面向智能任务的语义通信：理论、技术和挑战

doi:10.11959/j.issn.1000-436x.2022117

通信学报 ›› 2022, Vol. 43 ›› Issue (6): 41-57.doi: 10.11959/j.issn.1000-436x.2022117

• 专题：面向6G的智能至简网络关键技术 • 上一篇下一篇

面向智能任务的语义通信：理论、技术和挑战

刘传宏¹, 郭彩丽¹^,², 杨洋², 陈九九¹, 朱美逸¹, 孙鲁楠¹

¹ 北京邮电大学先进信息网络北京实验室，北京 100876
² 北京邮电大学网络体系构建与融合北京市重点实验室，北京 100876

修回日期:2022-05-09 出版日期:2022-06-01 发布日期:2022-06-01
作者简介:刘传宏（1998- ），男，安徽池州人，北京邮电大学博士生，主要研究方向为深度学习、语义通信、资源分配等
郭彩丽（1977- ），女，山西太原人，博士，北京邮电大学教授、博士生导师，主要研究方向为语义通信、无线移动通信技术、认知无线电、信号检测与估值、车联网、可见光通信、视觉智能计算、社交跨媒体数据挖掘与分析等
杨洋（1991- ），男，湖南娄底人，博士，北京邮电大学讲师，主要研究方向为可见光通信、室内定位技术、车联网技术、语义通信技术等
陈九九（1994- ），男，湖南平江人，北京邮电大学博士生，主要研究方向为车联网资源分配、语义通信、强化学习算法等
朱美逸（1999- ），女，湖北保康人，北京邮电大学博士生，主要研究方向为语义通信、车联网通信、强化学习算法等
孙鲁楠（1996- ），女，辽宁葫芦岛人，北京邮电大学博士生，主要研究方向为语义通信、图像传输、信源信道编码等
基金资助:
中央高校基本科研业务费专项资金资助项目(2021XD-A01-1);北京市自然科学基金资助项目(4202049);北京邮电大学博士生创新基金资助项目(CX2022101)

Intelligent task-oriented semantic communications:theory, technology and challenges

Chuanhong LIU¹, Caili GUO¹^,², Yang YANG², Jiujiu CHEN¹, Meiyi ZHU¹, Lu’nan SUN¹

¹ Beijing Laboratory of Advanced Information Networks, Beijing University of Posts and Telecommunications, Beijing 100876, China
² Beijing Key Laboratory of Network System Construction and Integration, Beijing University of Posts and Telecommunications, Beijing 100876, China

Revised:2022-05-09 Online:2022-06-01 Published:2022-06-01
Supported by:
The Fundamental Research Funds for the Central Universities(2021XD-A01-1);The Beijing Natural Science Foundation(4202049);BUPT Excellent Ph.D.Students Foundation(CX2022101)

摘要/Abstract

摘要：

目的：未来机-机、人-机万物智能互联对传统通信方式提出了挑战，提取信源语义信息进行传输的语义通信方法为6G通信系统提供了新的解决方法。然而如何度量语义信息、如何实现最优的语义编解码等均存在挑战，本文综述现有语义通信相关的论文，提出面向智能任务的语义通信方法和框架，为进一步推动语义通信的发展铺平道路。

方法：首先综述了语义通信的发展历程和研究现状，通过分析总结了语义通信面临的两大瓶颈问题，提出了面向智能任务的语义通信方法。针对语义熵难度量的问题，本文通过定义构成语义消息的最小基本单元为语义元，引入模糊数学理论刻画语义理解的模糊程度，给出语义信息熵的计算表达式。紧接着，本文基于信息瓶颈理论提出了语义信息编码方案和语义信道联合编码方案，分别考虑了接收端是否需要重建原始信源的场景；此外，从神经网络可解释性的角度出发，提出了基于可解释性的语义编码方法。最后，基于USRP和LabView等软硬件搭建了面向智能任务的语义通信平台，对所提算法进行性能验证。

