Abstract：A fully-integrated dual-mode low-noise RF receiver front-end for the 2.3 GHz～2.7 GHz band was presented based on a GaAs pHEMT technology. The receiver front-end consists of a single-pole double-throw(SPDT) transmit-receive switch and a low-noise amplifier(LNA)with bypass mode. The cascode common-source structure with inductive degeneration topology is adopted to achieve the LNA.The SPDT switch is embedded into the input matching network as a concurrent design to realize high Q matching network with minimum components for low noise, high gain and superior input return loss matching. Concurrent design of multiple switches for Bypass function is utilized to attenuate the high input power transmissions. The measurement results show that the presented FEM exhibits low noise figure of 1.53～1.64 dB and gain of 18.1～19.2 dB for 2.3～2.7 GHz frequency range in the high-gain LNA mode. At 2.5 GHz, the tested input 1 dB compression point at 2.5 GHz is -1.5 dBm. The insertion loss in Bypass mode typically varies from 6 to 7 dB in the operation frequency bands.

Abstract：In this paper, a Ka-band frequency source for miniaturization of rendezvous parameter detection unit is designed by using PLL and multiple frequency multiplication. The output frequency of the frequency source is 36 GHz and the output power is higher than 15 dBm. In this design, the mathematical modeling method is used to model each part of the PLL, and then the transfer function of the system is obtained. Through the analysis of the system function of the closed-loop system, the system parameters that can make the system work stably are obtained. Finally, dies are fabricated on the substrate and cavity of RO4350 and RO5880 in the form of micro assembly, and then be tested. The test results show that the stable working frequency source can be obtained by using mathematical modeling.

Abstract：Radiation interference is one of the bottlenecks that restrict the high frequency and miniaturization of power products. Based on the simulation idea of field circuit coupling, the electromagnetic field finite element model of MOSFET and the equivalent high frequency circuit model of high frequency transformer are established. Combined with the PCB network parameters extracted from SIwave,the board level radiation interference of a 5 W output flyback converter is co-simulated, and the influence of two high frequency transformer models on the far-field simulation results is compared. The experimental results show that the simulation results of 3 m far-field within 230 MHz are in good agreement with the measured results, which verifies the correctness of the simulation method, and the simplified two capacitance model of transformer has wider applicability; the obtained near-field electromagnetic field distribution shows that MOSFET is the main source of electric field radiation, and the rectifier diode at the secondary side of transformer is the main source of magnetic field radiation.

Abstract：Silicon-based radio frequency field effect transistors have the advantages of excellent linearity, simple drive circuit, fast switching speed, excellent thermal stability, no secondary breakdown, etc., and have broad application prospects in HF, VHF and UHF bands. In view of the application requirements of the RF field effect transistors with broadband, high gain and high efficiency, based on the standard planar MOS process, the split gate structure was adopted and a silicon-based RF vertical double-diffused metal oxide semiconductor field effect transistor with working voltage of 28 V was developed by optimizing the structure and process parameters. In the frequency range from 30 MHz to 90 MHz, the device can achieve small signal gain greater than 19 dB, the continuous wave output power can reach 87 W at the frequency of 60 MHz, the power added efficiency up to 72.4%, has excellent radio frequency performance.

Abstract：AI(Artificial Intelligence) makes significant improvements with fast development of hardware improvement and algorithm iteration. Google, Baidu and other high-tech company increase investment in AI applications year by year, considering AI as a key areas. At present, artificial intelligence has become the most important technological means to lead the revolution of the information industry. However, the threat of artificial intelligence to humans is also increasing. In this paper, we analize the existing AI algorithm, explain their defects, and discuss what kinds of risks would be caused. Finally, we propose some approaches to generate more explainable, more robust, and more ethical AI system.

Abstract：In view of the large workload of manual image reading, poor image reading quality, and prone to missed inspections and wrong judgments，in this paper, the faster RCNN target detection model is applied to the detection of hepatic echinococcosis CT images. And the target detection model is improved: based on the characteristics of low image resolution and different lesion sizes, the residual network with deeper network depth(ResNet101) is used to replace the original VGG16 to extract richer image features; according to the coordinate information of the lesion obtained by the object detection model, the LGDF model is introduced to further segment the lesion to assist doctors in diagnosing the disease more efficiently. The experimental results show that the object detection model based on the ResNet101 feature extraction network can effectively extract the features of the target, and the detection accuracy is 2.1% higher than the original detection model, and it has better detection accuracy. At the same time, the coordinate information of the lesion is introduced into the LGDF model. Compared with the original LGDF model, the segmentation of hepatic hydatid lesions is better completed, the Dice coefficient is increased by 5%, and the segmentation effect is better especially for the multi cystic liver hydatidosis CT image.

