Abstract：In this paper, a low-cost, tiled millimeter wave active phased array antenna(APAA) is presented. The APAA integrates CMOS multi-function chips, GaAs chips, antenna array and power division network on a microwave multilayer print circuit board by a three-dimensional integration technology, then the height of the APAA has sharply decrease. The RF part of the APAA is interconnected with the frequency conversion module, wave control module and power supply module through the vertical interconnection of radio frequency and low frequency pins. The measured results show that the APPA can achieve 7.7% bandwidth in a scanning range of ± 25° in all planes in Ka-band, and the gain changes smoothly in the scanning range, less than 1 dB gain variation, and the sidelobe level is better than 14.5 dB. The APPA has low profile of 52 mm, ERIP is larger than 66 dBm, G/T is better than 7.5 dB/K, and the transceiver switch time is less than 150 ns. In summary, this tiled active APAA has characteristics of high integration, miniaturization and low cost，and has high engineering application value.

Abstract：In order to meet the needs of microwave devices for high-resolution, non-destructive microwave vector near-field measurement, an all-optical microwave near-field vector measurement technique based on diamond nitrogen-vacancy(NV) centers is proposed. Using the sensitivity of NV centers to the circularly polarized electromagnetic field, the tapered optical fiber probe with diamond NV center is used as the sensor to measure the optical detection magnetic resonance(ODMR) spectrum with 8 peaks in the external static magnetic field environment, and the distribution of electromagnetic field components in different NV directions on the surface of microwave device is measured under the corresponding microwave frequency of each ODMR spectrum peak. The results of microwave near-field vector measurement are obtained. Finally, 3 μm diamond particles are used for near-field vector imaging of microstrip antenna with resonance frequency of 2.87 GHz. The experimental results show the effectiveness of the method, which can be widely used in chip electromagnetic compatibility(EMC) test, integrated microwave chip failure analysis, digital circuit signal integrity analysis, etc.

Abstract：In the face of the development of miniaturization and compactness of chipless RFID tags, and in order to reduce the cost of traditional RFID tags, this paper proposes a compact, polarization insensitive, passive chip less RFID tag based on 1/4 ring resonant unit. It has a high coding density of 4.39 bit/cm2 in the frequency range of 6 GHz～11.5 GHz. The pole of the tag is extracted by Matrix Pencil Method(MPM). Compared with the simulation data and the algorithm extraction data, the designed tag meets the requirements of identification accuracy and has good data coding performance.

Abstract：This paper introduces a design of RF Switchbox for RFID-OTA test system which can be programmed switching. According to the requirements of RFID and OTA test system for multi-band testing and path calibration，through reasonable structure design and system layout design, the programmable RF-Path switched of RF switchbox in the RFID-OTA test system is realized, which enables the test system to automatically complete the performance test and test path calibration of products in different frequency bands.

Abstract：In This paper, the development trend of automatic driving business, the limitations of single vehicle intelligence, and the advantages of vehicle road coordination are introduced. At the same time, it focuses on the technical architecture and system requirements of vehicle road collaboration, the multi-level multi-source data fusion perception analysis architecture based on "end edge cloud", and the implementation path of three-level/four-level edge deployment based on roadside+edge cloud + regional/central cloud. The networking scheme of multi-level multi-source sensing fusion analysis capability based on edge cloud is then proposed, and its specific functions and delay requirements are discussed in detail. This research is of some reference value to the project planner and designer of vehicle road collaboration.

Abstract：The development of autonomous driving, vehicle-road collaboration and Internet technology has given birth to the new power of Internet car building and promoted the building of a grand ecosystem of networked cars. This paper focuses on the intelligent automobile manufacturing industry, focuses on the digital upgrading of the industry and the intelligent transformation of the whole process, and relies on 5G + industrial Internet and artificial intelligence technology to carry out the practice of "digital transformation", focusing on the digital transformation of automobile business process and other aspects to improve the level of data empowerment. Through 5G networking upgrading, with equipment-end, factory-end, enterprise-end and inter-enterprise networking 5G upgrading as the focus, industrial chain connectivity will be enhanced. Through artificial intelligence technology and edge computing technology, focusing on the intelligence of production tools and decision-making means, the efficiency of enterprise resource allocation is improved to realize the flexible manufacturing of intelligent vehicles.

