The challenges in current WiFi based gait recognition models, such as the limited classification ability, high storage cost, long training time and restricted deployment on hardware platforms, motivate us to propose a lightweight gait recognition system, which is named as B-Net. By reconstructing original data into a frequency energy graph, B-Net extracts the spatial features of different carriers. Moreover, a Balloon mechanism based on the concept of channel information integration is designed to reduce the storage cost, training time and so on. The key benefit of the Balloon mechanism is to realize the compression of model scale and relieve the gradient disappearance to some extent. Experimental results show that B-Net has less parameters and training time and is with higher accuracy and better robustness, compared with the previous gait recognition models.
Authors: Yangjie Cao (School of Software, Zhengzhou University), Zhiyi Zhou (School of Software,Zhengzhou University), Pengsong Duan (School of Software, Zhengzhou University), Chao Wang (School of Sofeware,Zhengzhou University), Xianfu Chen (VTT Technical Research Centre of Finland),
Hide Authors & Abstract

Deep learning is one of the most effective and efficient methods for facial emo-tion recognition, but it still encounters stability and infinite feasibility problems for faces of different races. To address this issue, we proposed a novel bottleneck feature extraction (BFE) method based on the deep neural network (DNN) model for facial emotion recognition. First, we used the Haar cascade classifier with a randomly generated mask to extract the face and remove the background from the image. Second, we removed the last output layer of the VGG16 transfer learning model, which was applied only for bottleneck feature extraction. Third, we de-signed a DNN model with five dense layers for feature training and used the fa-mous Canade-Kohn dataset for model training. Finally, we compared the pro-posed model with the K-nearest neighbor and logistic regression models on the same dataset. The experimental results showed that our model was more stable and could achieve a higher accuracy and F-measure, up to 98.59%, than other methods.
Authors: Tian Ma (Xi’an University of Science and Technology, China), Kavuma Benon (Xi’an University of Science and Technology, China), Bamweyana Arnold (Xi’an University of Science and Technology, China), Keping Yu (Waseda University, Japan), Yan Yang (Xi’an University of Science and Technology, China), Qiaozhi Hua (Hubei University of Arts and Science, China), Zheng Wen (Waseda University, Japan), Anup Paul (East West University, Bangladesh),
Hide Authors & Abstract

In recent years, with the prominent population aging problem, health conditions of aged solitaries are inherently gaining more and more attentions. Among the techniques allowing real-time health monitoring, activity perception has become an important and promising field in both academia and industry. In this paper, a human activity perception recognition model, named MSHNet (Multi-Stream-Hybrid-Network) based on Deep Learning is proposed to solve the problems of difficulty in extracting perceptual features of Wi-Fi signals and low recognition accuracy in traditional Machine Learning methods. MSHNet adopts passive wireless sensing technology, it uses commercial off-the-shelf Wi-Fi devices to collect Channel State Information (CSI) based on underlying physical equipment and automatically extracts human activity features characterized by amplitude in CSI. Then MSHNet aggregates the data streams of the same receiving antenna using the wireless signal transceiving characteristics of Multiple Input Multiple Output (MIMO) and trains the aggregated data streams respectively. At last, the voting mechanism is adopted to select the best training result. The experimental results demonstrate that MSHNet’s results on the public dataset have reached the state-of-the-art and on the datasets of four environments collected by ourselves the average recognition accuracy rate has reached 97.41%, satisfying the daily activity monitoring of the elderly, especially those living alone.
Authors: Fuchao Wang (School of Software, Zhengzhou University), Pengsong Duan (School of Software, Zhengzhou University), Yangjie Cao (School of Software, Zhengzhou University), Jinsheng Kong (School of Software, Zhengzhou University), Hao Li (School of Software, Zhengzhou University),
Hide Authors & Abstract