Abstract：Feature engineering can automatically process and generate those highly discriminative features without human operation. Feature engineering is an inevitable and crucial part of machine learning. The article proposes a method based on reinforcement learning(RL), taking feature engineering as a Markov decision process(MDP), and proposes an approximate method based on the upper limit confidence interval algorithm(UCT) to solve the feature engineering of binary numerical data problem to automatically obtain the best transformation strategy. The effectiveness of the proposed method is verified on five public data sets. The FScore of the five public data sets is improved by an average of 9.032%. It is also compared with other papers that use finite element transformation for feature engineering. This method can indeed obtain highly discriminative features, improve the learning ability of the model, and obtain higher accuracy.

Abstract：A hysteresis comparator with wide input range for Controller Area Network(CAN) transceiver is designed to convert the -12~12 V signals on CAN bus into 0~5 V serial digital signals. The solution is to use resistance to linearly superimpose the signals on the power supply and bus to get the signals in the 5 V voltage, then use full differential amplifier to clamp, use common mode feedback to stabilize the common mode working point, and finally use the internal positive feedback hysteresis comparator that can compensate the temperature of the tail current to get the digital signal. Based on VIS 0.4 μm BCD process, Hspice is used for simulation and verification. The simulation results show that the circuit can get the correct digital signal by comparing the voltage of -12~12 V. The hysteresis threshold was 108.9 mV at room temperature(25 ℃) and TT process corner. When the temperature changes from -40 ℃ to 125 ℃，the change of hysteresis threshold voltage is smaller than 9.16 mV，the temperature coefficient is 0.055 5 mV/℃.

Abstract：This paper analyzes and designs a silicon based substrate integrated waveguide(SIW) dual passband broadband bandpass filter chip based on Micro-Electro-Mechanical System(MEMS) process with double layer resonant cavity filter structure. The designed filter chip has an insertion loss of less than 2 dB in the 23～29 GHz and 43～47 GHz dual passband bands with relative bandwidths of 23% and 9%, respectively, and an out-of-band suppression of more than 20 dB, and generates a zero point between the dual passbands with a size of only 4.2 mm×1.1 mm×0.8 mm. The filter is a planar stacked structure, with the advantages of small size, wide band and easy integration, which has good application value.

Abstract：The time-interleaved SAR ADC is sensitive to mismatch between channels. The key factor is the capacitor mismatch, both in channel and between channels. This paper analyzes the impacts of capacitor mismatches in DAC of SAR(Successive-Approximation-Register) ADC on time-interleaved SAR ADC. For improving performance of time-interleaved SAR ADC, a calibration technique based on calibration of capacitor mismatches in single low-speed SAR ADC is proposed, which results in improvement of SFDR(Spurious Free Dynamic Range) exceed 9 dB and SNDR(Signal to noise distortion ratio) exceed 2.5 dB。

Abstract：In order to meet the high efficiency and high power requirement of current power supply module, this paper designs a high frequency drive circuit for gallium nitride device, and builds a high frequency and high efficiency power supply module.By adjusting dead zone minimization processing module and non-overlapping module, this circuit can suppress the high-current pass-through phenomenon of power devices, which is helpful to improve the efficiency of power module.Using the high frequency characteristic of gallium nitride device, the working frequency of power supply system is increased greatly.The test results of the power system show that the time of the output waveform rising edge and falling edge is 10 ns and 5 ns respectively at 1 MHz. At 10 MHz, the time of output waveform rising edge and falling edge is 14 ns and 8 ns respectively. The system can achieve about 10 W high power output. The efficiency of the system at 1 MHz and 10 MHz is 93.7% and 83.5% respectively.

Abstract：GMSK(Gaussian filtered minimum shift keying) modulation technology has high spectrum modulation efficiency, which is suitable for aerospace high bit rate telemetry. Gaussian low pass filter is the key unit of GMSK modulation technology. The design principle of Gaussian low pass filter was introduced, a Gaussian low-pass filter with arbitrary bit rate and low resource consumption was implemented in Xilinx FPGA, with LUT look-up table method and XOR operation used. Arbitrary rate switching function was achieved without using DSP48E1s resources and less LUT resources through the simulation and implementation results.

Abstract：In view of the fact that the quality of the elevator system does not reach the threshold of the elevator system, but the internal space of the elevator is occupied, resulting in responding to external control and making the door control invalid open only according to the quality index. The elevator control system is optimizedand designed. The system uses STM32 microprocessor as the main control chip, and the optical sensor is designed by using photosensitive resistor to perceive the space occupancy of elevator，and the elevator space occupancy is sensed by using LabVIEW to build the upper computer to monitor the elevator operation data in real time. The test results show that the system can accurately detect the elevator space occupancy rate and control the elevator operation according to the space occupancy rate, which solves the problem of invalid opening of the door control caused by the inability to determine the space occupancy rate of the elevator.

