基于多流ConvNeXt网络和马氏距离度量的未知信号增量识别

doi:10.16183/j.cnki.jsjtu.2022.426

上海交通大学学报 ›› 2024, Vol. 58 ›› Issue (4): 481-491.doi: 10.16183/j.cnki.jsjtu.2022.426

基于多流ConvNeXt网络和马氏距离度量的未知信号增量识别

肖易寒¹, 刘序斌¹, 于祥祯², 赵忠凯¹()

1.哈尔滨工程大学先进船舶通信与信息技术工业和信息化部重点实验室,哈尔滨 150000
2.上海无线电设备研究所,上海 201100

收稿日期:2022-10-28 修回日期:2022-12-24 接受日期:2022-12-30 出版日期:2024-04-28 发布日期:2024-04-30
通讯作者: 赵忠凯,副教授;E-mail:zhaozhongkai@hrbeu.edu.cn.
作者简介:肖易寒(1980-),博士,从事基于机器学习的雷达信号识别研究.
基金资助:
国防科技基础加强计划(2019-JCJQ-ZD-067-00)

Unknown Signal Incremental Recognition Based on Multi-Flow ConvNeXt Network and Mahalanobis Distance Metric

XIAO Yihan¹, LIU Xubin¹, YU Xiangzhen², ZHAO Zhongkai¹()

1. Key Laboratory of Advanced Marine Communication and Information Technology of the Ministry of Industry and Information Technology, Harbin Engineering University, Harbin 150000, China
2. Shanghai Radio Equipment Research Institute, Shanghai 201100, China

Received:2022-10-28 Revised:2022-12-24 Accepted:2022-12-30 Online:2024-04-28 Published:2024-04-30

摘要/Abstract

摘要：

为解决现阶段基于深度学习网络的信号识别技术无法实现未知信号增量识别的问题,提出了基于多流ConvNeXt网络和马氏距离度量(MDM)相结合的未知信号增量识别方法.首先,利用改进的多流ConvNeXt网络提取信号的属性特征;其次,使用马氏距离度量判决方法进行未知信号检测进而实现已知信号和未知信号的二分类;最后,该方法根据不断增加的未知信号对模型的参数进行自动更新,使模型具备了自我进化的能力,进而可以识别出不断增加的新的未知信号类别,实现对未知信号的增量识别.仿真实验结果表明,该方法对未知信号的平均识别率达到97%以上.

关键词: 未知信号, 多流ConvNeXt网络, 马氏距离度量, 增量识别

Abstract:

In order to solve the problem that the signal recognition technology based on deep learning network cannot currently realize the incremental recognition of unknown signals, a method for incremental recognition of such unknown signals, based on the combination of the multi-flow ConvNeXt network and Mahalanobis distance metric (MDM) is proposed. First, the improved multi-flow ConvNeXt network is used to extract the attribute features of signals. Then, the MDM judgement method is used to detect unknown signals, and apply the binary classification for known and unknown signals. Finally, the parameters of the model is automatically updated according to the increasing number of unknown signals. In such way, the model has the ability of self-evolution, and it has the ability to recognize incrementally more types of unknown signals.The simulation results show that the average recognition rate of unknown signals is more than 97%.

Key words: unknown signal, multi-flow ConvNeXt network, Mahalanobis distance metric, incremental recognition

中图分类号:

TN911.3

肖易寒, 刘序斌, 于祥祯, 赵忠凯. 基于多流ConvNeXt网络和马氏距离度量的未知信号增量识别[J]. 上海交通大学学报, 2024, 58(4): 481-491.

XIAO Yihan, LIU Xubin, YU Xiangzhen, ZHAO Zhongkai. Unknown Signal Incremental Recognition Based on Multi-Flow ConvNeXt Network and Mahalanobis Distance Metric[J]. Journal of Shanghai Jiao Tong University, 2024, 58(4): 481-491.

