Search

Hybrid precoding is a credible solution to address the trade-off between spectral efficiency and power consumption in mmWave massive MIMO systems. Though traditional hybrid precoding with fixed architectures has reached its limits, the flexibility of hybrid precoding has not been fully exploited, which gives insight to improve the performance further. In this paper, we adopt a novel architecture called dynamic-subarray, where the connection between radio-frequency (RF) chains and antennas can be dynamically changed by switches. We propose a low-complexity scheme to design the hybrid precoder with dynamic subarrays. Simulation results demonstrate that our proposed method has an excellent performance on the achievable rate and energy efficiency even with lower complexity.

In this paper, we consider a distributed rate control problem with two selfish users. One user focuses on optimizing the data freshness while the other one aims at optimizing the “power” of throughput and the transmission delay. Two users aim at achieving their own goals by controlling the source sending rate. We formulate the problem as an uncooperative rate control game with two users and show that the Nash Equilibrium (NE) of the game is inefficient compared to the global optimum solution. To improve the efficiency, we propose a linear packet-dropping incentives in the uncooperative rate control game with two users, so that users are encouraged to control their rates near the global optimum sending rates. We show that the proposed incentives can improve the Price of Anarchy (PoA) performance.

Machine vision has become the development trend of strip surface defect detection technology because of its intelligence, real-time, high precision, reliability and other advantages. In this technology, due to the environmental interference of the production line, instability of optical elements and other factors, the surface defect image often with the problems such as uneven illumination, high noise, low contrast between the defect and background, which is not conducive to the subsequent processing. In order to solve these problems, this paper first normalizes the defect image by gray scale, then introduces the adaptive threshold gray transformation into the enhancement of the defect image on the steel surface. Finally, the image is processed by Gabor filter and image segmentation. The experimental results show that the image enhancement algorithm based on adaptive threshold gray transformation can effectively improve the contrast, enhance the image quality and facilitate subsequent processing.

As industrial production becomes more modern and intelligent today, the inspection of product quality of the workshop is becoming more and more accustomed to replacing the old manual visual inspection methods with automated inspection systems. In the welding field, automated welding robots are not only used in traditional large-scale automobile assembly lines. In more general welding work, welding robots also plays an important role. The inspection of the welding quality of the welding robot is mainly to detect the four main types of weld defects. Compared to traditional defect classification based on support vector machines and defect detection based on template matching, this paper uses a welding surface defect detection system designed based on deep learning methods. By working with workshop welding experts, a large-scale image of nearly 5000 pictures is built. Large-scale weld defect datasets, while using the real-time and accuracy of the YOLO series of deep learning object detection frameworks, the weld defects detection model reaches 75.5% mean average precision(mAP) in constructed weld defect data set. In addition, the construction cost of the detection model and the deployment time of the detection system are greatly reduced. During the field test of the system in the workshop, among a batch of welding workpieces provided by the factory, the detection accuracy of weld defects reached 71%, which initially met the requirements of the workshop for an automated defect detection system.

Fusion algorithm for foggy image enhancement based on transmittance weight factor is proposed. Firstly, the rough transmittance image of foggy image is obtained by the dark channel prior dehazing algorithm, and then the fine transmittance image is obtained by the guided filter. The dehazing image is obtained by the dark channel prior dehazing method based on the guided filter, and the enhanced image is obtained by histogram equalization based on the HSI color model. Finally, the weight factor is constructed by the fine transmittance, the above dehazing image and enhanced image are processed by weighted fusion. The algorithm gives full play to the respective advantages of two kinds of algorithms in foggy image processing, and significantly improves the fog removal and enhancement effect of foggy image.

Focused on the issues that the sparseness of the data obtained after factorization is reduced and with the increasing of training samples, the computing scale increases rapidly, an incremental learning algorithm based on graph regularized non-negative matrix factorization with sparseness constraints was proposed. It not only considers the geometric structure in the data representation, but also introduces sparseness constraint to coefficient matrix and combines them with incremental learning. Using the results of previous factorization involved in iterative computation with sparseness constraints and graph regularized, the cost of the computation is reduced and the sparseness of data after factorization is highly improved. Experiments on both ORL and PIE face recognition databases demonstrate the effectiveness of the proposed method.

At present, the development of IoT has been highly concerned by the domestic and even the world. The IoT has become a new engine of economic growth. The emergence and development of 5G technology have widely expanded the application fields of IoT technology. The intelligent feeding system is one of the application fields of IoT technology. The current intelligent feeding system is characterized by providing services based on multiple systems. In order to expand the range of services and enable multiple intelligent feeding systems to accept the same services, a standardized platform is needed and services are provided on that platform. This paper considers providing a service layer using a platform based on the oneM2M standardization protocol. Based on this platform, an intelligent feeding system is designed. In addition, we implement the intelligent pets feeding system based on the oneM2M, and use this system as an example.

