Table of Content

15 March 2025 Volume 38 Issue 3

  

Electromagnetic Signal Modulation Recognition Based on Complex-Valued Deep Neural Network

YUAN Depin, ZHAO Liang, GE Xiansheng

Electronic Science and Technology. 2025, 38(3):  1-6.  doi:10.16180/j.cnki.issn1007-7820.2025.03.001

Abstract ( 58 )   HTML ( 9 )   PDF (2836KB) ( 65 )  

Figures and Tables | References | Related Articles | Metrics

In the region of complex electromagnetic environment, it is difficult to obtain the signal modulation type. The traditional recognition and classification methods of modulated signals are not successful because of their own defects. The current deep learning methods usually used for signal modulation are based on real values for characterization and processing, which results in recognition bias due to the loss of the original intrinsic connection of complex values. To solve this situation, the complex deep neural network is applied to the modulation recognition of electromagnetic signals, complex convolutional deep neural networks such as complex convolutional deep neural networks, batch normalization and fully connected networks are designed, and the final classification task is completed by softmax function. The standard data set RML2016.10a is used to complete the training as well as testing of the network. The experimental results show that the trained complex deep neural network is significantly better than traditional recognition algorithms, and can effectively improve the recognition rate of electromagnetic signals.

A Nonlinear Representation-Based Probabilistic Latent Factorization Tensor Model

DONG Jiaying, SONG Yan, LI Ming

Electronic Science and Technology. 2025, 38(3):  7-15.  doi:10.16180/j.cnki.issn1007-7820.2025.03.002

Abstract ( 30 )   HTML ( 4 )   PDF (981KB) ( 10 )  

Figures and Tables | References | Related Articles | Metrics

In view of the filling problem of non-negative incomplete data with extremely sparse and unbalanced data, a probabilistic potential factor tensor model is proposed. The data sparsity is mitigated by reasonably assuming the probability distribution of the data as a priori information. Nonlinear mapping is used to realize nonlinear characterization of each non-negative element in the data and improve the characterization ability of the model. Considering the unbalance of data, the weights based on instance frequency are added to the traditional regularization terms to increase the effectiveness and pertinence of regularization terms. The experimental results show that the proposed model has obvious improvement over the existing model in terms of completion accuracy and time cost.

Research on MFC-Based Piezoelectric Vibration Energy Harvester

HE Yue, YUAN Tianchen, YANG Jian, SONG Ruigang

Electronic Science and Technology. 2025, 38(3):  16-21.  doi:10.16180/j.cnki.issn1007-7820.2025.03.003

Abstract ( 28 )   HTML ( 4 )   PDF (2447KB) ( 14 )  

Figures and Tables | References | Related Articles | Metrics

Mechanical vibration is a common form of motion in the environment, and energy harvesting of it can satisfy the power needs of low-power devices such as wireless sensor nodes. Piezoelectric power generation has a good application prospect. In view of the problem that piezoelectric power generation is limited by the lack of piezoelectric materials, this study designs a piezoelectric vibration energy harvester based on MFC (Macro Fiber Composite), which utilizes the large deformation vibration of the circular structure and the good flexibility of MFC to realize the low-frequency vibration energy harvesting. The output performance of the MFC piezoelectric sheet at different positions of the circular ring is investigated, and the results show that the output voltage of the collector reaches the peak value when the angle between the MFC piezoelectric sheet and the guide post is 97°. Through finite element simulation and experimental comparison verification, it is found that the resonance frequency of the collector is 3.8 Hz, and the output voltage of the experimental prototype at resonance is 6.60 V and the power reaches 0.17 mW at 0.3g excitation, which is basically in line with the simulation and experimental results.