结果：在需要重建信源的通信场景中，本文所提的语义通信方法可以大大提升信源数据的压缩比，大幅降低传输的数据量；在相同的压缩比下可以提升接收端执行后续智能任务的性能，同时提升信源重建的性能。在无需重建信源的场景中，语义通信方式可以在极大压缩比的情况下，较好地完成智能任务，这是因为语义通信传输图像的语义信息而非图像的所有数据，大大减小了其带宽需求，实现语义通信的带宽利用率超出传统通信方式的100倍。此外，语义通信方式抗噪声性能远远好于传统通信方法，这是因为语义通信方法传输的数据保留了图像的语义特征，且模型训练时考虑了信道噪声的影响，使智能任务性能更优，具有更好的鲁棒性。语义通信方式由于传输数据量大大减少，因此在带宽资源相同的情况下，传输时延显著下降；此外，由于不需要进行图像的重构，软硬件的处理负荷减小，处理时延也有所下降。因此本文所提的方案可以在保证高精度分类性能的同时，大幅减少了端到端智能任务的时延。

结论：面向智能任务的语义通信方法相较于传统通信方法具有明显优势，可以在大幅降低传输数据量的同时提升接收端智能任务的性能，因此语义通信将继续保持快速发展的趋势。然而，语义通信中仍有大量的基础概念和基础问题亟需进一步讨论和完善，如语义信息的基础理论、语义通信的统一架构和语义通信中的资源分配策略等等，对这些问题进行探讨和研究对推动6G时代的技术创新和突破具有重要意义，需要学术同仁共同推动实现。

关键词: 6G, 语义熵, 语义通信, 语义编码, 智能任务

Abstract:

Objectives: In the future, intelligent interconnection of all things, such as machine-to-machine and human-to-machine, poses challenges to traditional communication methods. The semantic communication method that extracts semantic information from source information and transmits them provides a novel solution for the sixth generation (6G) communication system. However, there are challenges in how to measure semantic information and how to achieve optimal semantic codec.This paper reviews the existing works related to semantic communication,and proposes a semantic communication method and framework for intelligent tasks,paving the way for further development of semantic communication.

Methods: Firstly, the development history and research status of semantic communication are reviewed, the two bottleneck problems faced by semantic communication are analyzed and summarized, and a semantic communication method oriented to intelligent tasks is proposed. Aiming at the difficulty of semantic entropy,this paper defines the smallest basic unit of semantic message as semantic element,introduces fuzzy mathematics theory to describe the fuzzy degree of semantic understanding, and gives the calculation expression of semantic information entropy. Then, based on the information bottleneck theory, this paper proposes a semantic information coding scheme and a semantic channel joint coding scheme,respectively considering whether the receiver needs to reconstruct the original source. Furthermore, from the perspective of neural network interpretability,an interpretability-based semantic encoding method is proposed.Finally, a semantic communication platform for intelligent tasks is built based on software and hardware such as USRP and LabView,and the performance of the proposed algorithm is verified.

Results:In the communication scenario where the source needs to be reconstructed,the semantic communication method proposed in this paper can greatly improve the compression ratio of the source data and greatly reduce the amount of transmitted data.Under the same compression ratio, the performance of the receiver to perform subsequent intelligent tasks can be improved,and the performance of source reconstruction can be improved at the same time.In scenarios where there is no need to reconstruct the source,the semantic communication method can better accomplish intelligent tasks with a large compression ratio.This is because semantic communication transmits the semantic information of the image instead of all the data of the image,which greatly reduces its bandwidth requirements,and the bandwidth utilization rate of semantic communication is 100 times higher than that of traditional communication methods. In addition, the anti-noise performance of the semantic communication method is much better than that of the traditional communication method, because the data transmitted by the semantic communication method retains the semantic features of the image,and the influence of channel noise is considered during model training, which makes the performance of intelligent tasks better and makes the communication system more robust. The semantic communication method greatly reduces the amount of data transmitted, so the transmission delay is significantly reduced under the same bandwidth resources.In addition,since image reconstruction is not required,the processing load of software and hardware is reduced, and the processing delay is also reduced. Therefore, the scheme proposed in this paper can greatly reduce the delay of end-to-end intelligent tasks while ensuring high-precision classification performance.

Conclusions: Compared with traditional communication methods, the semantic communication method oriented to intelligent tasks has obvious advantages,which can greatly reduce the amount of transmitted data and improve the performance of intelligent tasks at the receiving end. Therefore, semantic communication will continue to maintain the trend of rapid development. However,there are still a lot of basic concepts and basic problems in semantic communication that need to be further discussed and improved,such as the basic theory of semantic information,the unified architecture of semantic communication,and the resource allocation strategy in semantic communication. Research is of great significance to promoting technological innovation and breakthroughs in the 6G era,and academic colleagues need to jointly promote the realization.