Aimed at the problems of traditional data feature recognition schemes such as feature extraction difficulties, poor classification and recognition effects, and high sensitivity, a data feature recognition scheme based on deep learning is proposed to realize feature extraction and classification of the data information automatically. By analyzing the characteristics of the data and the problems of the deep neural network, the attention mechanism is introduced in the convolutional neural network to extract more important networks, and the long and short-term memory network is improved, reducing the complexity of the model, and avoiding the problem of information loss caused by long sequences.

As one of the key features in 5G network, Millimeter wave (mmWave) technology can provide the ultra-wide bandwidth to support higher data rate. However, for high frequency band, mmWave signal still suffers from the high pathloss, the multipath fading and the signal blockage issue, especially in the indoor environment. For different application scenarios, the channel conditions and quality of services (QoS) are quite different. Therefore, it is essential to investigate the impact of the mmWave channel on the system performance. This paper investigates and measures the performance of a 60GHz mmWave module that is exploited for the downlink and uplink high data rate transmission in the Internet of Radio-Light (IoRL) project. The coverage area and the throughput of the mmWave module is estimated by measuring the error vector magnitude (EVM) of received signals with different transmitter (TX) and receiver (RX) angles and at different locations in a laboratory. In this paper, the measurement environment and system setup are introduced. After that, the waveform design for the measurement is also discussed. The measurement results show that this 60GHz mmWave module can provide an acceptable performance only in some cases, which restricts its application scenarios.

In this work, we consider a scenario where a time sensitive source is broadcasted to multiple receivers by a base station (BS) over unilateral networks. A recently proposed metric-the Age of Information (AoI) is adopted to measure data freshness from the perspective of receiver. Unlike previous work, we consider that the BS receives no feedback from receivers and thus broadcasts every fixed interval. We derived the optimum fixed interval such that the average AoI can be minimized. Our work suggests that, when the transmission delay is highly random, the optimum fixed interval is larger than the expected transmission delay so that success delivery to more users can be guaranteed. Both theoretical analysis and simulation results show that without feedback, the average AoI performance following the proposed policy is near to transmission policy that utilizes receiver feedback.

This paper introduces a 8K ultra-high definition digital television transmission system based on DTMB-A. The system transmits 8K digital TV programs through two DTMB-A signals, and the data rate can reach 80 Mb/s. The bit error rate of the system is measured to be zero, and digital TV signals with 8K resolution can be transmitted normally without mosaic.

MmWave MIMO communications has been recognized as one revolutionary new technology in wireless communication area, which can support wide ban width and high data rate transmission compared with conventional communications. Recently, hybrid precoding is widely utilized to reduce the number of RF chains and power consumption, which has perfect performance on spectral efficiency (SE) and energy efficiency (EE). In this paper a new GenSM-aided low-complexity hybrid precoding algorithm is proposed, which enjoys higher spectral efficiency and lower complexity compared with the existed low-complexity methods.

With the development of artificial intelligence, convolutional neural networks (CNN) have been widely used in image processing and other aspects due to their excellent performance. However, as a computationally intensive algorithm, CNN face huge challenges in the realization of mobile devices. FPGA have the advantages of high performance, reprogramming, and low power consumption, and have becoming suitable choices for CNN deployment. Compared with various CNN algorithms, the YOLO algorithm regards target detection as a regression problem. It is a one-step algorithm with fast execution speed and small amount of calculation. It is suitable for implementation on FPGA hardware platforms. This paper proposes an improved algorithm for deep learning YOLO network based on Xilinx ZYNQ FPGA. By optimizing the YOLO network model and fixed-point, etc., the problem of large computational of CNN and limited resources on FPGA chips is solved, and the parallelism of FPGA is used to accelerate the CNN. Experimental results show that the method proposed in this paper has greatly improved the operation rate while maintaining accuracy, and has important practical value in the realization of mobile terminals of CNN and real-time computing.

In order to improve the network learning effect, based on the existing target learning framework MobileNet-V2, an ELU function is inserted as an activation function. Firstly, use the expansion convolution to increase the number of channels to get more features to activate and output through the ELU function, which can alleviate the disappearance of the gradient of the linear part, the nonlinear part is more robust to the noise of the input change. Then, the way of the residual connection combine high-level features with low-level features and then output. Finally, output to softmax using global pooling. The experimental data shows that compared with the current mainstream lightweight deep learning target recognition algorithm,E-MobileNet has improved the accuracy of recognition and the frame rate per second in the same test environment of the same test set. The experimental data fully demonstrates that the use of the ELU activation function and the global pooling layer reduces the number of parameters, enhances the generalization ability of the model, and improves the robustness of the algorithm. On the basis of ensuring the light weight of the neural network model, the recognition accuracy of the target is effectively improved.