Abstract：This paper proposes a spectrum sensing scheme based on wireless fingerprint database. Firstly, in the target geographic area covered by the cellular cognitive radio network(CCRN), the secondary user equipment(SUE) collects a large number of spectrum observation data, and processes the spectrum observation data based on various machine learning algorithms to obtain the transmission mode of primary user transmit(PUT) on the authorized spectrum. Then, in the different transmission modes of put, the geographic location area is divided, and the grid label is obtained based on the spatial distance calculation method, and the wireless fingerprint database is established. Finally, the secondary user equipment(SUE) with sensing needs obtains its wireless fingerprint according to the time of arrival(TOA) estimation of the base station(BS) reference signal, and then compares it with the wireless fingerprint in the wireless fingerprint database(WFPD) to determine its geographical location, and thus to determine the access tag of authorized frequency band. Simulations verify that the proposed scheme increases the spectrum access opportunity under the premise of minimizing the interference to the primary user.

Abstract：To evaluate the task performance of covert communication, this paper firstly proposes the task performance evaluation index under covert communication, and the performance of covert communication under three different strategies of on-off control, power control and combined power and rate control are compared. Finally, the effects of different control strategies are verified through simulation. The simulation results show that the communication duration of the combined power and rate control strategy is 2.24 times that of the on-off control strategy and 1.57 times that of the power control strategy, which greatly improves the communication time of covert communication and the task performance is the best.

Abstract：At present, there are many problems in the traceability of digital energy meters. In view of this situation, this article describes a design method of an analog-digital hybrid standard energy meter, which makes the traceability of digital energy meters unified into the traditional analog traceability system. The article introduces the method of power angle self-compensation to improve the measurement accuracy of the power angle; through a series of technical measures, it overcomes the difficulty of the low sampling rate of 61850-9-2 digital message, making the standard table accuracy level reach 0.01，which is of great significance for the verification work of the 0.05 class digital energy meter calibrator commonly used by power companies.

Abstract：With the wide application of electric bus in urban traffic, the charging work is increasing gradually. As the charging power is large, the charging cable is relatively heavy, and requires a large plug-in force, the labor intensity of manual work is great. It has a wide range of requirements to use general mechanical arm instead of manual to realize automatic plug-in of charging gun. Identifying the charging socket is the premise to realize automatic plug-in. Taking charging socket as identification object, a method of sub-pixel precision contour geometric matching detection based on Canny operator is proposed. Through the application of image processing and machine vision technology in Halcon software, the feature extraction and template matching of the target area are carried out to realize the automatic recognition of charging socket. The experimental results show that the method has good robustness in the case of external light environment.

Abstract：In recent years, big data, natural language processing and other technologies have been developed rapidly. As one of the cutting-edge technologies in the field of natural language processing, emotion analysis has received great attention. However, High precision and high performance are still the key factors restricting emotional analysis. In order to achieve high-precision emotion analysis, based on the feature-level neural network, this paper improves the reset feature level attention mechanism, and proposes an aspect level emotion classification model based on the reset feature level attention(RFWA) and an aspect level emotion classification model based on the reset feature level self-attention(RFWSA). Finally, combined with Bi-LSTM-CRF,high quality aspect level emotion analysis is realized by aspect level phrase extraction in the network. The experimental results show that compared with the existing mainstream emotion analysis model, the model proposed in this paper has obvious advantages. Especially when the classification effect is quite good, the parameters of the model are only 1/4 of the AOA Network.

Abstract：Extracting effective characteristics of load operation data plays an important role in improving the accuracy of non-intrusive load identification.In this paper, a ReliefF-DDC feature selection algorithm was proposed to reduce feature dimension, reduce complexity and improve load recognition.Firstly, ReliefF algorithm was used to analyze the relationship between each feature and category, calculate feature weight, and screen irrelevant features.Secondly, DDC algorithm is used to calculate the mutual information analysis correlation between features and categories, and redundant features are removed according to feature subset evaluation measurement. Finally, twin support vector machine(TWSVM) is used as classifier for load recognition. Experiments show that the algorithm proposed in this paper improves the classification effect and reduces the running time.