Key words: 6G, semantic entropy, semantic communication, semantic coding, intelligent task

中图分类号:

TN929.5

刘传宏, 郭彩丽, 杨洋, 陈九九, 朱美逸, 孙鲁楠. 面向智能任务的语义通信：理论、技术和挑战[J]. 通信学报, 2022, 43(6): 41-57.

Chuanhong LIU, Caili GUO, Yang YANG, Jiujiu CHEN, Meiyi ZHU, Lu’nan SUN. Intelligent task-oriented semantic communications:theory, technology and challenges[J]. Journal on Communications, 2022, 43(6): 41-57.

图/表 22

图1

图2

图3

表1

图4

表2

3种语义失真度量方法的总结与对比"

语义失真度量方法	数学表达形式	优缺点
基于孪生神经网络	$d (Φ (X), Φ (X'))$	便于计算，但算力需求较大
基于语义互信息	$I (X; Y) - I (X; Z)$	难以计算，可基于神经网络估计
基于语义信息熵	$H (X) - H (X')$	度量结果准确，但难以计算

表2

图5

图6

图7

图8

图9

图10

图11

图12

表3

表4

图13

图14

图15

图16

图17

图18

参考文献 55

[1]	CHEN M Z , CHALLITA U , SAAD W ,et al. Artificial neural networks-based machine learning for wireless networks:a tutorial[J]. IEEE Communications Surveys ＆ Tutorials, 2019,21(4): 3039-3071.
[2]	STRINATI E C , BARBAROSSA S . 6G networks:beyond Shannon towards semantic and goal-oriented communications[J]. Computer Networks, 2021,190: 107930.
[3]	WEAVER W . Recent contributions to the mathematical theory of communication[J]. ETC:A Review of General Semantics, 1953,10(4): 261-281.
[4]	XIE H Q , QIN Z J , LI G Y ,et al. Deep learning enabled semantic communication systems[J]. IEEE Transactions on Signal Processing, 2021,69: 2663-2675.
[5]	CARNAP R , BAR-HILLEL Y , . An outline of a theory of semantic information[R]. 1952.
[6]	POPPER K . Conjectures and refutations[M]. New York: Routledge, 2002.
[7]	ZADEH L A . Fuzzy sets[J]. Information and Control, 1965,8(3): 338-353.
[8]	ZADEH L A . Probability measures of fuzzy events[J]. Journal of Mathematical Analysis and Applications, 1968,23(2): 421-427.
[9]	LUCA A D , TERMINI S . Algebraic properties of fuzzy sets[J]. Journal of Mathematical Analysis and Applications, 1972,40(2): 373-386.
[10]	FLORIDI L . Outline of a theory of strongly semantic information[J]. Minds and Machines, 2004,14(2): 197-221.
[11]	LU C G . From Bayesian inference to logical Bayesian inference:a new mathematical frame for semantic communication and machine learning[J]. arXiv Preprint,arXiv:1809.01577, 2018.
[12]	LU C G , . The third kind of Bayes’ theorem links membership functions to likelihood functions and sampling distributions[C]// International Conference on Cognitive Systems and Signal Processing. Berlin:Springer, 2019: 268-280.
[13]	钟义信 . 信息科学原理(第 3 版)[M]. 北京: 北京邮电大学出版社, 2002.
	ZHONG Y X . Principles of information science(3rd edition)[M]. Beijing: Beijing University of Posts and Telecommunications Press, 2002.
[14]	钟义信 . 面向智能研究的全信息理论:纪念 Shannon 信息论 50 周年[J]. 北京邮电大学学报, 1998,21(4): 1-6.
	ZHONG Y X . Intelligence oriented comprehensive information theory:in memory of the 50th anniversary of Shannon information theory[J]. Journal of Beijing University of Posts and Telecommunications, 1998,21(4): 1-6.
[15]	钟义信, 张瑞 . 信息生态学与语义信息论[J]. 图书情报知识, 2017(6): 4-11.
	ZHONG Y X , ZHANG R . Information ecology and semantic information theory[J]. Documentation,Information ＆ Knowledge, 2017(6): 4-11.
[16]	石光明, 李莹玉, 谢雪梅 . 语义通讯:智能时代的产物[J]. 模式识别与人工智能, 2018,31(1): 91-99.
	SHI G M , LI Y Y , XIE X M . Semantic communications:outcome of the intelligence era[J]. Pattern Recognition and Artificial Intelligence, 2018,31(1): 91-99.
[17]	BAO J , BASU P , DEAN M K ,et al. Towards a theory of semantic communication[C]// Proceedings of 2011 IEEE Network Science Workshop. Piscataway:IEEE Press, 2011: 110-117.
