基于改进深度残差网络算法的智能干扰识别

doi:10.11959/j.issn.1000-0801.2022233

摘要/Abstract

摘要：

现网运维人力成本高、效率低，如何快速精准识别网络干扰类型、提高运维人员工作效率，成为亟待解决的问题。提出一种基于改进深度残差网络（ResNet）的智能干扰识别方法，通过对接运营商北向网管的通信干扰数据接口，对干扰数据进行采集和预处理，并结合现网专家经验对历史干扰数据类型进行标注和校正，形成离线干扰数据集。再将干扰频域信息进行图像化生成干扰频谱波形图，并针对不同干扰类型进行图像处理和数据处理。之后根据业务特点对传统ResNet算法进行改进，通过提取单一干扰类型特征，确定各特征在复合干扰类型中的权重，达到对任意干扰类型识别的目的。最后通过导入已训练好的模型对干扰数据进行在线识别，有效提高干扰识别的准确率和效率。

关键词: 5G, 干扰识别, 深度残差网络

Abstract:

The labor cost of the current network operation and maintenance is high and the efficiency is low.How to quickly and accurately identify the type of network interference, and improve the work efficiency of the maintenance personnel has become an urgent problem to be solved.An intelligent interference identification method of improved deep residual network (ResNet) was studied.The interference data was collected and preprocessedby connecting with the communication interference data interface of the operator's northbound network management.The type of itwaslabelled and corrected by the current network experts to form an offline interference data set.Then the interference frequency domain information was used to generate the interference spectrum waveform image, and performed image processing and data processing for different interference types.After that, the traditional Res Net algorithm was improved according to the business characteristics to extract the features of single interference type.The features were weighted in the compound interference type to identify any type of interference.Finally, the interference data was identified online by importing the trained model, which effectively improved the accuracy and efficiency of interference identification.

Key words: 5G, interference identification, ResNet

中图分类号:

TN929.5

马键, 张广晋, 张磊, 戴经纬. 基于改进深度残差网络算法的智能干扰识别[J]. 电信科学, 2022, 38(10): 98-106.

Jian MA, Guangjin ZHANG, Lei ZHANG, Jingwei DAI. Improved ResNet algorithm based intelligent interference identification[J]. Telecommunications Science, 2022, 38(10): 98-106.

图/表 8

图1

图2

图3

图4

图5

图6

表1

常见5G干扰类型"

类型	特征	干扰源	常见场所
LTE D频干扰	D1/D2 频段对应的底噪明显抬升，多为三段	LTE小区的上行终端	未完成 LTE 清频的区
	式或凹形		域
NR系统内干扰	随NR系统内上行资源分配波动	系统内上行负载高站点	重叠覆盖区域
无线网桥/视频监控干扰	宽带干扰，一个或多个直方凸起，中间一般	视频监控的无线网桥、无线回	楼宇、小区，如电
	有一个裂缝	传等设备	梯、停车场等
干扰器	全频段或大带宽底噪抬升且有些波动	干扰器产生的外部干扰	学校附近
路由器	呈现尖峰状，尖峰的底部稍宽，上头变窄，	路由器	楼宇、小区
	频域位置不固定
智能路灯干扰	左低右高，规律性波动干扰	智能路灯	城市、农村街道
伪基站干扰	干扰波形呈现规律的1.4 MHz /3 MHz /	伪基站	公安局附近
	5 MHz /10 MHz 等带宽的矩形波形
时钟源故障	前12个RB受较高的干扰，有的是前	常见于基站时钟源抖动/跳	随机出现
	35个RB	变、GPS故障等产生的上下行
		交叉时隙干扰
广电信号干扰	左边8 MHz一个方波	广电信号
电信800 MHz干扰	右边带杂散干扰特征	电信800 MHz
LTE D频干扰、NR系统	LTE D频干扰和NR系统内的复合干扰	系统内上行负载高站点LTE小	重叠覆盖区域未完成
内干扰		区的上行终端	LTE清频的区域
LTE D频干扰、干扰器	LTE D频干扰和干扰器叠加的复合干扰
NR系统内干扰、干扰器	NR系统内干扰和干扰器叠加的复合干扰
LTE D 频干扰、NR 系统	LTE D 频干扰、NR 系统内干扰、干扰器叠
内干扰、干扰器	加的复合干扰
LTE D频干扰、NR系统	LTE D频干扰、NR系统内干扰、无线网桥/
内干扰、无线网桥/视频	视频监控干扰叠加的复合干扰
监控干扰