Accurate demand prediction of bike-sharing is crucial to reduce the cost of bicycle scheduling and improve users’ satisfaction with social public services. It is difficult to make the prediction enough accurate due to the randomness and nonlinearity in bike-sharing systems. In this paper, we propose a prediction model based on pseudo-double hidden layer feedforward neural network. We improve extreme learning machine with adaptive particle swarm optimization further. Finally, on the basis of fully mining the massive operational data of "Shedd Aquarium" station in Chicago (USA), we provide a case to verify the performance by comparing with other neural networks and other learning rules.
Authors: Fan Wu (Anhui University of Technology), Si Hong (Anhui University of Technology), Wei Zhao (Anhui University of Technology), Xiao Zheng (Anhui University of Technology), Xun Shao (Kitami Insitute of Technology), Wen Qiu (School of Regional Innovation and Social Design Engineering, Kitami Insitute of Technology),
Hide Authors & Abstract

In recent years, with the growth of mobile network traffic and the development of various power interconnection services in heterogeneous smart grids, video traffic optimization in smart grids has attracted extensive research. Caching video at the edge of smart grid is a common optimization method. Transcoding cached video can meet the needs of different power interconnection services, thus effectively improving the cache hit rate. This paper proposes a cooperative transcoding and distribution mechanism based on overlapping clustering. First, the effect of computing and channel resources on the initial latency of the video is studied. Secondly, the model of topology and resource distribution in the network is established, with the goal of minimizing the initial buffer time of the video, the overall network delay optimization problem is divided into two sub-problems of grouping and resource allocation. Thirdly, an algorithm combining KM matching and genetic algorithm is proposed. Nodes in an ultra-dense network are overlapping clustered by KM matching algorithm, and genetic algorithm is used to allocate heterogeneous resources of the grouped base stations. Finally, the simulation results show that the proposed algorithm optimizes the user's average initial buffer delay.
Authors: Shuiyao Chen (State Grid Zhejiang Electric Power Co., Ltd., Hangzhou, China), DI ZHAI (State Grid China), Wei Bai (Global Energy Interconnection Research Institute. co.Ltd), Haochen Guan (BUPT), Ping Ma (State Grid Shaoxing Electric Power Supply Company Shaoxing, China), Weiping Shao (State Grid Zhejiang Electric Power Co., Ltd., Hangzhou, China),
Hide Authors & Abstract

Edge computing can reduce service latency through task offloading. Since computational resources on the edge of the network are scarce, selecting a node with rich computational capability is the key for getting a high-quality service. In this paper, we propose a virtual edge scheme where a node can offload its tasks to a virtual edge node that consists of multiple vehicles in vicinity. The relative vehicle velocity is considered in the virtual edge selec-tion. We compare our proposed scheme with several baseline schemes and show the superiority of the scheme.
Authors: Narisu Cha (The University of Electro-Communications), Celimuge Wu (The University of Electro-Communications), Tsutomu Yoshinaga (The University of Electro-Communications), Yusheng Ji (National Institute of Informatics),
Hide Authors & Abstract

Mobile Edge Computing (MEC) is regarded as a promising technology that migrates cloud computing platforms with computing and storage capabilities to the edge of the wireless access network, enabling rich applications and services in close proximity to the mobile users (MUs). There are a lot of literature that have studied computation offloading. Different from them, this paper performs a novel research on multi-level computation offloading taking account into the heterogeneity of computation tasks and computation resource backup pool, and introducing opportunistic networks in multi-access networks simultaneously. Firstly, we describe the computation offloading model. Then, we formulate the multi-level computation offloading problem as a Stackelberg game and demonstrate the existence of the game Nash equilibrium. In order to solve above problem, we design a global optimal algorithm based on game theory. Finally, the performance of the proposed algorithm is verified by comparing with other algorithms. Simulation results corroborate that the algorithm can not only decrease the energy consumption, but also is stable.
Authors: Wei Bai (Global Energy Interconnection Research Institute. co.Ltd), Yang Lu (Global Energy Interconnection Research Institute Co., Ltd., Beijing, China), Donglei Zhang (Global Energy Interconnection Research Institute Co., Ltd., Beijing, China), Jiawei Li (BUPT), Ping Ma (State Grid Shaoxing Electric Power Supply Company Shaoxing, China), Shuiyao Chen (State Grid Zhejiang Electric Power Co., Ltd., Hangzhou, China), Weiping Shao (State Grid Zhejiang Electric Power Co., Ltd., Hangzhou, China),
Hide Authors & Abstract

