Abstract：Multi-access edge computing(MEC) as one of key technologies of 5G，has attracted wide attention from related researchers. Based on edge networks and edge computing resources, MEC provides a combination of connectivity, computing, capabilities and applications, provides services to users nearby, and supports the introduction and exposure of cooperation between network capabilities and business capabilities. In the 5G mobile communication network, it has become a hot issue that how to exposure the edge capabilities safely and efficiently. Given that, this paper firstly introduces the MEC system architecture, network capabilities and business capabilities, and then studies the capability exposure architecture and procedure. Finally, we give the security strategy and introduce a specific example.

Abstract：With the advent of 5G, the new application form of edge computing and Internet of Things has become the focus of attention of the whole industry. The video security industry has become a key area of application because its core business trends and needs match the technical capabilities such as 5G, edge computing, Internet of Things, cloud computing, big data, and artificial intelligence deeply. This paper mainly introduces a cloud-side collaborative PaaS and corresponding technology prototype system based on 5G MEC infrastructure, as well as the application practice of the cloud-side collaborative PaaS in the field of intelligent security.

Abstract：Multi-access edge computing(MEC) enables applications, services and content to be deployed locally by migrating computing storage and business service capabilities to the edge of network. MEC can solve the problem that traditional content delivery network(CDN) nodes cannot sink into mobile network. However, after the CDN node sinks to the edge of the network, the CDN needs the network and terminal to assist the service to identify the user′s location and realize the traffic offload. This paper proposes an edge CDN service scheduling scheme based on MEC. Test results show that the proposed scheme can greatly reduce the first packet delay and increase the download rate, and significantly improve the user′s business experience.

Abstract：The advance of 5G technology can help to accelerate the development of remote applications in medical industry. Some telemedicine technologies are gradually improved, such as teleconsultation, remote diagnosis, remote ambulance, remote surgery and so on. As one of the main scenarios of 5G applications, 5G video services are required to achieve rapid transmission of high-resolution and high-frame-rate videos, as a result to optimize diagnostic results and improve medical efficiency. Virtual reality(VR) is a significant part of 5G video services, which is able to work with medical treatment to achieve functions like VR live broadcast in hospital, VR visit and so on. However, VR requires high capability of network, and the bandwidth and delay achieved by traditional 4G network cannot meet its requirements in medical application. This paper takes examples of VR visit and proposes a network solution to 5G medical applications based on SA smart medical private networks. Through Flex-E network slicing, edge computing and cloud application of VR, the medical resources are isolated from public resources to ensure the efficiency of VR data transmission, and a slicing scheme specific for VR scenarios is developed, which provides a fast and stable network solution for data transmission of VR applications in medical industry.

Abstract：Internal carrier control methods for Insulated Gate Bipolar Transistors(IGBTs) are critical for the device performances such as on-state voltage drop, turn-off losses, SOA, thermal reliability and transient ruggedness, etc. Numerous carrier control methods focusing on the design of the emitter(or cathode), collector(or anode), and drift regions have been reported. This paper focuses on the carrier control methods of current and future generations of IGBTs.In particular, the designs of the emitter, collector, and drift regions and how they affects the trade-off between the forward voltage drop and the turn-off energy loss are reviewed. Finally, the development trend of carrier control in high-power IGBT devices is summarized look forward to.

Abstract：Spoofing face can be used to deceive face authentication system for illegal purposes, and thus it poses a serious threat to the face recognition system. Most of the existing methods are training and testing in the same dataset, and the effect is not ideal when they are used in cross dataset scenario. In order to solve this problem, this paper proposes to use histogram of oriented gradients(HOG) to extract the cue information in the context and then send the extracted features to the one-class support vector machine(OCSVM) for training and classification. The classification results are combined with the abnormal cues detected in the context. And the algorithm is verified on the public database NUAA and CASIA-FASD. The experimental results show that the generalization ability and detection accuracy of the proposed algorithm has improved over the existing method when used in cross dataset scenario.

Abstract：In order to improve the accuracy of text similarity detection algorithm, this paper proposes a text similarity detection method combining latent Dirichlet Allocation(LDA) and Doc2Vec model, and names the model obtained by the algorithm HybridDL model. This algorithm obtains the document vector through Doc2Vec training of the document, and then obtains the probability of the occurrence of the document topic and the feature words under each topic with the LDA model, calculates the probability weighted sum of each topic and feature words in the document, and maps them to the Doc2Vec document vector. Experimental results show that the new algorithm model is more sensitive to the judgment of similar text than the traditional Doc2Vec model, and has higher accuracy in the detection of text similarity.

