Robust Loss Functions for Signal Modulation Recognition with Noise Labels

doi:10.19596/j.cnki.1001-246x.8437

Abstract

Abstract:

In view of the fact that the labeling of signal modulation type is prone to errors in applications, that is, the underlying training data set has label noise, we propose l₁ norm based loss function and its extended form as robust loss function of deep convolutional neural network, which is recognized as one of the most excellent feature extraction network, to classify signal modulation types with label noisy. The algorithm achieves high accuracy even if the label noise level of training data set is up to 50%. By contrast, it is unable to predict the type of signal modulation by using usual cross entropy as the loss function of the deep convolutional neural network. Robustness of the algorithm is verified with numerical examples on public available benchmark data sets.

Key words: l₁ norm based loss function, q loss function, signal modulation, label noisy, signal recognition

Xiao-bo WANG, Jun-ping YIN, Yan XU. Robust Loss Functions for Signal Modulation Recognition with Noise Labels[J]. Chinese Journal of Computational Physics, 2022, 39(4): 386-394.

Figures/Tables 8

Fig.1 Architecture of CNN

Fig.2 Classification accuracy using cross entropy loss function under different label noise rate and SNR

Fig.3 Local comparison of classification accuracy using cross entropy loss function under different label noise rate and SNR

Fig.4 Classification accuracy using l1 loss function under different label noise rate and SNR

Fig.5 Local comparison of classification accuracy using l1 loss function under different label noise rate and SNR

Fig.6 Classification accuracy using q loss function under different label noise rate and SNR

Fig.7 Local comparison of classification accuracy using q loss function under different label noise rate and SNR

Table 1 Average classification accuracy with cross entropy loss function, l1 loss function and q loss function under different label noise rate

标签噪声率	交叉熵损失函数/%	l₁模损失函数/%	q损失函数/% (q = 0.7)
0.0	54.98	47.23	53.18
0.1	53.82	41.28	54.21
0.2	53.10	41.11	54.01
0.3	52.38	38.72	52.73
0.4	9.09	9.09	51.21
0.5	9.09	38.18	52.63
0.6	9.09	9.09	9.09

References 24

1	ZHU Z , NANDI A K . Automatic modulation classification: Principles, algorithms and applications[M]. Hoboken: John Wiley and Sons, 2015.
2	邓兵, 张韫, 李炳荣. 通信对抗原理及应用[M]. 北京: 电子工业出版社, 2017.
3	WEAVER C S, COLE C A, KRUMLAND R B, et al. The automatic classification of modulation types by pattern recognition[R]. Stanford Electronics Laboratories, Technical Report, 1969.
4	WU H C , SAQUIB M , YUN Z F . Novel automatic modulation classification using cumulant features for communications via multipath channels[J]. IEEE Transactions on Wireless Communications, 2008, 7 (8): 3098- 3105. DOI
5	NIE Y N, SHEN X, HUANG S, et al. Automatic modulation classification based multi-ple cumulants and quasi-Newton method for MIMO system[C]//2017 IEEE Wireless Communications and Networking Conference (WCNC), 2017: 1-5.
6	KISHORE T R , RAO K D . Automatic intrapulse modulation classification of advanced LPI radar waveforms[J]. IEEE Transactions on Aerospace and Electronic Systems, 2017, 53 (2): 901- 914. DOI
7	HAN L , GAO F , LI Z , et al. Low complexity automatic modulation classification based on order-statistics[J]. IEEE Transactions on Wireless Communications, 2017, 16 (1): 400- 411. DOI
8	LU X R , HUANG H D , LI S , et al. Salt-body classification method based on U-Net[J]. Chinese Journal of Computational Physics, 2020, 37 (3): 327- 334.
9	ZHOU Y H , LIU H Q , QI P , et al. Forecast of oil production in fractured-vuggy reservoir by using recurrent neural networks[J]. Chinese Journal of Computational Physics, 2018, 35 (6): 668- 674.
10	O'SHEA T J, CORGAN J, CLANCY T C. Convolutional radio modulation recognition networks[C]//International Conference on Engineering Applications of Neural Networks, 2016: 213-226.
11	O'SHEA T J, WEST N. Radio machine learning dataset generation with gnu radio[C]//Proceedings of the GNU Radio Conference, 2016: 213-226.
12	RAJENDRAN S , MEERT W , GIUSTINIANO D , et al. Deep learning models for wireless signal classification with distributed low-cost spectrum sensors[J]. IEEE Transactions on Cognitive Communications and Networking, 2018, 4 (3): 433- 445. DOI
13	HAUSER S C, HEADLEY W C, MICHAELS A J. Signal detectlon effects on deep neural networks utlizing raw IQ for modulation clasfcatlon[C]//IEEE Military Commmunicationa Conference, 2017: 121-127.
14	ALI A , FAN Y Y . Automatic modulation classification using deep learning based on sparse autoencoders with nonnegativity constraints[J]. IEEE Signal Processing Letters, 2017, 24 (11): 1626- 1630. DOI
15	ALI A , FAN Y Y . Unsupervised feature learning and automatic modulation classification using deep learning model[J]. Physical Communication, 2017, 25 (1): 75- 84.
16	任思睿, 黄铭. 基于改进的长短期记忆网络的调制识别算法[J]. 云南大学学报(自然科学版), 2021, 43 (1): 39- 45.
17	张婷婷, 方宇强, 韩蕾. 基于深度残差网络的自动调制识别方法研究[J]. 计算机仿真, 2021, 38 (1): 178- 181.
18	许华, 苟泽中, 蒋磊, 等. 适用于样本分布差异的迁移学习调制识别算法[J]. 华中科技大学学报(自然科学版), 2021, 49 (4): 127- 132.
19	FRENAY B , VERLEYSEN M . Classification in the presence of label noise: A survey[J]. IEEE Transactions on Neural Networks and Learning Systems, 2014, 25 (5): 845- 869. DOI
20	ANGELOVA A, ABU-MOSTAFAM Y, PERONA P. Pruning training sets for learning of object categories[C]//2005 IEEE Computer Society Conference on Computer Visionand Pattern Recognition (CVPR'05), 2005: 494-501.
21	GHOSH A , MANWANI N , SASTRY P S . Making risk minimization tolerant to label noise[J]. Neurocomputing, 2015, 160 (Jul.21): 93- 107.
22	GHOSH A, KUMAR H, SASTRY P S. Robust loss functions under label noise for deepneural networks[C]//Proceedings of the Thirty-First AAAI Conference on Artifcial Intelligence, 2017: 1919-1925.
23	ZHANG Z L, SABUNCU M R. Generalized cross entropy loss for training deep neural networks with noisy labels[C]//Proceedings of the 32nd International Conference on Neural Information Processing Systems, 2018: 8792-8802.
24	O'SHEA T J, WEST N. Deep architectures for modulation recognition[C]//IEEE International Symposium on Dynamic Spectrum Access Networks, 2017: 1-6.