With the continuous development and improvement of 5G networks, many emerging technology architectures have been introduced to support 5G service requirements. As one of them, mobile edge computing can meet the exponentially increasing computing requirements, and with its advantages of being more efficient, smarter, and more flexible, it can be well adapted to smart grid scenarios. However, most of the existing research contents of the eIoT focus on the research of computing offloading and content caching separately, ignoring the problem of reusability of some computing results. This paper considers the certain content caching capabilities of the MEC system itself, and aims to design a cache-enhanced MEC eIoT . The model includes offloading, calculating and backhaul for uncached task and downloading of cached content. On the other hand, the problem of task diversity and inspection robot mobility is fully analyzed. Subsequently, we studied the impact of caching capabilities on computing power to get the best MEC server parameter information. Based on the above research, this paper proposes a cache enhanced offload strategy and a collaborative scheduling algorithm to optimize the total delay of all tasks of the inspection robot in the eIoT .Simulation results show that the strategy can effectively reduce the computational offloading latency.
Authors: Zhifeng Li (State Grid Jibei Tangshan Power Supply Company,Tangshan, China), Jie Bai (State Grid Jibei Tangshan Power Supply Company,Tangshan, China), Haonan Zhang (BUPT), Wei Bai (Global Energy Interconnection Research Institute. co.Ltd), Yongmin Cao (State Grid Jibei Tangshan Power Supply Company,Tangshan, China), Liwen Wu (State Grid Jibei Tangshan Power Supply Company,Tangshan, China), Jianying Dong (State Grid Jibei Tangshan Power Supply Company,Tangshan, China), Yanshan Deng (State Grid Jibei Tangshan Power Supply Company,Tangshan, China),
Hide Authors & Abstract

At present, the crowdsourcing-based indoor navigation system has attracted extensive attention from both the industry and the academia. The crowdsourcing-based indoor navigation system commendably solves the dilemma (e.g., high cost, etc.) of traditional navigation methods. Unfortunately, the system that relys on crowdsourced data is vulnerable to the collusion attack that threaten the security of the system. In this paper, we propose a crowdsourcing-based secure indoor navigation system. Specially, we first propose a new reputation mechanism. Then, we employ the offensive and defensive game to model the interactions between the fog service platform and responders. The optimization problem of the system is established to maximize the total utility of the system. Finally, the simulation results verify that the proposed mechanism can effectively encourage responders to choose the non-collusion strategy.
Authors: Liang Xie (School of Mechatronic Engineering and Automation, Shanghai University, China), Zhou Su (-Shanghai), Qichao Xu (School of Mechatronic Engineering and Automation, Shanghai University, China),
Hide Authors & Abstract

Smart grid is an energy network which integrates advanced power equipment, communication technology and control technology. The existing smart grid communication technologies include power line carrier communication, Industrial Ethernet, passive optical networks and wireless communication, which have different advantages. Due to the complex application scenarios, massive sampling points and high transmission reliability requirements, one single communication method can’t fully meet the communication requirements of smart grid, and heterogeneous communication modes are required. Besides, with the development of cellular technology, LTE based standards have been identified as a promising technology to meet the strict requirements of various operations in smart grid. In this paper, we analyzes the advantages and disadvantages of PLC and LTE communication, and designs a network framework for PLC and LTE communication uplink heterogeneous communication in smart grid, then puts forward an uplink scheduling transmission method of sampling data with optimized throughput according to the requirements of system delay and reliability. Then we use the formula derivation to prove the stability and solvability of the scheduling system in theory. Finally, the simulation results show that under the condition of satisfying the delay requirement, our proposed framework can optimally allocate the wireless communication resource and maximize the throughput of the uplink transmission system.
Authors: Qiyue LI (Hefei University of Technology), Tengfei CAO (Hefei University of Technology), Wei SUN (Hefei University of Technology), Weitao LI (Hefei University of Technology),
Hide Authors & Abstract