Abstract：In the field of vehicle networking, vehicle-road collaboration is an inevitable technical approach to realize intelligent transportation. With the development of Internet of vehicles(VV) business, rW equipment manufacturers and rW equipment types are increasing day by day, which brings new challenges to the traditional operation and maintenance. Through industry research, this paper studies the current requirements of RCC RCC equipment on operation and maintenance management, analyzes the problems faced by current RCC equipment operation and maintenance management, identifies key technical fields that need to be solved urgently, and provides theoretical basis and technical support for the realization of intelligent transportation operation and maintenance.

Abstract：This paper introduces the status of high-precision map production，and puts forward some problems and demands. It is very important for automatic driving. This paper focuses on a diversified map collection and update solution and related technologies based on 5G edge cloud collaboration. It would provid some reference for the construction and operation of high-precision map system in the future.

Abstract：With 5G commercialization, automatic driving based on intelligent vehicle infrastructure cooperative systems(IVICS) is considered as the one of the main direction of future development of IOV. The Road Side Unit is the key of IVICS. Based on the future development of 5G and C-V2X, eNB-type RSU is analyzed in this article, the technical scheme and function of 5G eNB-type RSU are discussed, and the application scenarios of 5G eNB-type RSU is also forecasted.

Abstract：Ambient backscatter technology brings new solution for those devices largely deployed with low power consumption required by Internet of Things. This paper firstly introduces an overview for basic theory and categories of backscatter technology. Then based on Wi-Fi signal transmission with ambient backscatter, the paper gives the analysis how backscatter device handles the messages encoded from Wi-Fi packets and reflects the signal for information transmitting. Then the paper provides the proposal of relevant hardware framework and software architecture, and potential deployment user cases. From perspective of product delivery, the paper also summarizes the technical challenges and key improvement areas. Finally, the paper shares the view for technical evolution trend and key points for next study.

Abstract：In this paper, a micro-expression and vehicle status recognition method based on blocked local binary pattern from three orthogonal planes(LBP-TOP) features and weighted sparse representation as the classifier is proposed. First of all, the effective block is selected from the blocked image. Then, the features, which are extracted from LBP-TOP feature descriptor, are used as a dictionary. Then the combined weighted sparse representation(WSRC) and the dual augmented lagrangian multiplier(DALM) algorithm performs sparse representation classification. Finally, the images are divided to different sizes blocks, then the effective block is chosen from these blocks, and the features are merged as the input to the classifier. The experiments are carried out on the CASME Ⅱ，SAMM and vehicle databases using leave-one-subject-out cross validation(LOSOCV). When classifying the micro-expressions into five categories, the classification accuracy can reach separately 77.30% and 58.82%, and the experiment on the database of vehicle state detection reaches 84.60% detection rate. Experimental results show the effectiveness of the proposed algorithm.

Abstract：Image quality degradation caused by rain streaks seriously affects the effective application of image and computer vision algorithm, so image deraining is very necessary. At present, mainstream deraining methods based on deep learning are only effective for single size rain streaks, and there are problems such as incomplete rain streaks removal and fuzzy background. Aiming at these difficulties, a single image deraining algorithm based on deep controlled dense connection network is presented. Through the introduction of multi-scale block, the ability to extract rain streaks of different sizes was enhanced. And attention mechanism module was injected to pay more attention to raining areas. What is more, controlled dense connection block was also introduced to fully represent the rain streaks characteristics. Experiments show that the proposed method outperforms some mainstream methods both on the synthetic dataset and the real dataset.

Abstract：Based on the SMIC 0.13 μm process，an analog integer frequency PLL with 1.5 V working voltage is designed, and a new type of current steering charge pump that can effectively solve the charge shunt and improve the switching speed is proposed. At the same time, a voltage controlled oscillator with low phase noise is designed based on the ring differential inverter. On the basis of the integer frequency-division PLL, the digital DSM with MASH3 structure is added to realize a digital-analog mixed fast-locked decimal frequency-division phase-locked loop, which can be the output frequency range from 0.8 GHz to 2.0 GHz at 26 MHz， and has the advantages of low noise, low frequency stepping, and fast lock.