Self-Supervised Network Intrusion Detection Model Based on Graph Contrastive Learning

WANG Ziyi, CHEN Shiping

Electronic Science and Technology. 2025, 38(3):  22-31.  doi:10.16180/j.cnki.issn1007-7820.2025.03.004

Abstract ( 32 )   HTML ( 6 )   PDF (3059KB) ( 15 )  

Figures and Tables | References | Related Articles | Metrics

Traditional methods for detecting network traffic anomalies suffer from issues such as neglecting network topology and high costs associated with acquiring labeled data, leading to lower model accuracy and generalization. This study proposes a detection approach based on graph neural networks and self-supervised learning. Based on the characteristics of network traffic data, the self-supervised graph comparison learning task is constructed, and the traffic graph is enhanced by edge feature transformation and edge masking to generate comparison samples. The graph encoder based on GraphSAGE(Graph SAmple and aggreGatE)is improved to make full use of correlation to enrich the feature representation of nodes, and the parameters of the graph encoder are trained with InfoNCE loss function suitable for comparative learning to achieve self-learning feature representation, get rid of the dependence on network traffic label data, and improve the accuracy of network abnormal traffic detection. The experimental results show that the proposed model performs well in detecting abnormal network traffic without label data, with F1 values reaching 92.64% and 90.97% on two public data sets, respectively.

Research on Elastic Scaling Method of Kubernetes Container Cloud

LI Jiaying, YANG Zemin, SONG Zhedai, ZHU Jinrong

Electronic Science and Technology. 2025, 38(3):  32-39.  doi:10.16180/j.cnki.issn1007-7820.2025.03.005

Abstract ( 30 )   HTML ( 2 )   PDF (998KB) ( 11 )  

Figures and Tables | References | Related Articles | Metrics

Kubernetes container cloud is currently a popular cloud computing technology, and its default elastic scaling method HPA(Horizontal Pod Autoscaler) can horizontally expand and shrink cloud native applications.However, this method is based on a single load index, which is difficult to apply to diversified cloud-native applications. In addition, the method performs elastic expansion based on the current load, so that the process of expansion and contraction has obvious hysteresis. This method is based on the sliding time window algorithm for elastic shrinkage, which is slow and easy to waste system resources.To solve these problems, an improved elastic stretching method is proposed in this paper. A dynamic weighted fusion algorithm is designed to fuse multiple load indicators into comprehensive load factors, which can fully reflect the comprehensive load of cloud native applications.CEEMDAN(Complete Ensemble Empirical Mode Decomposition with Adaptive Noise)-ARIMA(Autoregressive Integrated Moving Average Model) prediction model is proposed, and elastic expansion is realized in advance to cope with the burst traffic based on the predicted load value of the model.A method combining rapid capacity reduction and sliding time window is proposed to reduce system resource waste on the basis of ensuring application service quality.Experimental results show that compared with the HPA mechanism, the improved elastic scaling method shortens the average response time by 336.55% when dealing with the first burst traffic, reduces system resource usage by 50% after the traffic ends, and can quickly expand the capacity when encountering burst traffic again, with an average response time shortened by 66.83%.

A Remote Sensing Image Water Extraction Method by Combining Atrous Convolution and Pooling Models

ZHAO Yunfei, XUE Cunjin

Electronic Science and Technology. 2025, 38(3):  40-46.  doi:10.16180/j.cnki.issn1007-7820.2025.03.006

Abstract ( 26 )   HTML ( 2 )   PDF (2425KB) ( 11 )  

Figures and Tables | References | Related Articles | Metrics

The interference of spectral foreign objects such as vegetation, shadows and clouds leads to the low integrity and poor extraction effect of remote sensing images. A remote sensing image water extraction model MAP_UNet(A UNet of Combining Multi Atrous Convolution and Pooling Model) is proposed in this study, which combines multi-level cavity convolution and pooling model. The model uses UNet as the reference codec network to extract different dimensional features of water bodies, introduces double recursive residual module to prevent gradient disappearance of deep network, and uses multiple modules to integrate spatial cavity convolution and maximum pooling to capture a larger range of feature information and further strengthen the feature semantic relationship of adjacent scales. In order to verify the effectiveness and advance of the proposed method, experiments are carried out using high-resolution visible remote sensing image data and the results are compared with open deep learning semantic segmentation algorithm. The experimental results show that the MAP_UNet model has achieved good results in extracting accuracy and preventing misdetection of different objects of the same spectrum. The accuracy rate, recall rate, F1-Score and MIoU(Mean Intersection over Union) are 96.20%, 92.64%, 87.27% and 89.10%, respectively. Compared with UNet(U-shaped Network), UNet++ and UNet ASPP(UNet with Atrous Spatial Pyramid Pooling Network) networks, the performance of proposed method has significant improvement.