Abstract：Using TSMC 180 nm CMOS technology, a three-stage operational amplifier with high efficiency and large capacitive load driving capability is designed and implemented. This paper proposes an ACBC-C compensation structure based on CMC(Cascode Miller Compensation) and ACBC(AC Boosting Compensation). Among them, ACBC by adding an AC path increases the GBW and the capacitive driving ability. The output stage uses a class AB structure to achieve high efficiency. And CMC can adapt to the class AB output stage structure, maximize efficiency while keeping great linearity, and remove the zero caused by Miller capacitor, and better compensate the stability. The simulation results show that when the phase margin is set to 45 degrees, the maximum load capacity can be as high as 3 100 pF, and that when a 1 000 Ω load resistor is connected to the output, the circuit efficiency achieves 95.299%.

Abstract：Based on the spin transfer torque-magnetic random access memory of 1T1MTJ, an improved bit logic operation scheme in the memory is proposed in this paper. This scheme improves the storage array density by simplifying the 2T2MTJ bit logic operation scheme in the memory and increases the operation functions of "NAND " and "NOR" through the complementary readout circuit. In addition, an improved high-speed sense amplifier suitable for the above scheme is proposed by adding a branch voltage stabilizing circuit. Simulation results based on 55 nm LL logic process of SMIC show that, compared with traditional sense amplifiers, this scheme not only improves the reading speed by 33%, but also has stronger reading ability and better PDP when it is suitable for large storage arrays(CB≥0.8 pF).

Abstract：Based on the 0.7 μm InP HBT process of Nanjing Electronic Device Research Institute, a 1:4 quantized speed-down chip with a data conversion rate of 50 Gb/s is designed. At the same time, the chip integrates a front-end high-speed and high-sensitivity comparator with a 1:4 demultiplexer, so it is possible to directly achieve reliable reception and speed-down processing of analog input signals with a bandwidth of 2～18 GHz. The sensitivity of the input signal reaches 1 mV at the maximum operating rate of the chip, the operating voltage is 3.3 V, and the chip consumes 1.5 W. The highest data conversion rate reaches 50 Gb/s, and the amplitude of the output data signal and the clock signal reach 200 mV.

Abstract：With the development of technology and the decrease of the process size, the scale of the single memory cell becomes smaller, causing the charge sharing and the SEU more serious, and increasing the difficulty of radiation harden design. Therefore, the accurate simulation for estimating the SEU of module level SRAM is needed to provide basis and guidance for radiation harden design. This paper summarizes the structure and the layout of SRAM, proposes a simulation of the SEU of module level SRAM using Cogenda for sensitivity analysis and the SEU LET threshold estimate. Comparison between the simulation and the heavy ion tests shows that the difference is about 13.3%.

Abstract：Based on the principle of time difference method of ultrasonic flow measurement, the measurement errors caused by the convexity and convexity installation and installation position deviation of transducer are analyzed, the system model of flow measurement is established, and the flow calculation method based on BP neural network is put forward to compensate the measurement errors.Taking Senscomp Tao2008 ultrasonic probe as an example, the flow is measured in the software and hardware of the system designed based on the neural network algorithm.The experimental results show that the neural network has stronger nonlinear compensation ability than the traditional weight system.

Abstract：In order to solve the cumbersome problem of manually measuring the body size of dairy cows during dairy farming, reduce the manpower, workload and stress response of dairy cows, a measurement method based on machine binocular vision and stress-free cow body size data is proposed. The method mainly includes the calibration of the camera, the acquisition of the image, the preprocessing of the image, the extraction of the external contour of the cow, the matching of the characteristic points of the cow by the SIFT algorithm, and the body size of the cow can be calculated by calculating the three-dimensional coordinates of the matching point. Through the picture collection of 20 cows, the average error of the body size of the cow is less than 1.21%. Machine vision technology is applied in the field of dairy cattle breeding, which accelerates the fine feeding of dairy cows.

Abstract：OBD vehicle terminal can realize the capture of vehicle operation information, and then realize the status monitoring of vehicle, human and environment. Aiming at the present situation of single function of OBD products, a set of multifunctional OBD vehicle terminal equipment is designed and developed. Based on Infineon′s XMC4500 series 32-bit high performance processor, the hardware platform is equipped with 4G communication module and WiFi module to realize vehicle data upload and wireless network environment coverage. It is equipped with the MPU-9250 gyroscope to realize the detection of the 9-axis motion of the vehicle and support the read and write operation of SDHC large-capacity data. In order to enhance the user′s experience, digital FM audio module and tire pressure monitoring module are added to improve the user′s driving experience. This design can realize the collection and upload of vehicle engine parameters, calculate the data through the algorithm of upper computer platform, and provide reference for driving behavior and UBI insurance business.