Abstract：The realization of coherent population trapping(CPT) atomic clock with miniaturized, low-power is of great significance for its applications in the fields of precision measurement such as communication and position. The peripheral circuit controlling accurate frequency and laser output is the key technology to achieve coherent population trapping atomic clock. A peripheral control system of CPT Cesium atomic clock is developed in this paper. The system uses two levels for fine-tuning and coarse-tuning to drive the VCSEL laser to work, which is more accurate than single current adjustment and uses phase-locked loop circuit to generate modulated microwave signal. PIN diodes are used to adjust the microwave power attenuation in order to realize the adjustment of microwave power. In the meanwhile, software modulation and demodulation methods is applied to replace traditional hardware modulation and demodulation circuits, which saves circuit logic resources.The testing results show that the output of the microwave power reaches -3.33 dBm when the microwave frequency is 4.596 267 GHz. The phase noise of the radio frequency circuit is better than -80 dBc/Hz under the frequency offset of 1 kHz which has better capability.

Abstract：Since the development of modern technology, the number of high buildings and complex structures keeps ascending, which leads to the vital hazard of firefighters. This paper proposes a complete aid system to duel with the issues on vital signal detection and indoor location for the firefighters. In this paper, LSTM Neural Network is employed to classified the attitude and movement of the firefighter, and particular photoelectric sensors are used to detect the heard rate，blood oxygen saturation and the gas surrounded. At the meantime, this paper proposes a indoor location system based on the data fusion of the Ultra-WideBand location and the inertial navigation. Finally data above are transmitted via LoRa-170M Network to the monitoring platform which is built up with the help of LabVIEW. At the end, experiments are designed to verify the ability and stability of the system.

Abstract：In this paper, according to the research of four-rotor UAV control system, LabVIEW software is designed to build the data acquisition panel and query playback panel of the control system.The flight attitude parameters of the four-rotor UAV are collected in real time by the control system. The real-time data of flight altitude, flight speed, pitch angle and rolling angle are displayed in the data acquisition subpanel, and the functions of data display, processing, storage and playback are realized. The system has been tested, it realizes the attitude monitoring of UAV and meets the actual needs of controlling small four-rotor UAV.

Abstract：Deep learning has achieved great success in many areas such as computer vision, speech signal processing and natural language processing. However, as the neural architecture becomes more complex, the number of parameters increases rapidly. Designing efficient neural architectures requires expertise knowledge and quantities of repetitive experiments, which leads to high cost. Besides, the scale of neural networks running in mobile devices is strictly limited due to the power consumption limit and relatively small storage space. This paper proposes an efficient architecture searching algorithm for mobile devices. The main contributions include:(1)We propose an algorithm for calculating the number of floating point operations in neural networks under given architecture.(2)We improve an existing gradient based neural architecture search algorithm and propose a constraint NAS algorithm.(3)We search several efficient neural architectures by adding the constraint of the number of floating point operations to the architecture searching process and adjusting its strength. We train the searched neural networks, test their performance on image classification tasks and compare with other neural networks which are commonly used in industry. The experimental results show that the performance of the model searched by our method can reach the performance of mainstream models in industry.

Abstract：During the process of large building disaster, due to the adverse effects of toxic smoke, noise, fire, electricity leakage, light and other factors, as well as the complex internal structure of large buildings, it is difficult for many rescuers to obtain reliable information. Considering the above situation, wireless sensor networks can play their advantages in positioning indoor complex environments. But there is a challenge. Although their positioning accuracy is very high in the LOS environment, their measurement may be polluted by non-line-of-sight propagation in the NLOS environment, which results in a decrease in positioning accuracy. To solve this problem, we propose an improved location method based on unscented Kalman filter(UKF). Firstly, the propagation state between mobile node and beacon node is identified by means of test statistics. Secondly, the linear Kalman filter(LKF) is used to measure the distance smoothly. On this basis, a modified Kalman filter(MKF) is used to weaken the influence of NLOS on the measurement. Then, the UKF method is used to determine the location information of the unknown mobile node. Finally, the effectiveness of the proposed algorithm is verified by numerical simulation.

