the interleaved Spinal-Polar codes can effectively improve the BER performance of the coded non-coherent UWB systems using polar codes with medium and short code length.

In ultra dense networks (UDNs), base stations (BSs) with mobile edge computing (MEC) function can provide low latency and powerful computation to energy and computation constrained mobile users. Meanwhile, existing wireless access-oriented mobility management (MM) schemes are not suitable for high mobility scenarios in UDNs. In this paper, a novel dynamic programming based MM (DPMM) scheme is proposed to optimize delay performance considering both wireless transmission and task computation under an energy consumption constraint. Based on markov decision process (MDP) and dynamic programming (DP), DPMM utilizes statistic system information to get a stationary optimal policy and can work in an offline mode. Cooperative transmission is further considered to enhance uplink data transmission rate. Simulations show that the proposed DPMM scheme can achieve close-to-optimal delay performance while consume less energy. Moreover, the handover times are effectively reduced so that quality of service (QoS) is improved.

High definition (HD) map attracts more and more attention of researchers and map operators in recent years and has become an indispensable part for autonomous or assistant driving. Different from existing navigation map, HD map has the features of high precision, large-volume data and real-time update. Therefore, the real-time HD map transmission to the vehicles becomes one main challenge in vehicular networks. This paper considers the scenario that a RSU at the crossroad caches and transmits HD maps to its covered vehicles in four directions. To reduce the average delay of HD map delivery, the transmission power allocation for vehicles and the cache allocation for HD maps of different road segments are optimized by leveraging the traffic density and vehicle positions. Simulation results indicate that the proposed scheme has lower latency than that of equal power allocation scheme based on real traffic data.

Under natural disaster or other emergency situations, the fixed communication infrastructures are unavailable, which brings great inconvenience to information interaction among people. In this paper, we design a UAV relay network, using a small-scale UAV fleet serves as communication relays of a team of ground users performing collaborate tasks. Aiming at the user’s requirement for high communication capacity for multi service transmission, we present a distributed gradient-based algorithm of finding the optimal positions of UAV in UAV relay network to improve the network average end-to-end throughput in real-time. The system optimization objective is formulated by using Shannon-Hartley Theorem and received signal-to-noise ratio (SNR) that incorporates with UAV positions and ground user positions. Due to the non-smoothness of the objective function, we use generalized gradient instead. Each UAV moves along the generalized gradient direction of objective function to optimize the target locally, and finally, all UAV convergence to stable positions of optimizing the network throughput. Simulation results show the effectiveness of our method in improving the network average end-to-end throughput.

Aiming at the contradiction between the convergence rate and steady state mean square error of adaptive filter based on Maximum Versoria Criterion (MVC), this paper introduces the multi-convex combination strategy into MVC algorithm, and proposes a multi-convex combination MVC (MCMVC) algorithm. Simulation results show that compared with the existing MVC algorithm, MCMVC algorithm can select the best filter more flexibly under different weight change rates, and thus it has faster convergence speed and stronger tracking ability. Moreover, compared with the existing multi-convex combination maximum correntropy criterion (MCMCC) algorithm, MCMVC algorithm not only ensures the tracking performance, but also has lower exponential computation and steady-state error.

Most k-anonymization mechanisms that have been developed presently are vulnerable to re-identification attacks, e.g., those generating a generalized value based on input databases. k-anonymization mechanisms do not properly capture the notion of hiding in a crowd, because they do not impose any constraints on the mechanisms. In this paper, we define $$(k,\rho )$$-anonymization that achieves secure k-anonymization notion linked with differential identifiability under the condition of privacy parameter $$\rho $$. Both differential identifiability and k-anonymization limit the probability that an individual is re-identified in a database after an adversary observes the output results of the database. Furthermore, differential identifiability can provide the same strong privacy guarantees as differential privacy. It can make k-anonymization perform securely, while $$(k,\rho )$$-anonymization achieves the relaxation of the notion of differential identifiability, which can avoid a lot of noise and help obtain better utility for certain tasks. We also prove the properties $$(k,\rho )$$-anonymization under composition that can be used for application in data publishing and data mining.

