Current Issue
15 October 2024, Volume 37 Issue 10
  • Privacy-Preserving Data Style Transfer Method for Artificial Intelligence of Things
    CHENG Jinke, LI Gaolei
    2024, 37(10):  1-5.  doi:10.16180/j.cnki.issn1007-7820.2024.10.001
    Abstract ( 39 )   HTML( 26 )   PDF (1628KB) ( 26 )  

    In the artificial intelligence of things, traditional privacy protection technologies mainly focus on the transmission, storage, and analysis stages of the data lifecycle, while ignoring the importance of protecting data privacy at the source. This study proposes a privacy-protecting data style transfer method for artificial intelligence of things. Based on cycle-consistent adversarial networks, a new loss function is added to obfuscate identity information, allowing real-style images and animation-style images to visually transform into each other. Animation-style data can be used to construct various virtual entities in the digital world (such as metaverses), and malicious users cannot reverse the original data based on the virtual entities or correctly identify the original data using the original deep learning model, thereby enhancing privacy protection for real entities in the physical world. Experimental results on a face dataset show that the transformed data reduces the accuracy of the ArcFace face recognition model by 30% without significantly reducing visual distortion.

    Heterogeneous Ant Optimization Based on Dynamic Entropy Evolution
    WANG Shike, YOU Xiaoming, YIN Ling, LIU Sheng
    2024, 37(10):  6-14.  doi:10.16180/j.cnki.issn1007-7820.2024.10.002
    Abstract ( 22 )   HTML( 11 )   PDF (3430KB) ( 11 )  

    In view of the overcome the problem of the slow convergence speed and low precision of ant colony algorithm in solving TSP(Traveling Salesman Problem), a heterogeneous ant optimization based on dynamic entropy evolution is proposed. In this algorithm, a heterogeneous double population is comprised of ACS(Ant Colony System) and MMAS(Max-Min Ant System),which is helpful to promote the complementary advantages between the populations. And the dynamic entropy evolution strategy is introduced to dynamically control the communication frequency between the populations by information entropy. The pheromones of the two populations' optimal common paths are fused to adjust the distribution of pheromones on the optimal paths of the low entropy populations, thereby effectively preserving the historical search information of the two populations and accelerating the convergence of the algorithm.The non-common path of the optimal solution of low entropy population is pseudo-initialized to expand its search range near the optimal solution and improve the accuracy of the solution, so as to realize the co-evolution of two populations.Simulation results show that the proposed algorithm can effectively balance the relationship between algorithm diversity and convergence when solving large-scale traveling salesman problems.

    Path Planning of Improved Artificial Potential Field Method Based on Deflection Angle Suppression
    JIN Tao, YU Lianzhi
    2024, 37(10):  15-22.  doi:10.16180/j.cnki.issn1007-7820.2024.10.003
    Abstract ( 31 )   HTML( 16 )   PDF (2462KB) ( 16 )  

    In view of the local minimum value, unreachable target and path oscillation in the practical application of traditional artificial potential field method, this study proposes an improved artificial potential field method based on deflection angle suppression. Based on the traditional artificial potential field method, this method adopts the improved repulsive potential field function to ensure that the target point is the lowest point of the whole situation. The deflection angle inhibitor is introduced in the path solution to suppress the excessive deflection angle during the driving process of the unmanned vehicle. After the unmanned vehicle falls into the local minimum point, a new virtual target point is added to the path solution at each step until the unmanned vehicle accumulates a certain deflection angle to get rid of the obstacle group. The simulation results show that the algorithm can reduce the volatility of the planning path without affecting the obstacle avoidance of the unmanned vehicle, and enable the unmanned vehicle to smoothly escape the complex obstacle group, smoothly reach the target point, and plan an effective, short and less oscillating path.