Abstract：Broadening the structure of the neural network will lead to the increase of the amount of computation and the decrease of the computational performance, it is necessary to optimize and schedule the parallel network more effectively. By analyzing the throughput and bandwidth of convolutional neural network on FPGA platform, the roof model is used to explore the design space under the limitation of computing resources and memory access bandwidth. It is proposed to use different cycle expansion factors in different branches of convolutional neural network, so as to realize the parallel computing of different branches of neural network in the same convolution layer and ensure the computing efficiency reasonable allocation of resources and memory resources. The experimental results show that the performance of the proposed design is improved by 1.31×compared with the previous research.

Abstract：In order to solve the shortcoming of narrow scanning angle of micro-electro-mechanical system(MEMS) micromirror,a synchronous scanning system of MEMS micromirror and brushless motor was designed.The system consists of STM32 microcontroller,direct digital synthesis(DDS) drive circuit,and phase detection circuit.The speed measurement of the brushless motor is realized by designing the grating code disc.The proportion integration differentiation(PID) algorithm controls the motor speed so that the motor feedback signal and the MEMS micromirror feedback signal have the same frequency and the phase difference of the two feedback signals can be detected. Adjust the phase of DDS drive signal to achieve phase synchronization.Experiments show that the system works stably and the phase error does not exceed 1.5%.

Abstract：The high-precision synchronous collection of multi-source heterogeneous data is a key technology in the rapid detection system of combined track motion and static. In this paper, a set of track dynamic and static combined rapid detection multi-source data acquisition module is designed. This module uses Xilinx′s Artix-7 series FPGA as the main control chip, NVIDIA TX2 high-performance computer as the core data processing unit, and uses high-stability crystal oscillators and PPS pulse and NEMA data output by GPS establish a high-precision time reference, and realize high-precision real-time synchronous collection of multi-sensor data such as displacement sensors, temperature sensors, and laser displacement sensors by controlling AD sampling frequency. The module is small in size and low in cost. Tests show that the module can achieve high-precision synchronous collection of multi-sensor data, with a synchronization accuracy of microseconds.

Abstract：An oscilloscope channel protector is developed based on the THS3491 operational amplifier, which is used in the diagnosis of pulse radiation field. By comparing and analyzing the bode plots of THS3491 in SOIC and VQFN packages, the THS3491 in VQFN package is used as the core chip of the protector. The output voltage range of the developed protector is ±10 V(50 Ω load), the peak-to-peak voltage noise is about 2 mV; the -3 dB small signal bandwidth is about 850 MHz, and the 0.3 dB flatness bandwidth is about 710 MHz. The paper discovered and theoretically explained the peak shift phenomenon of the protector output voltage. The protector was tested by a delta signal source to achieve a distortion-free delta response. When the output voltage amplitude range exceeds ±10 V, the protector enters the output protection state, and the recovery time is less than 7 ns, which can well meet the oscilloscope channel protection requirements in the pulse radiation field diagnosis.

Abstract：Based on the analysis of the shortage of power consumption of the increasingly miniaturized and integrated space image equipment, this paper designs an interface scheme based on the domestic video processing chip Hi3516D. So that the equipment can use this interface to realize the function of low-power image acquisition and transmission. The selection of low-power devices, power consumption hardware adaptation, dynamic voltage regulation hardware circuit and data transceiver flow are analyzed in detail; the calculation of dynamic voltage regulation hardware circuit parameters, the software design flow of low-power module and the information configuration of uboot register and kernel module are given; the image quality and power consumption under different resolutions are tested. After design analysis and test verification, when the interface works within 3 Mb/s bit rate, the power consumption of the whole machine is less than 2.5 W, and the power consumption is reduced by 8%～16%, the image quality PSNR is better than 36 dB, and the interface performance meets the design requirements.

Abstract：This paper summaries InGaAs ternary compound material which can be used to detect near-infrared light，and solves the wave function equation of the InGaAs/GaAs single quantum well by using mathematics software. The InGaAs/GaAs multiple quantum wells structure is designed and fabricated by MBE on the basis of the calculated conclude. The In component and the thickness of the well and barrier can be calculated by double crystal X-ray diffraction analysis. The results are consistent with theoretical design and could guide the infrared detector development.

Abstract：Fingerprinting location based on received signal strength indication(RSSI) of Bluetooth low energy(BLE) is one of the most popular and effective indoor positioning methods. However, the BLE communication protocol is not specifically designed for location, and the ranging performance of the Bluetooth devices under different communication connection modes varies greatly, which significantly affects the positioning accuracy. In response to this problem, the paper tests and analyzes the BLE RSSI ranging performance under different connection modes. Further, experiments based on a self-built indoor positioning environment are carried out to verify its impact on the user′s actual positioning accuracy. The test results show that when the BLE module is set to "un-connectable mode", the RSSI ranging stability of the device is improved by about 60%, and the user′s online positioning accuracy is improved from 1.40 m to 0.58 m, which significantly improve the positioning performance.