In this paper, a distributed power and spectrum allocation scheme is proposed for Device-to-Device communication on unlicensed bands (D2D-U) enabled networks. To make full use of the spectrum resources on the unlicensed bands while guaranteeing the fairness among D2D-U links and the harmonious coexistence with WiFi networks, an online trained Neural network (NN) is first utilized on each D2D-U pair to determine the price to use the unlicensed channels according to the channel state and traffic loads. Then, a non-convex optimization problem can be formulated and solved on each D2D-U link to determine the optimal spectrum and power allocation scheme which can maximize the its transmission data rate. Numerical simulation results are demonstrated to verify the performance of the proposed method which enables each D2D-U link to maximize its own data-rate individually under the constraint of the fair coexistence with other D2D-U devices and WiFi networks.
Authors: Zhiqun Zou (College of Information Science and Electronic Engineering, Zhejiang University), Rui YIn (College of Information & Electronic Engineering, Zhejiang University City College), Celimuge Wu (The University of Electro-Communications, Graduate School of Informatics and Engineering), Jiantao Yuan (Institute of Ocean Sensing and Networking of the Ocean College, Zhejiang University), Xianfu Chen (VTT Technical Research Centre of Finland),
Hide Authors & Abstract

In the future mobile communication system, inter-cell interference becomes a serious problem due to the intensive deployment of cells and terminals. Traditional interference coordination schemes take long time for optimization in ultra-dense networks. Meanwhile, due to the increase of factors affecting communication and in order to better meet the communication needs of each terminal, an interference coordination scheme needs to fully consider multiple characteristic parameters of the terminal, which will further increase the scheme's computational time. Therefore, we should compress all the data through sparsification of parameters before optimization. There are many terminal parameters, and the essence of sparsification of parameters is to rank terminals. In this paper, a graph-based terminal ranking scheme is designed. First, each terminal can be represented by its multiple parameters. Then, all terminals are used as the vertexes in the graph to form a complete weighted graph, and edge weights represent the degree of dissimilarity between terminals. A ranking of terminals is obtained by finding a minimum Hamiltonian path in the graph. Finally, the ranking of all parameter sequences is obtained according to terminals ranking, which makes the sparsity of all parameter sequences better. Simulation results show that the proposed scheme can accomplish sparsification of multiple parameters effectively and keep good performance in terms of system capacity and fairness.
Authors: Junxiang Gou (Hefei University of Technology), Lusheng Wang (Hefei University of Technology), Hai Lin (Key Laboratory of Aerospace Information Security and Trusted Computing, Ministry of Education. School of Cyber Science and Engineering, Wuhan University, Wuhan, China), Min Peng (Anhui Province Key Laboratory of Industry Safety and Emergency Technology, School of Computer Science and Information Engineering, Hefei University of Technology, Hefei, China),
Hide Authors & Abstract

For an RFID system that contains a large population $S$ of $N$ tags, the tag-information sampling problem is to randomly choose $K$ distinct tags from $S$ to form a subset $T$, and then inform each tag in $T$ of a unique integer from $\{1,2,...,K\}$. This is a fundamental problem in many real-time analysis applications in RFID systems. Because it enables rapidly selecting a random subset $T$ and collecting the tag-information from $T$ (typically used for estimating the characteristics or status of $S$). However, existing protocols for this problem are far from satisfaction because of their high communication cost. In this paper, our objective is to solve this problem by using a minimal amount of communication cost. We first obtain a lower bound on communication cost, denoted by $C_{\rm{lb}}$, on this problem. Then we design a protocol, denoted by $P_{\text{s}}$, to solve this problem, and prove that the communication cost of $P_{\text{s}}$ stays within a factor of $1.88$ of $C_{\rm{lb}}$. Extensive simulations verifies the advantages of $P_{\text{s}}$ comparing with other protocols.
Authors: Xiujun Wang (Anhui University of Technology), Yan Gao (Anhui University of Technology), Yangzhao Yang (Research Institute of Cyberspace Security of CETC, Beijing, China), Xiao Zheng (Anhui University of Technology), Xuangou Wu (Anhui University of Technology), Wei Zhao (Anhui University of Technology),
Hide Authors & Abstract

