A novel multi-feature extraction based automatic modulation classification

doi:10.23919/JSEE.2025.000032

Journal of Systems Engineering and Electronics ›› 2025, Vol. 36 ›› Issue (6): 1407-1427.doi: 10.23919/JSEE.2025.000032

• ELECTRONICS TECHNOLOGY • Previous Articles

A novel multi-feature extraction based automatic modulation classification

Peng SHANG¹^,²(), Lishu GUO¹^,²^,*(), Decai ZOU¹^,²^,³(), Xue WANG⁴(), Shuaihe GAO¹^,²(), Pengfei LIU¹^,²()

¹ National Time Service Center, Chinese Academy of Sciences, Xi’an 710600, China
² Key Laboratory of Time Reference and Applications, Chinese Academy of Sciences, Xi’an 710600, China
³ School of Astronomy and Space Science, University of Chinese Academy of Sciences, Beijing 100049, China
⁴ Institute of Information Sensing, Xidian University, Xi’an 710600, China

Received:2024-07-24 Accepted:2024-07-24 Online:2025-12-18 Published:2026-01-07
Contact: Lishu GUO E-mail:shangpeng@ntsc.ac.cn;guolish@ntsc.ac.cn;zoudecai@ntsc.ac.cn;xuewang@xidian.edu.cn;gaoshuaihe@ntsc.ac.cn;liupengfei@ntsc.ac.cn
About author:
SHANG Peng was born in 1991. He received his Ph.D. degree in communication and information system from the National Time Service Center, University of Chinese Academy of Sciences in 2023. He is currently a assistant researcher in the National Time Service Center. His research interests are machine learning and intelligent signal recognition. E-mail: shangpeng@ntsc.ac.cn

GUO Lishu was born in 1985. She received her Ph.D. degree in mechanical engineering from Harbin Engineering University, in 2012. She is a senior engineer at the National Time Service Center, Chinese Academy of Sciences. She mainly engages in the research of signal intelligence processing and identification technology. E-mail: guolish@ntsc.ac.cn

ZOU Decai was born in 1979. He received his Ph.D. degrees in telecommunication technology from the National Time Service Center, Chinese Academy of Sciences, in 2009. He is a researcher of the National Time Service Center, Chinese Academy of Sciences. His current research interests include the development theory of satellite navigation, cooperative localization in wireless Ad Hoc networks, indoor positioning, and high precision time transfernavigation. E-mail: zoudecai@ntsc.ac.cn

WANG Xue was born in 1978. He received his Ph.D. degrees in telecommunication technology from the National Time Service Center, Chinese Academy of Sciences, in 2011. He is a researcher at the Institute of information sensing, Xidian University. His research interests are including the global navigation satellite system (GNSS) signal quality monitoring and evaluation, the GNSS signal design and verification, GNSS reception technology and anti-jamming and detection and identification for the space signal. E-mail: xuewang@xidian.edu.cn

GAO Shuaihe was born in 1986. He received his B.S. degree in aerospace engineering and automation from Jilin University, in 2008，received his Ph.D. degree from Harbin Engineering University, in 2012. He is senior engineer at the National Time Service Center of the Chinese Academy of Sciences. His main research interests include satellite navigation and time-frequency transmission. E-mail: gaoshuaihe@ntsc.ac.cn

LIU Pengfei was born in 1999. He received his B.E. degree in communication engineering from Liaocheng University, Liaocheng, China, in 2020. He is currently pursuing his Ph.D. degree in the National Time Service Center, Chinese Academy of Sciences, Xi’an, China. His current research interests include satellite signal processing and space target recognition. E-mail: liupengfei@ntsc.ac.cn
Supported by:
This work was supported by the National Natural Science Foundation of China(12273054).

Abstract

Abstract:

Automatic modulation classification(AMC) is an essential technique in both civil and military applications. While deep learning has surpassed traditional methods in accuracy, distinguishing high-order modulations remain challenging. Current efforts prioritize complex network designs, neglecting the integration of deep features and tailored feature engineering to reslove high-order ambiguities. Therefore, a multi-feature extraction framework is proposed, which directly concatenates the deep feature extracted by a newly designed lightweight neural network and the proposed spectrum secondary features or de-noised high-order statistical features. The proposed features and lightweight network both demonstrate superior overall accuracy than other competing features or networks. Furthermore, the effectiveness of the feature extraction framework is also validated. The average classification accuracy on high-order modulation sets reaches 67.39% on the RadioML2018.01A dataset, increasing more than 2% compared with the other competitive networks under the framework. The results indicate the effectiveness of the proposed feature extraction framework for its representational ability by combing the deep features with the proposed domain features.