Abstract：To accurately measure the temperature and humidity in the elevator car, a simple fiber Bragg grating(FBG) temperature and humidity sensor was developed. The sensor was composed of FBG with a grating length of 20 mm, where in the region with a grating length of 12 mm was etched to 7.8 μm by wet etching, and a polyimide moisture sensitive material was coated on the etched surface to measure relative humidity. The unetched 8 mm FBG was used to detect temperature. In the experiments, firstly, the response characteristics of the etched and unetched FBGs to the temperature and humidity were studied. Secondly, the influence of the thickness of the polyimide film on the humidity response of the etched FBG was examined. Thirdly, the response of the etched FBG with polyimide coating to temperature was checked. Fourthly, we investigated the response of the sensor that includes etched and unetched FBGs in variable temperature and humidity environment. The experimental results show that when the thickness of the humidity sensitive film was 80 μm, the response sensitivity of the etched FBG to humidity was 17.1 pm/(%RH). The sensitivity of the unetched FBG to temperature was 9.9 pm/℃. The sensor can accurately measure temperature and humidity in the temperature range of 30 to 70 ℃ and in the humidity range of 20~90%RH.

Abstract：In order to improve the effect of medical image segmentation, for the effect of two-dimensional Tsallis entropy threshold method，two-dimensional Tsallis Entropy improved by cloudmodel firefly algorithm is applied to medical image segmentation algorithm. Firstly, in order to improve the convergence speed and optimization ability，the cloud model is introduced into the Firefly algorithm. Secondly, the homogeneity measure was chosen as the evaluation index of medical image segmentation, and the parameter q of the two-dimensional Tsallis entropy threshold method was optimized by CMFA algorithm. The results show that CMFA-Tsallis has the highest homogeneity measure compared with FA-Tsallis and Tsallis, and the result boundary is clear, thus proving the effectiveness of this algorithm.

Abstract：To realize the end positioning of the manipulator with the hand-eye relationship uncalibrated, an image-based uncalibrated visual servoing system is designed and implemented for the monocular hand-eye system. The paper performs modular system design, uses the Kalman filter to estimate the joint-image Jacobian matrix online, designs the visual servoing controller according to the joint-image velocity mathematical model, and develops each parallel algorithm module by C++ multi-threading technology, which realizes the complete high-precision positioning of the arm end without calibration. Experiments show that the positioning accuracy of the system is 0.1 pixels.

Abstract：Aiming at the defect that the traditional BP neural network model cannot effectively express the historical features existing in time series data, a method with the combination of BP neural network and grey forecast principle was proposed. Furthermore, grey forecast principle has the advantage of discovering the laws of historical changes, which can overcome the weakness of BP neural network prediction model and this method have higher prediction accuracy. The neural network prediction model was established by using the PM2.5 mass concentration of ten monitoring stations in Beijing in January and February 2018, as well as the corresponding hourly air pollutant concentration and meteorological factors. Meanwhile, the grey forecast algorithm was used to improve the neural network model. The results indicate that the improved method has the features of lower system error, and better fitting degree between the predicted result and the measured result.

Abstract：While autonomous driving in highway scenarios is already possible using only the vehicles sensors, auto nomous driving in the complex scenarios of big cities with all its different trafic participants is still a vision. Cellular vehicle-to-everything(C-V2X) communication is a necessary enabler of this vision and an emerging field of interest in today′s research. However, to the best of our knowledge open source simulators essential for open research do not exist yet. In this work we present our open source C-V2X mode 4 simulator based on the discrete-event network simulator NS-3. To analyze the performance of C-V2X mode 4 using our simulator, we created a worst case scenario and the 3GPP reference Manhattan grid scenario using the microscopic traffc simulator SUMO. In addition, we investigated the impact of the resource reservation period and the resource reselection probability on the system′s packet reception ratio.

Abstract：In order to identify the electromagnetic interference sources around the airport，the spectrum data of various interference sources around airports are collected. The characteristics of interference sources are summarized and artificial features are carried out. The peak value, spectrum energy and coefficient of variation are selected as identification features. Entropy method is used to measure the discreteness of feature set, and the feature points with similar feature values within class and big differences between classes are highlighted. Finally, a double weighted feature template is obtained by weighting. Each feature template is mapped to a point in the high-dimensional space, the Euclidean distance between the template point and the test point is calculated, the threshold is divided by the mean variance method, and the pattern recognition is carried out by the template matching method. The experimental results show that this method is stable, fast and efficient, and has a high recognition rate of wide-band EMI sources, which has practical value for interference sources identification around the airport.

Abstract：In order to realize the self-monitoring of human glucose, a glucose detection system based on MSP430 microcontroller was designed. Based on a conductive carbon cloth, a flexible glucose sensing electrode was constructed to fit the human skin; polyaniline and gold nanoparticles were loaded on the electrode, which increased the stability and sensitivity of the electrode instead of glucose oxidase; the host computer used a smartphone application to read the body′s glucose index at any time. The results showed that the system showed excellent response of measuring current, the detection range of glucose was 0.5 mM~1.5 mM, and the correlation coefficient R2 was 0.995 4. The system could complete the detection of glucose in human body.

Abstract：In automated testing,in order to avoid modifying test reports artificially,it is an optimal choice to generate PDF files directly.Based on this requirement,it studies the method of PDF document generation by PDFlib library under embedded system VxWorks,and carries out an application test in a practical example in this paper.The results show that this research can provide a practical method in automatic test and PDF report file generation.