[18]	POPOVSKI P , SIMEONE O , BOCCARDI F ,et al. Semantic-effectiveness filtering and control for post-5G wireless connectivity[J]. Journal of the Indian Institute of Science, 2020,100(2): 435-443.
[19]	KOUNTOURIS M , PAPPAS N . Semantics-empowered communication for networked intelligent systems[J]. IEEE Communications Magazine, 2021,59(6): 96-102.
[20]	YANG Y , GUO C L , LIU F F ,et al. Semantic communications with AI tasks[J]. arXiv Preprint,arXiv:2109.14170, 2021.
[21]	KALFA M , GOK M , ATALIK A ,et al. Towards goal-oriented semantic signal processing:applications and future challenges[J]. Digital Signal Processing, 2021,119: 103134.
[22]	牛凯, 戴金晟, 张平 ,等. 面向6G的语义通信[J]. 移动通信, 2021,45(4): 85-90.
	NIU K , DAI J C , ZHANG P ,et al. 6G-oriented semantic communications[J]. Mobile Communications, 2021,45(4): 85-90.
[23]	ZHANG P , XU W J , GAO H ,et al. Toward wisdom-evolutionary and primitive-concise 6G:a new paradigm of semantic communication networks[J]. Engineering, 2022,8: 60-73.
[24]	石光明, 肖泳, 李莹玉 ,等. 面向万物智联的语义通信网络[J]. 物联网学报, 2021,5(2): 26-36.
	SHI G M , XIAO Y , LI Y Y ,et al. Semantic communication networking for the intelligence of everything[J]. Chinese Journal on Internet of Things, 2021,5(2): 26-36.
[25]	UYSAL E , KAYA O , EPHREMIDES A ,et al. Semantic communications in networked systems[J]. arXiv Preprint,arXiv:2103.05391, 2021.
[26]	LAN Q , WEN D , ZHANG Z ,et al. What is semantic communication? a view on conveying meaning in the era of machine intelligence[J]. Journal of Communications and Information Networks, 2021,6(4): 336-371.
[27]	FARSAD N , RAO M , GOLDSMITH A . Deep learning for joint source-channel coding of text[C]// Proceedings of 2018 IEEE International Conference on Acoustics,Speech and Signal Processing. Piscataway:IEEE Press, 2018: 2326-2330.
[28]	XIE H Q , QIN Z J . A lite distributed semantic communication system for internet of things[J]. IEEE Journal on Selected Areas in Communications, 2021,39(1): 142-153.
[29]	WENG Z Z , QIN Z J . Semantic communication systems for speech transmission[J]. IEEE Journal on Selected Areas in Communications, 2021,39(8): 2434-2444.
[30]	BOURTSOULATZE E , KURKA D B , GüNDüZ D . Deep joint source-channel coding for wireless image transmission[J]. IEEE Transactions on Cognitive Communications and Networking, 2019,5(3): 567-579.
[31]	LEE C H , LIN J W , CHEN P H ,et al. Deep learning-constructed joint transmission-recognition for Internet of things[J]. IEEE Access, 2019,7: 76547-76561.
[32]	JANKOWSKI M , GüNDüZ D , MIKOLAJCZYK K . Wireless image retrieval at the edge[J]. IEEE Journal on Selected Areas in Communications, 2021,39(1): 89-100.
[33]	刘传宏, 郭彩丽, 杨洋 ,等. 人工智能物联网中面向智能任务的语义通信方法[J]. 通信学报, 2021,42(11): 97-108.
	LIU C H , GUO C L , YANG Y ,et al. Intelligent task-oriented semantic communication method in artificial intelligence of things[J]. Journal on Communications, 2021,42(11): 97-108.
[34]	XIE H Q , QIN Z J , LI G Y . Task-oriented multi-user semantic communications for VQA[J]. IEEE Wireless Communications Letters, 2022,11(3): 553-557.
[35]	XIE H , QIN Z , TAO X ,et al. Task-oriented multi-user semantic communications[J]. arXiv Preprint,arXiv:2112.10255, 2021.
[36]	李蕾, 周延泉, 钟义信 . 基于语用的自然语言处理研究与应用初探[J]. 智能系统学报, 2006,1(2): 1-6.
	LI L , ZHOU Y Q , ZHONG Y X . Pragmatic information based NLP research and application[J]. CAAI Transactions on Intelligent Systems, 2006,1(2): 1-6.