Key words: automatic modulation classification (AMC), spectrum secondary features, de-noised high-order statistics, lightweight attention network

Peng SHANG, Lishu GUO, Decai ZOU, Xue WANG, Shuaihe GAO, Pengfei LIU. A novel multi-feature extraction based automatic modulation classification[J]. Journal of Systems Engineering and Electronics, 2025, 36(6): 1407-1427.

Figures/Tables 23

Fig 1

Fig 2

Fig 3

Table 1

Fig 4

Fig 5

Fig 6

Fig 7

Fig 8

Fig 9

Fig 10

Table 2

Fig 11

Fig 12

Fig 13

Table 3

Fig 14

Fig 15

Table 4

Fig 16

Fig 17

Table 5

Fig 18

References 40

1	GHASEMZADEH P, BANERJEE S, HEMPEL M, et al. Analysis of distribution test-based and feature-based approaches toward automatic modulation classification. Proc. of the IEEE 30th Annual International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC), 2019. DOI: 10.1109/PIMRC.2019.8904428.
2	ZHOU R Y, HE P F, CAI Z R. An improved automatic modulation classification method based on complex-valued capsule network. Proc. of the Global Reliability and Prognostics and Health Management Conference, 2023. DOI: 10.1109/PHM-Hangzhou58797.2023.10482607.
3	ELSHEBANI M S, ALI Y, AZROUG N, et al. Convolutional neural networks with feature fusion method for automatic modulation classification. Proc. of the 9th International Conference on Computer and Communication Engineering, 2023: 340−345.
4	ZHANG J W, WANG T T, FENG Z X, et al. AMC-Net: an effective network for automatic modulation classification. Proc. of the IEEE International Conference on Acoustics, Speech and Signal Processing, 2023. DOI:10.1109/ICASSP49357.2023.10097070.
5	DEVI P K , KUMAR N N , VARUN M S, et al. Automatic modulation classification using software defined radio system. Proc. of the 3rd International Conference on Innovative Mechanisms for Industry Applications, 2023: 1322−1330.
6	CHEN W H, XIE Z C, MA L, et al A faster maximum likelihood modulation classification in flat fading non-Gaussian channels. IEEE Communications Letters, 2019, 23 (3): 454- 457. doi: 10.1109/LCOMM.2019.2894400
7	ZHU D, MATHEWS V J, DETIENNE D H A likelihood-based algorithm for blind identification of QAM And PSK signals. IEEE Trans. on Wireless Communications, 2018, 17 (5): 3417- 3430. doi: 10.1109/TWC.2018.2811802
8	ZHENG J P, LYU Y F Likelihood-based automatic modulation classification in OFDM with index modulation. IEEE Trans. on Vehicular Technology, 2018, 67 (9): 8192- 8204. doi: 10.1109/TVT.2018.2839735
9	MOSER E, MORAN M K, HILLEN E, et al. Automatic modulation classification via instantaneous features. Proc. of the IEEE Aerospace & Electronics Conference, 2015: 218−223.
10	ZHOU S Y, WU Z L, YIN Z D, et al. Noise-robust feature combination method for modulation classification under fading channels. Proc. of the IEEE 88th Vehicular Technology Conference (VTC-Fall), 2018, DOI:10.1109/VTCFall.2018.8690662.
11	SWAMI A, SADLER B M Hierarchical digital modulation classification using cumulants. IEEE Trans. on Communications, 2000, 48 (3): 416- 429. doi: 10.1109/26.837045
12	HAN L B, XUE H F, GAO F, et al Low complexity automatic modulation classification based on order statistics. IEEE Trans. on Wireless Communications, 2017, 16 (1): 400- 411. doi: 10.1109/TWC.2016.2623716
13	SUETRONG N, TAPARUGSSANAGORN A, PROMSUK N. Deep learning-based robust automatic modulation classification Using higher order cumulant features. Proc. of the 15th International Conference on Information Technology and Electrical Engineering, 2023. DOI: 10.1109/ICITEE59582.2023.10317756.
14	MAJHI S, GUPTA R, XIANG W, et al Hierarchical hypothesis and feature based blind modulation classification for linearly modulated signals. IEEE Trans. on Vehicular Technology, 2017, 66 (12): 11057- 11069. doi: 10.1109/TVT.2017.2727858
15	CAMARA T, LIMA A, LIMA B, et al Automatic modulation classification architectures based on cyclostationary features in impulsive environments. IEEE Access, 2019, 7, 138512- 138527. doi: 10.1109/ACCESS.2019.2943300
16	GAO X C. Modulation recognition algorithm based on spectral feature extraction. Xi’an: XiDian University, 2020.
17	ZHANG D N, LU Y Y, LI Y D, et al High-order convolutional attention networks for automatic modulation classification. IEEE Trans. on Wireless Communications, 2023, 22 (7): 4600- 4610. doi: 10.1109/TWC.2022.3227518
18	HUYNH T T , HUA C H , PHAM Q V , et al. MCNet: an efficient CNN architecture for robust automatic modulation classification. IEEE Communications Letters, 2020, 24(4): 811−815.