In this letter, a new space time (ST) coding scheme is proposed to improve the system spectral efficiency (SE). To improve the SE while achieving full diversity gain, the ST coded symbol transmission sequence (denoted as ST coding pattern in time domain) is exploited to improve the transmission rate. Since the orthogonal construction of the ST codes is preserved, the simple decoding scheme (linear maximum likelihood detector) is still applicable. Based on the analysis and Monte Carlo simulations, we demonstrate that the data rate increases by 25% and the its bit error rate (BER) is close to the conventional ST codes when 16-QAM or 16-PSK is used in the system.
Authors: Rui YIn (College of Information & Electronic Engineering, Zhejiang University City College), Zhiqun Zou (Zhejiang University, Information Science and Electronic Engineering), Celimuge Wu (The University of Electro-Communications, Graduate School of Informatics and Engineering), Hongjun Xu (University of KwaZulu-Natal, The School of Electrical, Electronic and Computer Engineering), Chao Chen (Zhejiang Gongshang University, School of Information Science and Electronic Engineering),
Hide Authors & Abstract

In this paper we describe an airport ground movement surveillance network. Airport ground videos are captured by multiple cameras, and than transmitted to the airport control center based on a dedicated optical network. On the high-performance servers in the control center, various intelligent applications process video data, visualize the processing results and provide them to the air traffic controllers as a reference for airport management. Moving object detection is the foundation of many video based intelligent applications in airport surveillance. We propose detecting the moving objects in the airport ground by the use of the prior knowledge, that is, the airport ground made of cement has a gray-white color distribution. Based on this fact, firstly we use a dual-mode Gaussian distribution to fit the color distribution of the ground. Next, based on the fitted distribution we build a prior model, where pixels near the class boundary are more likely to be classified as the foreground. Finally, the prior model is used to detect moving targets within a Bayesian classification framework. Experiments are conducted on the AGVS benchmark and the results demonstrate the effectiveness of the proposed moving object detection algorithm.
Authors: Xiang Zhang (University of Electronic Science and Technology of China), Yi Qiao (The Second Research Institute of Civil Aviation Administration of China),
Hide Authors & Abstract

There has recently been an increasing number of blockchain applications in different realms. Among the popular blockchain technologies, Ethereum is an emerging platform featuring smart contracts with the public Ethereum associated to the Ether currency. Besides, the private Ethereum has been gaining interest due to its applicability to the Internet of Things. An Ethereum blockchain network includes distributed records that are immutable and transparent through replicating among network nodes. Ethereum manages information in blocks that are submitted to the chain as transactions. This paper aims to characterize latency performance in the private Ethereum blockchain network. Initially, we clarify two perspectives of latency according to the lifecycle of transactions (transaction-oriented and block-oriented latency). We then construct a real private blockchain network with a laptop and Raspberry Pi 3b+ for the latency measurement. We write and deploy a smart contract to read and write data to the blockchain and measure the latencies in a baseline and realistic scenario. The experiment results reveal the latencies-hop correlation, as well as the latencies' relation in different workloads. Moreover, the blockchain network spends averagely 63.92 milliseconds (except the mining time) to take one transaction into effect in one hop.
Authors: Xuan Chen (Graduate School of Engineering, Chiba University), Kien Nguyen (Graduate School of Engineering, Chiba University), Hiroo Sekiya (Graduate School of Engineering, Chiba University),
Hide Authors & Abstract