We study a network with a central controller collecting random updates from power-limited sensors. The time-varying channels between sensors and the central controller are modeled as ergodic Markov chains while packet-loss may happen due to decoding error. We measure the data freshness from the central controller by the metric Age of Synchronization (AoS), i.e., the average time elapsed since information about a sensor becomes desynchronized. To minimize the average AoS under all aforementioned bandwidth and power constraints, we first relax the hard bandwidth limit and decouple the multi-sensor problem into a single-sensor constrained Markov decision process (CMDP), which is then solved through linear programming (LP). We then propose an asymptotic optimal scheduling policy to solve the original hard-bandwidth-constrained problem. It is revealed that sensors are more likely to send updates under better channel states and higher AoS to save energy and avoid packet-loss.

In the future, Drones or general Unmanned Aerial Vehicles(UAVs) need to implement online real-time object detection, but their memory and computing power are limited. An improved YOLOv3 is adopted in this paper, making full use of the lightweight network instead of the feature extraction network of YOLOv3 to achieve high-efficiency object detection for remote sensing images. In the case of similar detection accuracy, parameters and FLOPs (floating point operations) of the improved model are 2.5 times and 3.3 times smaller than YOLOv3, respectively. In addition, an IoU K-medians algorithm is proposed, which improves mAP by 7.0% on YOLOv3 and by 2.3% on the improved YOLOv3.Experiments show that the detection speed of the improved YOLOv3 is 101 frame/s at the fastest, and it is still 1.6 times faster than that of YOLOv3 when its mAP is 6% higher than that of YOLOv3. The efficient remote sensing object detection method proposed in this paper lays the foundation for future UAVs.

Face liveness detection is the first stage of the whole face detection technology, and it is of great significance to the system security using face recognition technology. In this paper we use convolutional neural networks to extract facial features. Compared to ResNet, SqueezeNet and VGG Face networks, the network structure is lighter and the model training takes less time. After testing, it achieved an accuracy of 99.5% on the Nanjing University of Aeronautics and Astronautics (NUAA) face detection data set of China Southern Airlines. In order to improve the robustness of the model, we built our own dataset through collecting pictures from internet and camera shooting, and horizontally compared the accuracy of the above several models on the self-built dataset. Finally the model proposed in this paper achieved the highest 99.02% accuracy and achieved the best results in actual testing.

With the rise of intelligent and connected vehicles (ICVs), new vehicle applications continue to emerge, while the computing capability of vehicles remains limited. Mobile edge computing (MEC) is considered to be the most effective technique for mitigating vehicle computing pressure, with computation offloading being a key technology for MEC. To solve the problem of excessive task processing delay and energy consumption due to the vehicle-limited computing power in the vehicular network, we consider the tasks and the characteristics of MEC, and divide the tasks into indivisible tasks and divisible tasks according to the size of data (that is, whether it affects functionality after segmentation). Then, two computation offloading algorithms are proposed named binary offloading and partial offloading separately. The binary offloading unloads the task to the mobile edge computing server as a whole and selects only an optimal offloading site; thus, an improved upper confidence bound algorithm is adopted. The partial offloading divides the complex tasks with large data volumes through time slots processed by different MEC servers, and uses the Q-learning algorithm to find the most effective offloading strategy. The simulation results show that the total cost of delay and energy consumption of the binary offloading algorithm is lower when processing computationally intensive tasks. When addressing divisible and complex tasks, the partial offloading algorithm improves the real-time performance of the tasks significantly and conserves the energy of the vehicle terminal.

With the rapid development of modern scientific research, the research of infrared technology has achieved a lot of results, and infrared detectors and other devices have also been widely used. However, infrared images have characteristics such as low contrast, unclear information, and blurred edges, which seriously affect the observation effect of infrared images. In order to better achieve observation and monitoring, it is necessary to optimize and enhance the acquired images. In this paper, the infrared image acquired by FLIR-Tau camera is taken as the research object. According to the characteristics of infrared image and the structure of convolutional neural network, an improved super-resolution convolutional neural network algorithm is proposed to train the algorithm model for infrared image. The infrared image is enhanced by the learning characteristics of the convolutional neural network, which is better than the traditional algorithm in contrast enhancement and detail information enhancement.

In this paper, we present an experimental testbed of a Visible Light Communication (VLC) - based localization framework developed for the 5G Internet of Radio-Light (IoRL) European project. The established testbed is capable of providing decimeter-level positioning accuracy to support high level location-based services (LBS) as well as the data transmission service for the future IoRL system. The emulation results on the testbed confirm that the proposed system achieves high location accuracy with most positioning errors less than 0.15 m and only a few more than 0.25 m corresponding to the corner area.