[37]	LUCA A D , TERMINI S . A definition of a nonprobabilistic entropy in the setting of fuzzy sets theory[J]. Information and Control, 1972,20(4): 301-312.
[38]	吴伟陵 . 广义信息源与广义熵[J]. 北京邮电大学学报, 1982(1): 29-41.
	WU W L . Generalized information source and generalized entropy[J]. Journal of Beijing University of Posts and Telecommunications, 1982(1): 29-41.
[39]	BASU P , BAO J , DEAN M ,et al. Preserving quality of information by using semantic relationships[J]. Pervasive and Mobile Computing, 2014,11: 188-202.
[40]	HOANG D T , LONG P M , VITTER J S . Rate-distortion optimizations for motion estimation in low-bit-rate video coding[C]// Digital Video Compression:Algorithms and Technologies.[S.l.:s.n.], 1996: 18-27.
[41]	HUYNH-THU Q , GHANBARI M . Scope of validity of PSNR in image/video quality assessment[J]. Electronics Letters, 2008,44(13): 800.
[42]	WANG Z , BOVIK A C . A universal image quality index[J]. IEEE Signal Processing Letters, 2002,9(3): 81-84.
[43]	WANG Z , BOVIK A C , SHEIKH H R ,et al. Image quality assessment:from error visibility to structural similarity[J]. IEEE Transactions on Image Processing, 2004,13(4): 600-612.
[44]	KIM J , LEE S . Deep learning of human visual sensitivity in image quality assessment framework[C]// Proceedings of 2017 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway:IEEE Press, 2017: 1969-1977.
[45]	ZHANG R , ISOLA P , EFROS A A ,et al. The unreasonable effectiveness of deep features as a perceptual metric[C]// Proceedings of 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway:IEEE Press, 2018: 586-595.
[46]	HU Y Q , ZHOU W , ZHAO S X ,et al. SDM:semantic distortion measurement for video encryption[C]// Proceedings of 2018 13th IEEE International Conference on Automatic Face ＆ Gesture Recognition. Piscataway:IEEE Press, 2018: 764-768.
[47]	HUSSAIN A J , AL-FAYADH A , RADI N . Image compression techniques:a survey in lossless and lossy algorithms[J]. Neurocomputing, 2018,300: 44-69.
[48]	YANG Z H , WANG Y H , XU C ,et al. Discernible image compression[C]// Proceedings of the 28th ACM International Conference on Multimedia. New York:ACM Press, 2020: 1561-1569.
[49]	韦来生 . 数理统计(第2版)[M]. 北京: 科学出版社, 2015.
	WEI L S . Mathematical statistics(2nd edition)[M]. Beijing: Science Press, 2015.
[50]	SHWARTZ-ZIV R , TISHBY N . Opening the black box of deep neural networks via information[J]. arXiv Preprint,arXiv:1703.00810, 2017.
[51]	LI Y Y , ZHAO P P , WANG D Q ,et al. Learning disentangled user representation based on controllable VAE for recommendation[C]// International Conference on Database Systems for Advanced Applications. Berlin:Springer, 2021: 179-194.
[52]	TISHBY N , PEREIRA F C , BIALEK W . The information bottleneck method[J]. arXiv Preprint,arXiv:physics/0004057, 2000.
[53]	QIN Z J , TAO X M , LU J H ,et al. Semantic communications:principles and challenges[J]. arXiv Preprint,arXiv:2201.01389, 2022.
[54]	LIU F F , TONG W J , SUN Z F ,et al. Task-oriented semantic communication with semantic reconstruction:an extended rate-distortion theory based scheme[J]. arXiv Preprint,arXiv:2201.10929, 2022.
[55]	陈九九, 冯春燕, 郭彩丽 ,等. 车联网中视频语义驱动的资源分配算法[J]. 通信学报, 2021,42(7): 1-11.
	CHEN J J , FENG C Y , GUO C L ,et al. Video semantics-driven resource allocation algorithm in Internet of vehicles[J]. Journal on Communications, 2021,42(7): 1-11.