19	KUMAR A, SRINIVAS K K, MAJHI S Automatic modulation classification for adaptive OFDM systems using convolutional neural networks with residual learning. IEEE Acess, 2023, 11, 61013- 61024. doi: 10.1109/ACCESS.2023.3286939
20	HUANG S, DAI R, HUANG J, et al Automatic modulation classification using gated recurrent residual network. IEEE Internet of Things Journal, 2020, 7 (8): 7795- 7807. doi: 10.1109/JIOT.2020.2991052
21	KE Z, VIKALO H Real-time radio technology and modulation classification via an LSTM auto-encoder. IEEE Trans. on Wireless Communications, 2022, 21 (1): 370- 382. doi: 10.1109/TWC.2021.3095855
22	CHAUDHARI M S, MAJHI S, JAIN S CNN-Attention-DNN design for CFO estimation of non-Pilot-assisted OFDM system. IEEE Communications Letters: A publication of the IEEE Communications Society, 2023, 27 (2): 551- 555.
23	HU J, SHEN L, SUN G, et al Squeeze-and-excitation networks. IEEE Trans. on Pattern Analysis and Machine Intelligence, 2020, 42 (8): 2011- 2023. doi: 10.1109/TPAMI.2019.2913372
24	TENG C F, CHOU C Y, CHEN C H, et al Accumulated polar feature-based deep learning for efficient and lightweight automatic modulation classification with channel compensation mechanism. IEEE Trans. on Vehicular Technology, 2020, 69 (12): 15472- 15485. doi: 10.1109/TVT.2020.3041843
25	WU H, LI Y X, ZHOU L, et al Convolutional neural network and multi-feature fusion for automatic modulation classification. Electronics Letters, 2019, 55 (16): 895- 897. doi: 10.1049/el.2019.1789
26	CHEN T, GUESZTRIN C. XGboost: a scalable tree boosting system. ACM, 2016: 785−794.
27	YANG Z, YANG D, DYER C, et al. Hierarchical attention networks for document classification. Proc. of the Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, 2016: 1480−1489.
28	DEVELOPS T F. Tensorflow. https://tensorflow.google.cn/tutorials/keras/keras_tuner.
29	JIANG Y Z. Research on high-order modulation recognition based on machine learning. Beijing: Beijing University of Posts and Telecommunications, 2019.
30	ALI A K, ERELEBI E Automatic modulation recognition of DVB-S2X standard-specific with an APSK-based neural network classifier. Measurement, 2019, 151, 1- 13.
31	KHARBECH S, SIMON E P, BELAZI A, et al. Denoising higher-order moments for blind digital modulation identification in mutiple-atenna systems. IEEE Wireless Communication Letters, 2020, 9(6): 765−769.
32	PHUONG T T, OHISHI K, YOKOKURA Y. Deep learning based singular spectrum analysis for realization of wideband force sensing. Proc. of the IEEE International Conference on Mechatronics, 2021. DOI:10.1109/ICM46511.2021.9385603.
33	MONDINO L Y R, CORRAL B G Automatic modulation classification for low-power IoT applications. IEEE Latin America Transactions, 2024, 22 (3): 204- 212. doi: 10.1109/TLA.2024.10431424
34	XIAO C H, YANG S Y, FENG Z X Complex-valued depthwise separable convolutional neural network for automatic modulation classification. IEEE Trans. on Instrumentation and Measurement, 2023, 72, 2522310.
35	CHEN Y T, DONG B H, LIU C T, et al Abandon locality: frame-wise embedding aided transformer for automatic modulation recognition. IEEE Communications Letters, 2023, 27 (1): 327- 331. doi: 10.1109/LCOMM.2022.3213523
36	ZENG R, LU Z L, ZHANG X D, et al Convolutional neural network assisted transformer for automatic modulation recognition under large CFOs and SROs. IEEE Signal Processing Letters, 2024, 31, 741- 745. doi: 10.1109/LSP.2024.3372770
37	ZHANG F X, LUO C B, XU J L, et al An Efficient deep learning model for automatic modulation recognition based on parameter estimation and transformation. IEEE Communications Letters, 2021, 25 (10): 3287- 329. doi: 10.1109/LCOMM.2021.3102656
38	KINGMA D P, BA J. Adam: a method for stochastic optimization. Proc. of the 3rd International Conference for Learning Representations, 2015. https://arxiv.org/abs/1412.6980.
39	RAJENDRAN S, MEERT W, GIUSTINIANO D, et al Deep learning models for wireless signal classification with distributed low-cost spectrum sensors. IEEE Trans. on Cognitive Communications and Networking, 2018, 4 (3): 433- 445. doi: 10.1109/TCCN.2018.2835460
40	XU J L, LUO C B, PARR G, et al A spatiotemporal multi-channel learning framework for automatic modulation recognition. IEEE Wireless Communications Letters, 2020, 9 (10): 1629- 1632. doi: 10.1109/LWC.2020.2999453