Integration of CNN and Transformer for Retinal OCT Image Fluid Segmentation Method

CHEN Yuyang, LI Feng

Electronic Science and Technology. 2025, 38(3):  47-59.  doi:10.16180/j.cnki.issn1007-7820.2025.03.007

Abstract ( 26 )   HTML ([an error occurred while processing this directive] )   PDF (2708KB) ([an error occurred while processing this directive])  

Figures and Tables | References | Related Articles | Metrics

In view of the problems such as small size, heterogeneous shape and fuzzy details of the fluid accumulation area, this study integrates CNN(Convolutional Neural Networks) and Transformer to propose an innovative multi-branch segmentation network. The network consists of full convolutional path, Transformer path and CNN-Transformer fusion path. The fully convolutional path is used to capture detailed features of the lesion area, while the Transformer path extracts multi-scale non-local feature information with long-range dependencies. The fusion path takes advantage of both CNN and Transformer to make up for the shortcomings of other branches. The features of the three branches are integrated through the prediction head to generate the final segmentation map. The performance of retinal effusion segmentation is tested on Kermany, UMN and DUKE data sets for intraretinal effusion and subretinal effusion. The experimental results show that the Dice coefficient of the proposed method is 86.63%, the crossover ratio is 77.02%, the sensitivity is 89.47%, and the accuracy is 85.51%, which proves its effectiveness and provides a feasible solution for the automatic segmentation of retinal effusion.

Optimization of 3D Coverage in Wireless Sensor Networks Based on Improved SO Algorithm

GAO Zhixiang, PANG Feifei, SONG Peikun

Electronic Science and Technology. 2025, 38(3):  60-67.  doi:10.16180/j.cnki.issn1007-7820.2025.03.008

Abstract ( 25 )   HTML ( 2 )   PDF (1647KB) ( 9 )  

Figures and Tables | References | Related Articles | Metrics

In view of the problems such as slow convergence and incomplete coverage of the basic snake algorithm when applied to the coverage of 3D wireless sensor networks, an improved snake algorithm is proposed to optimize the coverage of spatial networks. Chebyshev polynomial is used to improve the initial population of the snake optimization algorithm, so that the population of the algorithm has a better initial position, which lays a foundation for the subsequent position update. In the exploration stage and the development stage of snake optimization algorithm, spiral sine strategy and piecewise chaotic mapping are introduced respectively to intervene in population position update, so that the population can keep a high search ability and search range. The improved snake optimization algorithm is applied to the coverage problem of wireless sensor three-dimensional space network. The simulation results show that compared with the basic snake optimization algorithm, the improved snake optimization algorithm can improve the coverage of space network by 20%, enhance the overall performance of the network, and has practicability and robustness.

Design of Ultrasonic Piezoelectric Transducer Impedance Matching

FU Yihui, LIU Yu, XIAO Hanlin, WANG Mingjie

Electronic Science and Technology. 2025, 38(3):  68-74.  doi:10.16180/j.cnki.issn1007-7820.2025.03.009

Abstract ( 23 )   HTML ( 1 )   PDF (2029KB) ( 11 )  

Figures and Tables | References | Related Articles | Metrics

To improve the measurement accuracy of ultrasonic flow meters,it is necessary to match the impedance of excitation signal source and ultrasonic piezoelectric transducer. However, the existing impedance matching devices of ultrasonic piezoelectric transducers not only rely on imports but also have the disadvantages of poor operation and flexibility. In order to meet the requirements of low manufacturing, maintenance cost and convenient operation, two impedance matching circuits of series inductor and parallel inductor are designed in this study. An analog switch circuit is designed to realize the transmission and reception control of ultrasonic signals, a signal amplifier circuit is responsible for the amplification of ultrasonic signals, and a single-chip microcomputer is used to realize channel switching of the analog switch circuit and data collection of ultrasonic propagation time. The simulation results show that the series inductor impedance matching optimizes the input signal waveform and improves the driving efficiency of the ultrasonic piezoelectric transducer. The experimental results show that after the series inductor impedance matching, the ultrasonic propagation time data measured by the microcontroller is advanced by 50 μs, the accuracy of the ultrasonic wave propagation time is improved. But the internal resistance of the excitation signal source is small, and the parallel inductance impedance matching effect is not good.