图/表 13

图1

图2

图3

图4

表1

表2

图5

表3

表4

表5

图6

表6

图7

参考文献 25

[1]	KAMENCAY P, BENCO M, MIZDOS T, et al. A new method for face recognition using convolutional neural network[J]. Advances in Electrical and Electronic Engineering, 2017, 15(4): 663-672.
[2]	TIAN Y, YANG G, WANG Z, et al. Apple detection during different growth stages in orchards using the improved YOLO-V3 model[J]. Computers and Electronics in Agriculture, 2019, 157: 417-426. doi: 10.1016/j.compag.2019.01.012 URL
[3]	KANG Y, CAI Z, TAN C W, et al. Natural language processing (NLP) in management research: A literature review[J]. Journal of Management Analytics, 2020, 7(2): 139-172. doi: 10.1080/23270012.2020.1756939 URL
[4]	肖易寒, 王亮, 郭玉霞. 基于去噪卷积神经网络的雷达信号调制类型识别[J]. 电子与信息学报, 2021, 43(8): 2300-2307.
	XIAO Yihan, WANG Liang, GUO Yuxia. Radar signal modulation type recognition based on denoising convolutional neural network[J]. Journal of Electronics & Information Technology, 2021, 43(8): 2300-2307.
[5]	全大英, 陈赟, 唐泽雨, 等. 基于双通道卷积神经网络的雷达信号识别[J]. 上海交通大学学报, 2022, 56(7): 877-885. doi: 10.16183/j.cnki.jsjtu.2021.209
	QUAN Daying, CHEN Yun, TANG Zeyu, et al. Radar signal recognition based on dual channel convolutional neural network[J]. Journal of Shanghai Jiao Tong University, 2022, 56(7): 877-885.
[6]	XIAO Y H, LIU W J, GAO L. Radar signal recognition based on transfer learning and feature fusion[J]. Mobile Networks and Applications, 2020, 25(4): 1563-1571. doi: 10.1007/s11036-019-01360-1
[7]	金丽洁, 武亚涛. 基于双CNN的雷达信号调制类型识别方法[J]. 空天防御, 2022, 5(1): 66-70.
	JIN Lijie, WU Yatao. Radar signal modulation type recognition based on double CNN[J]. Air & Space Defense, 2022, 5(1): 66-70.
[8]	郭业才, 姚文强. 基于信噪比分类网络的调制信号分类识别算法[J]. 电子与信息学报, 2022, 44(10): 3507-3515.
	GUO Yecai, YAO Wenqiang. Modulation signal classification and recognition algorithm based on signal to noise ratio classification network[J]. Journal of Electronics & Information Technology, 2022, 44(10): 3507-3515.
[9]	GUO Y, ZHANG Y, MA H. Modulation recognition of digital signals based on deep belief network[J]. IOP Conference Series: Materials Science and Engineering, 2019, 563: 052009. doi: 10.1088/1757-899X/563/5/052009
[10]	YASHASHWI K, SETHI A, CHAPORKAR P. A learnable distortion correction module for modulation recognition[J]. IEEE Wireless Communications Letters, 2018, 8(1): 77-80. doi: 10.1109/LWC.2018.2855749 URL
[11]	AFAN A, FAN Y, SHU L. Automatic modulation classification of digital modulation signals with stacked autoencoders[J]. Digital Signal Processing, 2017, 71: 108-116. doi: 10.1016/j.dsp.2017.09.005 URL
[12]	DONG Y, JIANG X, ZHOU H, et al. SR2CNN: Zero-shot learning for signal recognition[J]. IEEE Transactions on Signal Processing, 2021, 69: 2316-2329. doi: 10.1109/TSP.2021.3070186 URL
[13]	YOSHIHASHI R, SHAO W, KAWAKAMI R, et al. Classification-reconstruction learning for open-set recognition[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Long Beach, USA: IEEE, 2019: 4016-4025.
[14]	CHENG Z, WANG S, ZHANG P, et al. Improved autoencoder for unsupervised anomaly detection[J]. International Journal of Intelligent Systems, 2021, 36(12): 7103-7125. doi: 10.1002/int.v36.12 URL
[15]	YUE Z, WANG T, SUN Q, et al. Counterfactual zero-shot and open-set visual recognition[C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Beijing, China: IEEE, 2021: 15404-15414.
[16]	LIU Z, MAO H, WU C Y, et al. A ConvNet for the 2020s[C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. New Orleans, USA: IEEE, 2022: 11966-11976.
[17]	MAHALANOBIS P C. Mahalanobis distance[J]. Proceedings National Institute of Science of India, 1936, 49(2): 234-256.
[18]	HE K M, ZHANG X M, REN S Q, et al. Deep residual learning for image recognition[C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Las Vegas, USA: IEEE, 2016: 770-778.
[19]	LIU Z, LIN Y, CAO Y, et al. Swin transformer: Hierarchical vision transformer using shifted windows[C]// Proceedings of the IEEE International Conference on Computer Vision. Beijing, China: IEEE, 2021: 10012-10022.
[20]	MA L, YANG X, TAO D. Person re-identification over camera networks using multi-task distance metric learning[J]. IEEE Transactions on Image Processing, 2014, 23(8): 3656-3670. doi: 10.1109/TIP.2014.2331755 pmid: 24956368
[21]	LUO Y, LIU T, TAO D, et al. Decomposition-based transfer distance metric learning for image classification[J]. IEEE Transactions on Image Processing, 2014, 23(9): 3789-3801. doi: 10.1109/TIP.2014.2332398 pmid: 24968169
[22]	KALINTHA W, ONO S, NUMAO M, et al. Kernelized evolutionary distance metric learning for semi-supervised clustering[J]. Intelligent Data Analysis, 2019, 23(6): 1271-1297. doi: 10.3233/IDA-184283 URL
[23]	O’SHEA T J, CORGAN J, CLANCY T C. Convolutional radio modulation recognition networks[C]// International Conference on Engineering Applications of Neural Networks. Aberdeen, UK: Springer International Publishing, 2016: 213-226.
[24]	HOWARD A, SANDLER M, CHU G, et al. Searching for mobilenetv3[C]// Proceedings of the IEEE/CVF International Conference on Computer Vision. Seoul, South Korea: IEEE, 2019: 1314-1324.
[25]	SZEGEDY C, IOFFE S, VANHOUCKE V, et al. Inception-v4, Inception-ResNet and the impact of residual connections on learning[C]// Thirty-First AAAI Conference on Artificial Intelligence. San Francisco, USA: AAAI, 2017: 4278-4284.