Digital terrestrial/television multimedia broadcasting - advanced (DTMB-A), the new generation of digital terrestrial television broadcasting (DTTB) system proposed by China, has been officially accepted by the International Telecommunication Union (ITU) as the international DTTB standard in 2015. Compared with DTMB standard, DTMB-A has higher spectrum efficiency, stronger robustness and more flexible system parameter combinations. In August 2019, at the invitation of Television Broadcasts Limited (TVB), Tsinghua University organized the field trials to evaluate the coverage effect of DTMB-A in Hong Kong. By setting up the single frequency network (SFN) with 3 transmitting stations, 4K UHDTV program can be transmitted by DTMB-A effectively. In this paper, the DTMB-A SFN structure and measurement results in Hong Kong are analyzed in detail.

Vanadium Diselenide (VSe 2 ) has a high electrical conductivity owing to the electron delocalization of the vanadium lattice framework. Most of previous research has been focused on its electronic properties. The nonlinear optical properties of VSe 2 have not been investigated yet. Herein, we prepared few-layer VSe 2 nanoflakes by liquid phase exfoliation. Subsequently, we utilized Z-scan technique to study the nonlinear optical properties of few-layer VSe 2 nanoflakes. And we prepared VSe 2 based saturable absorbers (SA) and achieved passively mode-locked operation in Er-doped fiber laser. The results indicate that VSe 2 nanosheets own remarkable nonlinear optical properties and could be employed as ideal saturable absorption materials in pulsed fiber lasers.

In order to solve the problem of high time complexity and low generalization of traditional methods in the human-machine collaboration scene, this paper applies the YOLOv3 deep learning network to the part of workpiece recognition and detection of the robot workpiece capture. According to the specific application scenarios, the corresponding data set is created to train the YOLOv3 model, and the anchor value suitable for the data set is obtained by the iterative self-organizing data analysis(ISODATA) clustering algorithm. A systematic and comprehensive data augmentation of the data set is carried out for the case where the self-made data set is small and the scene is single. Considering that the target to be detected is small and the background of the detection scene is simple, the YOLOv3 basic network architecture is appropriately pruned. Combining the shallow features with the deep features makes the detection time reduced 4ms while the accuracy of the model is basically unchanged. The comparison experiment on the self-made dataset shows that the improved YOLOv3 algorithm has a mean average precision(mAP) of 0.990 and an average detection time of 60ms. Compared with the original YOLOv3 algorithm, the accuracy of the improved YOLOv3 algorithm is improved by 6%, and the average detection time is reduced by 8ms.

The study of frequency-selective hybrid precoding for broadband millimeter wave (mmWave) multiple-input multiple-output (MIMO) systems has recently attracted major research interest. In this paper, we seek to provide an efficient solution to simultaneously conduct user scheduling and frequency-selective hybrid precoding for a broadband multiuser mmWave MIMO system. More specifically, we re-formulate the issue using the framework of stochastic-optimization. By utilizing the powerful tool of Lyapunov optimization, a dynamic weighted sum-rate maximization problem is formulated, towards which we develop an algorithm that simultaneously conducts i) user scheduling, ii) multi-carrier resource allocation, and iii) frequency-selective precoder design. It is also demonstrated that the proposed framework is capable of asymptotically optimizing a given stochastic utility (e.g. long-term sum rate). Simulation results are provided to demonstrate the superiority of the proposed algorithm.

Massive multiple-input-multiple-output (MIMO) has been considered as one of the promising technologies for the future communications due to its high spectrum efficiency. In massive MIMO systems, hybrid precoding is widely adopted for it can reduce the unacceptable power consumption with a small number of radio- frequency (RF) chains. However, one RF chain can serve at most one user. Consequently, the number of simultaneously served users are strictly limited. To break this limitation, the idea of beamforming MIMO non-orthogonal multiple access (NOMA) has been proposed. In beamforming MIMO NOMA systems, one RF chain can support two or more users, where users with the same beam selection are paired into one cluster and served by an identical RF chain. However, the traditional beams are so narrow that only a few users can find a match while most of the users still remain single and perform orthogonal multiple access. To solve the problem, we propose a new scheme called multi-resolution beamforming and user clustering scheme. We first utilize beams with narrow angle range and higher array gain for user clustering. Those users which cannot be paired with these beams are clustered by a new beam codebook whose beams have a wider angle range. The selected wide beams are well-designed to avoid bringing interference to the previously paired users. The proposed scheme raises the possibility for users being paired into clusters while keeping a satisfying array gain. Therefore, more users in the cell can take advantages of NOMA to achieve a higher spectral efficiency. The simulation results demonstrate that the proposed scheme outperforms existing schemes under different scenarios.