Energy Harvesting Cognitive Wireless Sensor Network (EH-CRSN) is a novel network which introduces cognitive radio (CR) technology and energy harvesting (EH) technology into traditional WSN. Most of the existing works do not consider that battery capacity of the sensor is limited and will decay over time. Battery capacity degradation will reduce the lifetime of the sensor and affect the performance of the network. In this paper, in order to maximize the network utility of the energy harvesting sensor node in its life cycle, we are concerned with how to determine the optimal sampling rate of sensor node under the condition of battery capacity degradation. Therefore, we propose an optimal adaptive sampling rate control algorithm (ASRC), which can adaptively adjust the sampling rate according to the battery level and effectively manage energy use. In addition, the impact of link capacity on network utility is further investigated. The simulation results verify the effectiveness of the algorithm, which shows that the algorithm is more realistic than the existing algorithm. It can maximize the network utility and improve the overall performance of the network.

With the advent of new Global Navigation Satellite Systems (GNSS) and signals, the signal quality monitoring techniques for navigation signals also need to be updated. In the traditional satellite signal integrity detection, the multi-correlator processing method is commonly used in signal quality monitoring to detect if a signal is distorted. This method often assumes that multipath signals have been eliminated, avoiding multipath signals from interfering with the detection results. However, if there is a multipath signal that has not been eliminated, since the correlation functions of the multipath signal and the distorted signal have a certain similarity, if the detection method without considering the multipath effect is used, here is a case where the multipath signal is erroneously detected as a distorted signal. Since the influence of the multipath signal and the distorted signal on the positioning result is very different, it is necessary to distinguish the two signals during the detection process. In this paper, the model of multipath signal and distorted signal is discussed for the new generation GNSS signal (BOC signal). Based on the characteristics of the correlation functions of these two models, a multi-correlator range setting method is proposed, and the appropriate detection values are selected, which can effectively distinguish multipath signals and distorted signals at the relevant peak levels.

Cognitive vehicular ad-hoc networks (CVANETs) are expected to improve spectrum utilization efficiently and offer both infotainment and safety services for vehicles. In this paper, the joint route selection and resource allocation problem is considered for CVANETs. Taking into account the lifetime of transmission links, we first propose a candidate link selection method which selects the transmission links satisfying the link lifetime constraint. Then stressing the importance of transmission delay, we formulate the joint route selection and resource allocation problem as an end-to-end transmission delay minimization problem. As the formulated optimization problem is a complicated integer nonlinear problem, which cannot be solved conveniently, we equivalently transform the original problem into two subproblems, i.e., resource allocation subproblem for candidate links and route selection subproblem. Solving the two optimization subproblems by applying the K shortest path algorithm and the Dijkstra algorithm, respectively, we can obtain the joint route selection and resource allocation strategy. Simulation results demonstrate the effectiveness of the proposed algorithm.

In recent years, machine-to-machine (M2M) communications have attracted great attentions from both academia and industry. In M2M communication networks, machine type communication devices (MTCDs) are capable of communicating with each other intelligently under highly reduced human interventions. In this paper, we address the problem of joint resource allocation and clustering for M2M communications. By defining the system energy efficiency (EE) as the sum of the EE of MTCDs, the joint resource allocation and clustering problem is formulated as a system EE maximization problem. As the original optimization problem is a nonlinear fractional programming problem, which cannot be solved conveniently, we transform it into two subproblems, i,e., power allocation subproblem and clustering subproblem, and solve the two subproblems by means of Lagrange dual method and modified K-means algorithm, respectively. Numerical results demonstrate the effectiveness of the proposed algorithm.

The relationship between the latency and reliability of multi-antena diversity system is investigated in this paper. The system performance of diversity system is analysed with the outage probability chosen as the reliability metric. Two combining techniques are considered in the diversity system. It is proved that the latency-reliability trade-off degree (LRTD), i.e., the slope of the latency-outage curves with logarithmic scales, equals the number of the diversity order. In addition, the diversity system with considering system overhead is investigated. Golden section search algorithm and a simplified iterative method can be used to obtain the optimum diversity order of multiple-input and single-output (MISO) system adopted with maximal ratio combining and section combining techniques, respectively.