配置	型号	参数
处理器及内存	Intel(R) Xeon(R) Silver 4110	2.1 GHz 32 GB内存
操作系统	Ubuntu 18.04.1	64位
显卡	NVIDIA Tesla T4	15 GB显存
语言	Python	3.7.9
框架	Pytorch	1.7.1
运算平台	CUDA	10.1

类别	预测的已知信号	预测的未知信号
真实的已知信号	T_P	F_N
真实的未知信号	F_P	T_N

α₁	T_P	F_N	T_N	F_P	R_TK	R_TU
0.00	0	3 600	800	0	0	1
0.33	2 664	936	798	2	0.740	0.998
0.66	3 500	100	791	9	0.972	0.989
1.00	3 564	36	504	296	0.990	0.630

实际信号	预测信号										识别率/%
实际信号	8PSK	AM-DSB	BPSK	CPFSK	PAM4	QAM16	QAM64	QPSK	WBFM	未知信号	识别率/%
8PSK	385	0	1	0	0	1	0	1	0	12	96.3
AM-DSB	0	213	0	0	0	0	0	0	167	20	53.3
BPSK	2	0	374	0	0	0	0	2	0	22	93.5
CPFSK	0	0	0	394	0	0	0	0	0	6	98.5
PAM4	0	0	0	0	391	0	0	0	0	9	97.8
QAM16	8	0	0	0	0	342	23	2	0	25	85.5
QAM64	3	0	0	0	0	31	339	1	0	26	84.8
QPSK	0	0	1	0	0	0	2	386	0	11	96.5
WBFM	0	57	0	0	0	0	0	0	324	19	81.0
AM-SSB	0	0	0	0	0	0	1	0	0	399	99.8
GFSK	0	0	0	0	0	0	0	0	8	392	98.0

实际信号	预测信号
实际信号	未知1	未知2	未知3	未知4	未知5	未知6
8PSK	4	3	1	3	0	1
AM-DSB	10	5	0	2	3	0
BPSK	2	8	3	7	2	0
CPFSK	0	3	3	0	0	0
PAM4	0	3	5	1	0	0
QAM16	7	9	6	2	1	0
QAM64	1	6	12	2	4	1
QPSK	0	1	8	0	2	0
WBFM	0	13	0	3	2	1
AM-SSB	395	3	0	0	1	0
GFSK	2	386	1	2	1	0

基于多流ConvNeXt网络和马氏距离度量的未知信号增量识别

Unknown Signal Incremental Recognition Based on Multi-Flow ConvNeXt Network and Mahalanobis Distance Metric

RichHTML

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

图/表 13

参考文献 25

相关文章 1

编辑推荐

Metrics

本文评价

识别参数类型	识别率/%
识别参数类型	本文算法	SR2CNN	Inception-v4	MobileNetV3
8PSK	96.3	85.5	85.7	87.5
AM-DSB	53.3	73.5	65.0	72.3
BPSK	93.5	95.5	94.5	88.5
CPFSK	98.6	99.0	87.0	89.8
PAM4	97.8	94.5	85.2	87.0
QAM16	85.5	49.3	71.7	36.5
QAM64	84.8	44.0	72.3	44.5
QPSK	96.5	90.5	89.0	91.2
WBFM	81.0	32.0	66.1	22.5
平均识别率	87.5	73.7	79.6	68.8
R_TK	97.2	95.9	92.9	91.4
R_TU	98.8	99.5	97.8	92.6