    Research on Agricultural Crop Diseases and Pests Classification Based on Masked Autoencoding
    JU Ping, SONG Yan, ZHANG Yingjie, XU Yifu, SHAO Ha...
    2024, 37(10):  23-29.  doi:10.16180/j.cnki.issn1007-7820.2024.10.004
    Abstract ( 32 )   HTML( 10 )   PDF (2207KB) ( 10 )  

    Crop diseases and insect pests cause a large amount of economic losses in agricultural production activities, and it is difficult to meet field production requirements of the current society if only relying on manual surveys by agronomist staffs. Applications of the machine vision technology can realize the automatic classification and detection of crop diseases and insect pests, and provide the guarantee for accurate and efficient agricultural productions. However, existing detection methods based on the deep learning framework and convolutional neural networks are constrained by factors such as rigid convolutional receptive field, inefficient data enhancement operator, and small sample size. In order to make up for the above shortcomings of existing detection technologies in term of recognition accuracy, a method for the classification of agricultural economic crop diseases and insect pests based on the masked autoencoding learning paradigm is proposed in this study. Through local random content masking, semantic feature extraction, and global context reconstruction of high-dimensional mapping of input crop images, the proposed algorithm can fully mine implicit representations of high-level semantics of images and model the long-distance contextual relationship in the same map, so as to train a more robust model with less data samples. Moreover, the model eliminates the interference of the high-frequency noise on the pre-training feature extraction processing by the relative total variational transformation. The results of comparison between the proposed method and current methods based on mainstream convolutional networks show that the proposed method can significantly improve the performance of existing methods, and the accuracy rate is improved from 90.48% to 95.24% based on ResNet50 benchmark network.

    Detection of Cervical Lesions Based on Multi-Scale Features and Attention Mechanism
    FENG Ting, YING Jie, YANG Haima, LI Fang
    2024, 37(10):  30-39.  doi:10.16180/j.cnki.issn1007-7820.2024.10.005
    Abstract ( 25 )   HTML( 14 )   PDF (5214KB) ( 14 )  

    CIN(Cervical Intraepithelial Neoplasm) is a precancerous lesion of the cervix with a high correlation to invasive cervical cancer. Accurate detection and classification of CIN is helpful to reduce the rate of severe cervical cancer. YOLOv5-CBTR(You Only Look Once version 5-Convolutional Block Transformer) cervical lesion detection method is proposed to address the issues of low accuracy in detection and classification of cervical lesions by combining multi-scale features and multiple attention mechanisms. The backbone network employs the SE-CSP (SENet-BottleneckCSP) with SENet (Squeeze-and-Excitation Networks) attention mechanism for feature extraction. The Transformer encoder module is introduced to fuse and amplify multi-feature information, and multi-head attention mechanism is used to enhance the feature extraction ability of lesion regions. Convolutional attention modules are introduced into the feature fusion layer for multiscale fusion of lesion feature information. The power transformation is introduced into the calculation of the boundary regression box, which speeds up the convergence of the model's loss function and realizes the detection and classification of cervical lesions. The experimental results show that the accuracy, recall rate, mAP(mean Average Precision), and F value of YOLOv5-CBTR model for the detection and classification of RGB cervical lesion images are 93.99%, 92.91%, 92.80%, and 93.45%, respectively. The mAP and F values of the model in multispectral cervical image detection and classification are 97.68% and 95.23%, respectively.

    Distributionally Robust Low-Carbon Economic Dispatch Considering New Energy Output
    ZHANG Tangqian, HE Yu, JIANG Muning, QIN Tingxiang...
    2024, 37(10):  40-47.  doi:10.16180/j.cnki.issn1007-7820.2024.10.006
    Abstract ( 24 )   HTML( 17 )   PDF (1318KB) ( 17 )  

    In order to enhance the consumption capacity of wind power and reduce carbon emissions, this study proposes a two-stage distributionally robust low-carbon economic dispatch model that takes into account carbon trading and electric vehicles. Carbon trading costs are introduced in the proposed study, and the cooperation between electric vehicle energy storage and wind power generation is utilized to reduce the system's carbon emissions and to increase the wind power consumption capacity. Considering the uncertainty of wind power, a distributionally robust optimization method based on the general moment uncertainty is established. The fuzzy set of distributionally robust is established using the moment information of the historical wind power output data that can be obtained to characterize the uncertain wind power output characteristics. The distributionally robust model is transformed into a quadratic programming model using duality and linear decision rules, and the model is solved through CPLEX. Experimental results show that the wind power consumption capacity increases by 11.35% and carbon emissions decrease by 1 579 t under the proposed method, which verifies the effectiveness and superiority of the two-stage distributionally robust model.