Abstract：In the field of three-dimensional virtual experiment, human pose recognition is one of the key technologies. Applying this technology to modern medical treatment can provide certain assistance and promotion for medical rehabilitation. In this study, the Kinect human bone tracking technology is combined with medical rehabilitation. Based on the extensive collection of standard medical rehabilitation actions, a set of standard medical rehabilitation action database is constructed as the basis for comparison with the data collected by sensors. In order to improve the stability of joint angle measurement and the accuracy of pose recognition, a method of pose determination based on the Euclidean distance and angle between joint points is proposed. The results show that through the application of this technology, the recognition rate of posture is high, and the feedback can be given in time for the rehabilitation training of the trainer, which makes the movement evaluation and guidance for the trainer more convenient and efficient, and is conducive to the development of medical science and technology, with the promotion value.

Abstract：As the number of PCIe devices used in computer systems increases, timely adjusting the power management state of the device can effectively reduce power consumption. The operating system and drivers need to cooperate with each other to complete the control for the PCIe device′s power management. Based on the research and analysis of the Windows operating system′s power management technology and the WDF, this paper introduces the processing flow of I/O requests and the principle of queue′s usage, as well as designs the solutions based on I/O request processing, power state transitions and framework property settings to achieve device hibernation and wake-up. Moreover, there are some codes shown to prove the feasibility of the design.

Abstract：There are a lot of strong non-linear and non-stationary noises in the wind field data detected by radiosonde. In this paper, an adaptive and multi-scale wavelet denoising method combined with variational modal decomposition is used to extract features and denoise the wind field data of radiosonde at Syowa station in Antarctica in January 2019. Firstly, the signal is decomposed by the Variational Mode Decomposition, then modes related to the signal components are determined by the correlation coefficient method, and then the effective components and noise in the remaining modes are further separated by the wavelet analysis method. By merging and reconstructing all the effective components, the signal feature extraction and noise reduction are realized. Simulation experiments and real signal processing results show that this method has a good suppression effect on non-stationary nonlinear noise in radiosonde wind field signals, and has better feature extraction and noise reduction capabilities than traditional methods.

Abstract：In order to meet the needs of Internet information collection and mining, aiming at the problems of traditional web site information collection methods, such as mixed collection information, unable to be used directly, and the high cost and low efficiency of manual structured collection method, this paper proposes an adaptive method for extracting structured information from web pages. We implement web page classification algorithm, subtree based title item and content item structured information extraction algorithm. Based on the classification annotated dataset of typical website pages, the classification model can adapt to the differences of various web sites, classify the web pages, and extract the list structured information and content structured information in the web pages according to the web page classification. This technology plays an important role in improving the automation level and processing efficiency of website structured information collection and processing.

Abstract：In order to solve the problem of comprehensive effectiveness evaluation of the system with complex composition, diverse functions and poor samples, this paper constructs a system effectiveness evaluation model based on grey theory, RBF neural network and grey RBF neural network according to the three-tier structure of the system effectiveness evaluation index system. The simulation results show that the accuracy of the grey RBF neural network model is higher than that of the grey model and RBF neural network. Through the grey RBF neural network model, we can accurately evaluate the comprehensive effectiveness of the system with various functions, complex composition, few samples.

Abstract：Aiming to real-time requirement of image processing, an image compression coding and decoding system based on FPGA is designed. The system includes real-time image acquisition, JPEG compression and UART transmission. The D5M camera is configured by Verilog hardware description language to acquire image. We focus on improving two-dimension discrete cosine transformation(2D-DCT) design flow in the image compression module. The dual frequency signal controller is adopted to reduce the call of adder based on the Chen algorithm and realize its fast operation. The test shows that the compression ratio of the image is 26.3:1, and the average signal-to-noise ratio is more than 40 dB. The visual effect of the compressed image is good, which meets the design requirements.