Layer	Parameter	Value	Description
Conv1	N_F1	8	Number of filters
Conv1	N_K1	5	The kernel size
Conv2	N_F2	16	Number of filters
Conv2	N_K2	5	The kernel size
Maxpoo11	N_D1	2	Down-sampling factor
Maxpoo11	N_S1	2	The pooling stride
Conv3	N_F3	32	Number of filters
Conv3	N_K3	5	The kernel size
Maxpool2	N_D2	2	Down-sampling factor
Maxpool2	N_S2	2	The pooling stride
Conv4	N_F4	32	Number of filters
Conv4	N_K4	5	The kernel size
LSTM	N_O	32	Output dimension
Attention	−	−	Output from the previous layer
Den1	N_H1	64	Units of hidden layer
Den2	N_H2	10	The modulation types

Classifier	I	Y	F	H	D	O
DT	46.55	47.45	49.77	47.93	50.08	46.33
RF	48.93	47.61	50.17	46.98	49.64	48.96
XGBoost	51.90	48.49	54.68	52.34	55.69	53.20
SVM	51.48	48.26	55.91	53.49	57.58	53.46
DNN	51.86	48.23	54.26	53.61	57.56	52.11

Network	Accuracy/%	Parameter/K	Inference time/ms
CDSCNN	67.27	330	0.28
FEAT	70.79	177	0.20
PETCGDNN	69.39	74	0.18
TransGroupNet	67.59	583	0.25
AttCNN	70.69	21	0.08

Network	Feature set
Network	Z+D	Z+V
CDSCNN	69.19	68.41
TransGroupNet	69.20	69.24
FEAT	72.58	72..47
PETCGDNN	71.58	70.91
AttCNN	72.95	72..66

SNR/dB	Feature set
SNR/dB	CDSCNN+D	TransGroupNet+D	FEAT+D	PETCGDNN+D	AttCNN+D
−2	37.30	32.36.	46.37	46.17	47.98
0	59.71	60.54	68..80	70.14	75..00
2	81.00	83.23	85.26	83.64	89.23
4	89.42	91.83	94.03	92.64	96.47

A novel multi-feature extraction based automatic modulation classification

RichHTML

PDF (PC)

Knowledge

Abstract

Cite this article

Share this article

Figures/Tables 23

References 40

Related Articles 1

Recommended Articles

Metrics

Comments