    Character Enhancement Based on Named Entity Recognition for Industrial Equipment Faults
    ZHANG Yang, LIU Jin
    2024, 37(10):  48-54.  doi:10.16180/j.cnki.issn1007-7820.2024.10.007
    Abstract ( 11 )   HTML( 5 )   PDF (1183KB) ( 5 )  

    To address the issues of sparse training data, complex entity structures, and uneven entity distribution in the industrial equipment failure domain, this study constructs an industrial equipment failure named entity recognition corpus. Due to the difficulty of character-level named entity recognition models in representing the professional vocabulary information in the field of industrial equipment failure, this study proposes a character-enhanced industrial equipment failure named entity recognition model to address this problem. In the embedding layer, professional vocabulary information is directly fused between the Transformer layers of RoBERTa-WWM (Robustly Optimized BERT Pretraining Approach with Whole Word Masking) to allocate word information to each of its constituent characters for enhanced semantics. The global semantic information is obtained through a BiLSTM(Bidirectional Long Short-Term Memory), and the CRF(Conditional Random Field) is used to learn the dependency relationship between adjacent labels to obtain the optimal sentence-level label sequence. Experimental results demonstrate that the proposed model has good performance on industrial equipment fault named entity recognition tasks, with an average F1 score of 92.403%.

    Blockchain-Based Data Sharing Mechanism for Multi-Domain Internet of Things
    ZHANG Di, LI Yunfa
    2024, 37(10):  55-63.  doi:10.16180/j.cnki.issn1007-7820.2024.10.008
    Abstract ( 18 )   HTML( 3 )   PDF (1975KB) ( 3 )  

    In view of the security and privacy protection issues faced in sharing sensing information data between sensing devices in the internet of things, this study proposes a blockchain-based multi-domain data sharing mechanism. The mechanism treats the data sharing problem as a data transaction problem and introduces a trusted regulator to guarantee the interests of both buyers and sellers of the transaction. The mechanism provides a data publishing method in which buyers can securely publish data to a cloud storage service and do not disclose information about the data during the publishing process. A ring signature is used to sign and verify the buyer's licence in order to protect the buyer's privacy from being compromised during the data transaction. In addition, the mechanism provides an anonymous evaluation method that allows buyers to evaluate transactions and monitor seller behaviour without revealing their identity. The security analysis and simulation experiments results show that the mechanism is secure and efficient, meets the requirements for secure data sharing in the internet of things.

    Research on Action Recognition Method Based on Deep Learning
    XIN Tenghao, LI Feifei
    2024, 37(10):  64-70.  doi:10.16180/j.cnki.issn1007-7820.2024.10.009
    Abstract ( 35 )   HTML( 13 )   PDF (1324KB) ( 13 )  

    The key of current research on behavior recognition algorithms based on deep learning lies in enhancing the accuracy and stability of key point extraction, in order to achieve more accurate action recognition of targets. However, many current algorithms tend to just add attention mechanisms that appear to perform better in the feature extraction stage of the target, without considering the impact of different attention mechanisms on different models and tasks. Therefore, this study proposes an algorithmic model for pose estimation based on various attention mechanisms, which further highlights the importance of selecting an appropriate attention mechanism by comparing the impact of different attention mechanisms on the model. In addition, considering the stability of key point extraction, the initialization of the model is fine-tuned to select a more suitable initialization method that improves the performance by increasing the category of weights on network layer judgments. Compared with the performance of the benchmark network model, the model enhances all evaluation metrics on both multiscale and no-multiscale CrowdPose datasets, where the average accuracy improvement in both cases is more than 1%.

    Method of Suppressing Ripple of DC-Side Voltage of Angle-Connected SVG Based on Boost-Type APD
    ZHENG Shicheng, LIU Hairui, XU Quanwei, LANG Jiaho...
    2024, 37(10):  71-80.  doi:10.16180/j.cnki.issn1007-7820.2024.10.010
    Abstract ( 16 )   HTML( 5 )   PDF (2365KB) ( 5 )  

    In the cascaded H-bridge multilevel SVG(Static Var Generator), the inherent structural characteristics of the H-bridge lead to the presence of secondary ripple in the DC-side capacitor voltage of the angular cascaded SVG, and the voltage ripple affect the degree of compensation of power quality and even lead to system dysregulation. In order to suppress the secondary ripple of the voltage on the H-bridge DC side of the angular cascaded SVG, this study proposes a design solution for the angular cascaded H-bridge multilevel SVG based on Boost-type active power decoupling. The APD(Active Power Decoupling) circuit separates fluctuating power and stable power from each other and uses energy storage elements to absorb fluctuating power, which weakens the voltage ripple on the DC side of the H-bridge. The simulation platform was built and simulated using MATLAB/Simulink software. The simulation results show that the angular cascaded H-bridge multilevel SVG based on Boost-type active power decoupling can effectively suppress the secondary ripple of the voltage of the DC-side capacitor.