	指导理论	编码方式	算力需求	评价指标	编码目标
语义编码	广义率失真理论^[40]	基于神经网络	GPU/CPU	智能任务性能	提升码元的语义信息量
传统信源编码	率失真理论	非神经网络	CPU	编码效率	提高码元的平均信息量

训练参数	参数取值
Epochs	10
Batchsize	20
优化器	随机梯度下降
学习率	0.001
动量	0.9

实验参数	参数取值
天线距离/m	1.8
发收频率/GHz	5
调制方式	64QAM
码率	0.43
带宽/MHz	40

面向智能任务的语义通信：理论、技术和挑战

Intelligent task-oriented semantic communications:theory, technology and challenges

在线阅读

PDF下载

可视化

摘要/Abstract

引用本文

使用本文

图/表 22

参考文献 55

相关文章 15

Metrics

推荐阅读 0

[1]	李荣鹏, 汪丙炎, 张宏纲, 赵志峰. 知识增强的语义通信接收端设计[J]. 通信学报, 2023, 44(6): 70-76.
[2]	张平, 牛凯, 姚圣时, 戴金晟. 面向未来的语义通信：基本原理与实现方法[J]. 通信学报, 2023, 44(5): 1-14.
[3]	石光明, 杨旻曦, 高大化, 柴靖轩. 面向语义信息直传的通信架构[J]. 通信学报, 2023, 44(5): 15-27.
[4]	秦志金, 赵菼菼, 李凡, 陶晓明. 多模态语义通信研究综述[J]. 通信学报, 2023, 44(5): 28-41.
[5]	张平, 戴金晟, 张育铭, 王思贤, 秦晓琦, 牛凯. 面向语义通信的非线性变换编码[J]. 通信学报, 2023, 44(4): 1-14.
[6]	江沸菠, 彭于波, 董莉. 面向6G的深度图像语义通信模型[J]. 通信学报, 2023, 44(3): 198-208.
[7]	王晓云, 张小舟, 马良, 王亚娟, 楼梦婷, 姜涛, 金婧, 王启星, 刘光毅. 6G通信感知一体化网络的感知算法研究与优化[J]. 通信学报, 2023, 44(2): 219-230.
[8]	杨静雅, 唐晓刚, 周一青, 刘玲, Jiangzhou Wang. 意图抽象与知识联合驱动的6G内生智能网络架构[J]. 通信学报, 2023, 44(2): 12-26.
[9]	张海君, 陈安琪, 李亚博, 隆克平. 6G移动网络关键技术[J]. 通信学报, 2022, 43(7): 189-202.
[10]	廖建新, 付霄元, 戚琦, 王敬宇, 孙海峰. 6G-ADM：基于知识空间的6G网络管控体系[J]. 通信学报, 2022, 43(6): 3-15.
[11]	王志勤, 江甲沫, 刘沛西, 曹晓雯, 李阳, 韩凯峰, 杜滢, 朱光旭. 6G联邦边缘学习新范式：基于任务导向的资源管理策略[J]. 通信学报, 2022, 43(6): 16-27.
[12]	李昂, 陈建新, 魏昕, 周亮. 面向6G的跨模态信号重建技术[J]. 通信学报, 2022, 43(6): 28-40.
[13]	唐盼, 林佳欣, 张建华, 田磊, 常钊玮, 夏亮, 王启星. 面向6G的太赫兹信道反射特性研究[J]. 通信学报, 2022, 43(5): 102-109.
[14]	张晓茜, 徐勇军. 面向零功耗物联网的反向散射通信综述[J]. 通信学报, 2022, 43(11): 199-212.
[15]	周振宇, 贾泽晗, 廖海君, 赵雄文, 张磊. 基于上下文学习的电力物联网接入控制方法[J]. 通信学报, 2021, 42(3): 150-159.