Abstract：At present, in the research of heart sound recognition system based on medical big data and machine learning, the extraction of single cardiac cycle mostly relies on manual interception or segmentation based on synchronous ECG signals, which greatly reduces the practicability and ease of use of the whole system. In view of this, a hardware system with the function of single-cardiac periodic segmentation based on low-frequency extraction and MFCC(Mel Frequency Cepstral Coefficients) feature extraction is proposed, which realizes the single-cardiac periodic segmentation of heart sound and gives MFCC features more efficiently, and comprehensive segmentation accuracy reaches 98.3%, solves the problem of high purity of heart sound signal and no mature system in the segmentation of single heart sound cycle, meanwhile reduces the data storage cost，increases the system practicability and has potential application prospect.

Abstract：Millimeter wave(MMW) communication link is an important transmission technology of 5G/6G and wireless private network. Adaptive modem can resist environmental influences such as rain attenuation and effectively guarantee the quality of communication link. In this paper, the system generator, Vivado and other tools were used to realize and verify the adaptive modem module adapted to MMW communication link on FPGA, which supported seven modulation modes of BPSK to 4096QAM, realized the time delay less than 18 clock cycles, and could be automatically switched according to channel measurement and signaling instructions in the condition of ensuring the SNR is below 10-5.

Abstract：In order to meet the needs of laboratory teaching and research, based on phase-locked loop technology，this paper designed a secondary radar signal generator, to achieve the 700 MHz～1 100 MHz range of frequency synthesis. Firstly, the principle of frequency synthesis of phase-locked loop is analyzed, and the third-order passive loop filter is designed and realized. The high-frequency components generated by phase detector are filtered out and the phase-locked loop control circuit is formed with phase-locked loop chip ADF4350. Secondly, using the simulation tool ADI simPLL, the phase-locked loop is simulated and the parameters of the loop filter and the simulation output are obtained. Finally, the secondary radar signal generator system is built and the system output is analyzed by SPSS software. The results show that the system deviation is less than 0.02 MHz and the fixed deviation is in the range of 1 GHz～1.1 GHz band. After the compensation, system accuracy can be meet the needs of laboratory teaching and research well.

Abstract：In order to quickly identify moving objects within a certain range and improve the accuracy of detection, a close-range moving objects detection system based on STM32 and microwave radar is designed. The system uses the Doppler principle of electromagnetic waves to detect differences in the wavelength of reflected waves from moving objects. The received signals are processed through STM32 and optimization algorithms to improve the speed and accuracy of detection. Finally, the system test shows that the system has the ability of low cost, stable working environment, fast and accurate display.

Abstract：Adaptive sidelobe cancellation radar systems can effectively resist active jamming from the sidelobe direction. Based on the analysis of the operating principle of AESA adaptive sidelobe cancellation, on the base of auxiliary antenna configuration design, the cancellation performance of the adaptive sidelobe cancellation radar system was simulated systematically and comprehensively in view of the change of jamming source relative to radar angle and the difference of radar antenna pattern. The simulation results show that when multiple single polarization jamming sources and single polarization jamming source is used for jamming the adaptive sidelobe cancellation radar system respectively, the blinking jamming and polarization jamming are not ideal. Therefore, jamming combination of blinking jamming with multiple jamming sources and polarization jamming is advised as the sidelobe jamming strategy against adaptive sidelobe cancellation radar systems.

Abstract：Multi robots create large-scale map together, the key of which is to splice multiple local grid maps when the relative position of robots is unknown. In this paper, the method of image feature matching and the improved particle swarm optimization algorithm are combined. Firstly, the feature points of the two grid maps to be spliced are extracted for matching, and the effective feature point pairs are screened. Then, the information of the feature point pairs is used as the input parameters of the improved particle swarm optimization algorithm to calculate the best conversion matrix from the source image to the destination image. Finally, the converted source image and destination image are stitched together by using the grid superposition rule. Through simulation experiments and data analysis, the stability and accuracy of the algorithm are proved.