    Emotion Recognition Algorithm Based on Multimodal Cross-Interaction
    ZHANG Hui, LI Feifei
    2024, 37(10):  81-87.  doi:10.16180/j.cnki.issn1007-7820.2024.10.011
    Abstract ( 38 )   HTML( 11 )   PDF (1735KB) ( 11 )  

    Due to the limitations of single modality emotion recognition, many researchers have shifted their focus to the field of multimodal emotion recognition. Multi-modal emotion recognition focuses on two problems: The optimal extraction of the features of each mode and the effective fusion of the extracted features. This study proposes an emotion recognition method based on multimodal cross-interaction to capture the diversity of modality expressions. The editors of various modalities separately extract features with emotional information, and the stacked interaction modules based on the attention mechanism between modalities model the potential relationship among vision, text and audio. Experiments are conducted on CMU-MOSI and CMU-MOSEI datasets for emotion recognition based on text, audio and visual. The results show that the method achieved the scores of 86.5%, 47.7%, 86.4%, 0.718, 0.776, and 83.4%, 51.5%, 83.4%, 0.566, 0.737 on five indicators, Acc2(Accuracy2)、Acc7(Accuracy7)、F1、MAE(Mean Absolute Error) and Corr(Correlation). This demonstrates that the proposed algorithm significantly improves performance, and also validates that the cross-mapping mutual representation mechanism perform better than single-modal representation methods.

    Low Power Consumption Wearable Electrocardiogram Monitoring System
    ZHANG Peng, JIANG Mingfeng, LI Yang
    2024, 37(10):  88-94.  doi:10.16180/j.cnki.issn1007-7820.2024.10.012
    Abstract ( 21 )   HTML( 11 )   PDF (2442KB) ( 11 )  

    Daily electrocardiogram monitoring is of great significance for the prevention of cardiovascular diseases. However, medical-grade electrocardiogram monitoring systems are expensive, complicated to operate, and not suitable for home-based electrocardiogram monitoring. Wearable devices often ignore the interference of noise, and the signals collected are difficult to analyze and diagnose. This study presents a low-power wearable electrocardiogram monitoring system, which consists of four parts:Power management module, electrocardiogram acquisition module, data processing module, and delay switch module. The system uses the BMD101 chip to acquire human electrocardiogram signals and sends the electrocardiogram data to a mobile device via low-power Bluetooth chip nrf52832. To address the muscle electrical interference noise that is easily introduced into the electrocardiogram signals, a denoising algorithm based on VMD (Variational Mode Decomposition) is proposed. The NLM(Non-Local Mean) filter is used to remove the noise in the low-frequency mode of the electrocardiogram signal, and the DWT(Discrete Wavelet Transform) threshold denoising algorithm is used to eliminate the high-frequency mode noise of the electrocardiogram signal. The reconstructed signal quality is significantly improved. The experimental results show that the proposed algorithm has the characteristics of low cost, easy portability and convenient use, and can obtain high quality electrocardiogramsignals, meet the needs of users for long-range monitoring, and solve the problems of inconvenient daily electrocardiogrammonitoring and low quality electrocardiogram signal acquisition.

Download

Monthly,Founded in September 1987
Competent Authorities:
Ministry of Education of the People's Republic of China
Sponsored by:Xidian University
Chief Editor:Liao Guisheng
Executive Editor:Wan Liancheng
Editor:Hei Lei
Editor and Publisher:
The Editorial Department of Electronic Science and Technology
Distribution Abroad:
China Intermational Book Trading Corporation
P.O.BOx 399,Beijing 100044,China
Address:
P.O.Box 375,2 Taibai Road(South),Xi'an 710071,China
Tel/Fax:0086-029-88202440
Website:http://www.dianzikeji.org
E-mail:dzkj@mail.xidian.edu.cn
Unit Price:$20.00