The rapid development of automatic control, wireless communication and intelligent information processing promotes the prosperity of unmanned aerial vehicles (UAVs) technologies. In some applications, UAVs are required to fly from given source places to certain destinations for task execution, a reasonable path planning and charging stations (CSs) strategy can be designed to achieve the performance enhancement of task execution of the UAVs. In this paper, we consider joint UAV path planning and CS deployment problem. Stressing the importance of the total time of the UAVs to perform tasks and the cost of deploying and maintaining CSs, we formulate the joint path planning and CS deployment problem as a cost function minimization problem. Since the formulated optimization problem is an NP-hard problem which cannot be solved easily, we propose a heuristic algorithm which successively solves two subproblems, i.e, path planning subproblem and destination path selection subproblem by applying the A* algorithm, K-shortest path algorithm and genetic algorithm (GA), respectively. Simulation results validate the effectiveness of the proposed algorithm.

Mobile edge computing (MEC) is regarded as an emerging paradigm of computation that aims at reducing computation latency and improving quality of experience. In this paper, we consider an MEC-enabled heterogeneous cellular network (HCN) consisting of one macro base station (MBS), one small base station (SBS) and a number of users. By defining workload execution cost as the weighted sum of the energy consumption of the MBS and the workload dropping cost, the joint computation offloading and resource allocation problem is formulated as a workload execution cost minimization problem under the constraints of computation offloading, resource allocation and delay tolerant, etc. As the formulated optimization problem is a Markov decision process (MDP)-based offloading problem, we propose a hotbooting Q-learning-based algorithm to obtain the optimal strategy. Numerical results demonstrate the effectiveness of the proposed scheme.

With the rapid development of artificial intelligence, gesture recognition has become the focus of many countries for research. Gesture recognition using Wi-Fi signals has become the mainstream of gesture recognition because it does not require additional equipment and lighting conditions. Firstly, how to extract useful gesture signals in a complex indoor environment. In this paper, after de-noising the signal by Discrete Wavelet Transform (DWT) technology, Principal Component Analysis (PCA) is used to eliminate the problem of signal redundancy between multiple CSI subcarriers, further to remove noise. Secondly, the frequency domain features of the gesture signal are constructed by performing Short-Time Fourier Transform (STFT) on the denoised CSI amplitude signal. Then, the time domain features are combined with the frequency domain features, and the features are trained and classified using the Support Vector Machine (SVM) classification method to complete the training and recognition of gesture. The experimental results show that this paper can effectively identify gestures in complex indoor environments.

For the difficulty of separation of accompaniment from mono music, image filtering was applied into a novel approach to separate accompaniment music. Our approach presents how single channel music manifests in the 2D Fourier Transform spectrum. In image domain, the position of periodic peak energy was determined by image filtering, and then masking matrix was constructed by rectangular window to extract the constituent of the accompaniment music. We find that our system is more robust and very simple to describe. The simulation experiments show that the method in this work has an advantage over other separation algorithm.

In order to solve the problem of broadcast storm and broadcast unreliability in Vehicular Ad Hoc Networks (VANET) on highways, an improved algorithm based on Speed Adaptive Probabilistic Flooding (SAPF) [1], which is referred to as Average Speed Based Broadcast (ASBB), is proposed. Since the average speed of vehicles in the vicinity reflects the network congestion around the current node more accurately, ASBB dynamically calculates the forwarding probability according to the average speed of the current node and the corresponding neighbor nodes. To obtain the speed of neighbor nodes, each node encapsulates its speed into the header of packets it transmits, instead of employing new types of packet for exchanging speed. This approach alleviates the network load and reduces the complexity of implementation. Meanwhile, only the nodes located behind the current node may participate in the forwarding of the broadcast packet, which reduces the number of nodes participating in the forwarding and further mitigates the broadcast storm and improves the broadcast reliability. The simulation results show that ASBB performs well in terms of suppressing broadcast storms, increasing the reachability and reducing the end-to-end delay.