表1

表2

参考文献 21

[1]	段晓东, 孙滔 . 面向融合、智慧、低碳的5G技术演进[J]. 电信科学, 2022,38(3): 3-9.
	DUAN X D , SUN T . Towards converged,intelligent,low-carbon characterized 5G-Advanced technologies[J]. Tele-communications Science, 2022,38(3): 3-9.
[2]	DAHLMAN E , PARKVALL S , SKOLD J . 5G NR:the next generation wireless access technology[M]. New York: Academic, 2020.
[3]	宋心刚, 张冬晨, 李行政 ,等. 2.6 GHz频段5G上行干扰分析与识别研究[J]. 电信工程技术与标准化, 2021,34(4): 74-81.
	SONG X G , ZHANG D C , LI X Z ,et al. Research on interfer-ence analyzing and recognition of 5G in 2.6 GHz[J]. Telecom Engineering Technics and Standardization, 2021,34(4): 74-81.
[4]	陈凯 . 5G 干扰特征识别及解决方案研究[J]. 邮电设计技术, 2021(4): 34-39.
	CHEN K . Research on 5G network interference characteristics recognition and solution[J]. Designing Techniques of Posts and Telecommunications, 2021(4): 34-39.
[5]	DEB S , MONOGIOUDIS P . Learning based uplink interference management in 4G LTE cellular systems[J]. IEEE/ACM Transactions on Networking, 2015,23(2): 398-411.
[6]	DARWEESH T A , ELSAYED K . LTE uplink interference-aware scheduling using high interference and overload indicators[C]// Proceedings of IFIP Wireless ＆ Mobile Networking Conference. Piscataway:IEEE Press, 2016.
[7]	宋心刚, 张冬晨, 李行政 ,等. 5G NR 上行干扰问题研究[J]. 电信工程技术与标准化, 2020,33(3): 9.
	SONG X G , ZHANG D C , LI X Z ,et al. Research on uplink in-terference of 5G NR[J]. Telecom Engineering Technics and Standardization, 2020,33(3): 9.
[8]	韩文冬 . 5G 基站上行干扰分析工具设计与实现[J]. 现代信息科技, 2021,5(4): 53-55,59.
	HAN W D . Design and implementation of uplink interference analysis tool for 5G base station[J]. Modern Information Tech-nology, 2021,5(4): 53-55,59.
[9]	李攀攀, 谢正霞, 乐光学 ,等. 基于深度学习的无线通信接收方法研究进展与趋势[J]. 电信科学, 2022,38(2): 1-17.
	LI P P , XIE Z X , YUE G X ,et al. Research progress and trends of deep learning based wireless communication receiving method[J]. Telecommunications Science, 2022,38(2): 1-17.
[10]	SCHMIDHUBER J . Deep learning in neural networks:an overview[J]. Neural Networks, 2015(61): 85-117.
[11]	LU L Q , YI Y H , HUANG F L ,et al. Integrating local CNN and global CNN for script identification in natural scene images[J]. IEEE Access, 2019(7): 52669-52679.
[12]	SCHMIDHUBER J . Deep learning in neural networks[J]. Neural Netw, 2015.
[13]	YANG W , JING Z , YANG C ,et al. A deep CNN method for underwater image enhancement[C]// Proceedings of 2017 IEEE International Conference on Image Processing (ICIP). Piscataway:IEEE Press, 2017.
[14]	DOLZ J , AYED I B , JING Y ,et al. HyperDense-Net:a hyper-densely connected CNN for multi-modal image segmentation[J]. IEEE Transactions on Medical Imaging, 2018(99).
[15]	HE K M , ZHANG X Y , REN S Q ,et al. Deep residual learning for image recognition[C]// Proceedings of 2016 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway:IEEE Press, 2016: 770-778.
[16]	LI B Q , HE Y Y . An improved ResNet based on the adjustable shortcut connections[J]. IEEE Access, 2018(6): 18967-18974.
[17]	ROY S K , MANNA S , SONG T C ,et al. Attention-based adaptive spectral-spatial kernel ResNet for hyperspectral image classification[J]. IEEE Transactions on Geoscience and Remote Sensing, 2021,59(9): 7831-7843.
[18]	YOO H , HAN S , CHUNG K . Diagnosis support model of cardiomegaly based on CNN using ResNet and explainable feature map[J]. IEEE Access, 2021(9): 55802-55813.
[19]	WU Z , SHEN C , HENGEL A . Wider or deeper:revisiting the ResNet model for visual recognition[J]. Pattern Recognition, 2016(90): 119-133.
[20]	JIAO J B , TU W C , HE S F ,et al. FormResNet:formatted residual learning for image restoration[C]// Proceedings of 2017 IEEE Conference on Computer Vision and Pattern Recognition Workshops. Piscataway:IEEE Press, 2017: 1034-1042.
[21]	YUE Z , LUO J Q , HUSHENG F ,et al. Study on the identification method of human upper limb flag movements based on inception-ResNet double stream network[J]. IEEE Access, 2020(9): 85764-85784.

正确标签	是/个	否/个	总计/个	准确率
LTE D频干扰	14	1	15	93.3%
NR系统内干扰	232	16	248	93.5%
无线网桥/视频监控干扰	7	2	9	77.8%
干扰器	104	1	105	99.0%
路由器	1	0	1	100.0%
智能路灯干扰	12	1	13	92.3%
电信800 MHz干扰	3	0	3	100.0%
其他干扰	0	32	32	0
LTE D频干扰、NR系统内干扰	197	9	206	95.6%
LTE D频干扰、干扰器	2	0	2	100.0%
LTE D频干扰、其他干扰	0	1	1	0
NR系统内干扰、干扰器	15	0	15	100.0%
NR系统内干扰、路由器	23	2	25	92.0%
NR系统内干扰、电信800 MHz干扰	1	0	1	100.0%
NR系统内干扰、无线网桥/视频监控干扰	2	0	2	100.0%
LTE D频干扰、NR系统内干扰、干扰器	4	0	4	100.0%
总计	617	65	682	90.5%