Positive High Voltage Charge Pump Applied to Flash-Based FPGA

JIANG Shaoxiang, YU Shenglin, MA Jinlong, WU Chubin

Electronic Science and Technology. 2025, 38(3):  75-81.  doi:10.16180/j.cnki.issn1007-7820.2025.03.010

Abstract ( 24 )   HTML ( 1 )   PDF (1639KB) ( 10 )  

Figures and Tables | References | Related Articles | Metrics

When Flash-based FPGA(Filed Programmable Gate Array) performs programming operations, the charge pump provides positive high voltage to the gate end of the programming tube. In order to meet the power on timely operation, and programming stability of Flash-based FPGA, the charge pump is required not only be able to output high voltage, but also have fast startup speed and small output voltage ripple. This study proposes a positive high-voltage charge pump based on the traditional cross coupled charge pump. The main body of the charge pump adopts a parallel dual branch structure, reducing the output voltage ripple. It uses a six phase non overlapping clock and a new clock boost module to control the timing of the charge pump, eliminating the impact of reverse current and improving the start speed of the charge pump. A voltage stabilizing module is set at the output end for voltage regulation, ensuring programming stability. The simulation results show that under the conditions of 3.3 V supply voltage, 20 MHz clock frequency and 50 pF load capacitance, the charge pump startup time is 6.6 μs, the output voltage is stable to 15 V, and the output ripple is only 23 mV. After using 0.18 μm CMOS(Complementary Metal Oxide Semiconductor) technology, the test results meet the programming requirements of Flash FPGA.

Handshake Protocol Design of Transceiver Module Based on FPGA High-Speed Serdes Interface

LIU Zhengqiang, HONG Xujian, SUN Weihong

Electronic Science and Technology. 2025, 38(3):  82-87.  doi:10.16180/j.cnki.issn1007-7820.2025.03.011

Abstract ( 23 )   HTML ( 2 )   PDF (1583KB) ( 10 )  

Figures and Tables | References | Related Articles | Metrics

In order to improve the stability of FPGA(Field Programmable Gate Array) high-speed Serdes communication and monitor its communication status in real time, a communication protocol based on K-code control characters is designed. Two K-code control characters, dynamic SOF(Start of Frame) that marks the start of user data frames and static EOF(End of Frame) that marks the end, are created to facilitate the continuity detection of communication. K-code control characters TLINK(Transmit Link) and BLINK(Back Link) are created, where the TLINK control character is sent regularly at the Serdes sender. After the Serdes receives the TLINK control character, it controls the sender of its own side to output the BLINK control character for feedback, establishing a handshake relationship between the two communicating parties, which is beneficial for communication timeout and status detection. Verifying user data independent of SOF and EOF for CRC32(Cyclic Redundancy Check32) calculation is beneficial for communication error detection. The experimental results show that the protocol can accurately monitor the number of frame losses, error codes, timeouts, and communication disconnection duration of the Serdes link, with a minimum time accuracy of 10 μs.

Research on Iris Segmentation Algorithm Based on Ensemble Learning

SUN Jiaqian, ZHU Jinrong, ZHANG Xiaobao, ZHANG Yunkai, GONG Weijuan

Electronic Science and Technology. 2025, 38(3):  88-94.  doi:10.16180/j.cnki.issn1007-7820.2025.03.012

Abstract ( 26 )   HTML ( 4 )   PDF (2040KB) ( 21 )  

Figures and Tables | References | Related Articles | Metrics

In view of the problems such as inaccurate detail segmentation, unsmooth boundary segmentation and easy to be affected by noise, an iris segmentation algorithm based on ensemble learning is proposed in this study. Compared with the traditional ensemble learning algorithm, the appropriate model is selected as the base learner based on Pearson coefficient method, so as to improve the performance of ensemble learning. U²-Net, DeepLabv3+ and PSPNet are selected as homogeneous individual learners to train on the CASIA-Iris-Interval dataset, and the corresponding iris segmentation prediction results are obtained. CLAHE and Gamma correction and other image enhancement operations are performed to obtain a new prediction result graph. Weighted average method is selected as an integrated algorithm to integrate the prediction results of the basis learner, so as to obtain the final prediction results of iris segmentation. The test results show that the accuracy of the proposed algorithm is improved by 1%, the average crossover ratio is improved by 3.8%, the average macro score is improved by 2.4%, and the visual effect and objective evaluation index have better segmentation effect when compared with the base learner under three different evaluation indexes.