With the explosive growth of mobile video services, unmanned aerial vehicle (UAV) is flexibly deployed as a relay node to offload cellular traffic or provide video services for emergency scenario without infrastructures. In this paper, we propose a novel design framework for UAV enabled video relay system to minimize the UAV’s energy consumption while guaranteeing the QoE requirement of each GU. A dynamic resource allocation strategy is employed to model the bandwidth and power allocation of the UAV and the optimization problem is formulated as a non-convex problem, via optimizing the transmit power and bandwidth allocation of the UAV jointly with the UAV trajectory. To tackle this non-convex problem, the original problem is decoupled into two sub-problems: transmit power and bandwidth allocation optimization, as well as UAV trajectory optimization. We propose an efficient iterative algorithm to obtain a Karush-Kuhn-Tucker (KKT) solution via solving the two sub-problems with successive convex approximation and alternating optimization techniques. Simulations are conducted to corroborate our study and the results demonstrate that with the proposed joint design, the UAV’s energy consumption can be significantly reduced, by up to 30%, and the QoE requirement for GUs can be well satisfied simultaneously.
Authors: Han Hu (Beijing Institute of Technology), Cheng Zhan (Southwest University), Jianping An (Beijing Institute of Technology),
Hide Authors & Abstract

In smart grid systems, heterogeneous networks are considered as a promising solution to address the expeditious growth of mobile traffic. Considering the different user preferences, how to efficiently utilize the limited resource of small base stations (SBSs) becomes a challenge. In this paper, we investigate a joint handover and transmission strategy for users. We formulate the handover and transmission problem to minimize the video delivery delay, which considers the overlapped coverage of SBSs and the limited capacity of backhaul link. To solve this NP-hard problem, we design a heuristic algorithm with two phases. In content caching phase, the video encoding layers are cached at SBSs according to the greedy algorithm. In content delivery phase, a transmission strategy is designed to meet the user demands for videos with different quality level. Simulation results show that our proposed algorithm has the advantage of reducing video delivery delay and saving the backhaul traffic compared with other algorithms.
Authors: Weiping Shao (State Grid Zhejiang Electric Power Co., Ltd., Hangzhou, China), Donglei Zhang (Global Energy Interconnection Research Institute Co., Ltd., Beijing, China), Yang Lu (Global Energy Interconnection Research Institute Co., Ltd., Beijing, China), Yazhou Wang (BUPT), Wei Bai (Global Energy Interconnection Research Institute. co.Ltd), Ping Ma (State Grid Shaoxing Electric Power Supply Company Shaoxing, China), Shuiyao Chen (State Grid Zhejiang Electric Power Co., Ltd., Hangzhou, China),
Hide Authors & Abstract

Traditional medicine requires doctors and patients to have face-to-face palpation, which is hugely inconvenient for under-developed areas, primarily rural areas. Telemedicine provides an opportunity for patients using mobile devices or the Internet to connect with doctors who maybe thousands of miles away. In this way, elastography is a crucial medical imaging modality that maps the elastic properties of soft tissue, which will be sent to doctors remotely. Among them, ultrasound elastography has become a research focus because it can accurately measure soft tissue lesions. Also, displacement estimation is a key step in ultrasound elastography technology. The phase-zero search method is a popular displacement estimation method with accurate and rapid features. However, the method is powerless when the displacement is more than 1/4 wavelength. The block-matching method can make up for this shortcoming. It is suitable for large displacement, although it is not accurate. Notably, the quality-guided block matching method has good robustness under the complex mutational condition. In this paper, we propose a displacement estimation method, which combines the block-matching method and the phase-zero search method. The block-matching method provides prior knowledge to increase the robustness of phase-zero search under the condition of large displacement. The experimental results show that our proposal has stronger robustness, more accurate results, and faster calculation speed.
Authors: Hong-an Li (College of Computer Science and Technology, Xi'an University of Science and Technology, Xi'an 710054, China), Min Zhang (College of Computer Science and Technology, Xi'an University of Science and Technology, Xi'an 710054, China), Keping Yu (Global Information and Telecommunication Institute, Waseda University, Tokyo 169-8050, Japan), Xin Qi (Global Information and Telecommunication Institute, Waseda University, Tokyo 169-8050, Japan), Li Zhen (School of Communications and Information Engineering, Xi'an University of Posts and Telecommunications, Xi'an 710121, China), Yi Gong (School of Information and Communication Engineering, Beijing University of Posts and Communications (BUPT), Beijing, China), Jianfeng Tong (School of Information Science and Technology, Northwest University, Xi'an 710127, China),
Hide Authors & Abstract