18 May 2023 in IEEE Transactions on Aerospace and Electronic Systems
Automatic modulation recognition plays an important role in military and civilian applications, identifying the modulation format of received signals before signal demodulation. With the increasing complexity and density of the electromagnetic environment, the multi-component modulation radar signal recognition against various signal-to-noise ratio (SNR) conditions has become a practical and urgent problem. In this paper, we propose a dual-component modulation recognition framework, which incorporates the residual Swin transformer denoise network (ResSwinT), Swin transformer feature extraction network (SwinT), residual-attention (RA) modulation recognition head, and SNR level classifier and achieves robust recognition performance against various SNR conditions with tolerable complexity and accuracy trade-off. Firstly, the time-frequency analysis is employed to transform dual-component radar signals into time-frequency images (TFIs). Then, the TFIs at various SNR levels are applied to the SwinT, which generates shallow and deep feature representations for the SNR classifier and RA-modulation recognition head, respectively. The ResSwinT is initiated to reconstruct low SNR TFIs only, which are again processed by the SwinT. Finally, the RA-modulation recognition head provides modulation format predictions. The proposed framework can identify randomly combined dual-component radar signals from 12 modulation formats, meanwhile, improving the utilization of the SwinT feature and reducing unnecessary computation of the ResSwinT. Simulation results show that the proposed scheme can obtain an exact match ratio (EMR) of larger than $\bm {97\%}$ at SNR $\bm {>\!-6}$ dB. At low SNR condition ( $\bm {-12}$ dB), the ResSwinT can obtain about EMR gain of $\bm {20\%}$ and the overall framework can achieve EMR of more than $\bm {80\%}$ , which outperforms other state-of-the-art methods and obtains better generalization capability.
29 April 2024 in IEEE Wireless Communications Letters
Orthogonal frequency-division multiplexing (OFDM) has gained increasing attention for the automatic modulation classification (AMC) tasks in multiple-input multiple-output OFDM (MIMO-OFDM) systems. This letter proposes a MIMO-OFDM modulation classification network called 4D2DConvNet for MIMO-OFDM systems, which integrates multi-branch shallow two-dimensional convolution (2DConv) with four-dimensional convolution (4DConv) to extract channel-specific OFDM symbol features and cross-channel correlation. Simulation results demonstrate that the 4D2DConvNet achieves robust performance, which attains a classification accuracy exceeding 95% at a signal-to-noise ratio (SNR) of 8 dB over fifth-generation (5G) frequency-selective fading channels across various delay profiles and Doppler shift configurations.
22 January 2024 in IEEE Transactions on Vehicular Technology
Automatic modulation classification (AMC) plays a vital role in military and civilian applications, which identifies the modulation schemes of received signals before signal demodulation. Orthogonal frequency division multiplexing (OFDM) technology against the frequency-selective fading channel is widely applied in fourth generation (4G) and fifth generation (5G) communication systems, which may suffer from various hardware impairments, including in-phase and quadrature (IQ) imbalance and carrier frequency offset (CFO). This paper proposes an OFDM modulation classification framework, which incorporates the initial blind IQ imbalance and CFO compensation modules and the cross-selective kernel network (Cross-SKNet). Firstly, a small segment of OFDM signals is utilized for consecutive IQ imbalance and CFO compensation modules to estimate compensation coefficients and reconstruct the received burst OFDM symbols. Subsequently, the Cross-SKNet is initiated using the reconstructed signals. The network comprises the consecutive Cross-Down sampling (Cross-DS) module, the Cross-selective kernel (Cross-SK) module, and the channel and spatial attention classifier, which integrate asymmetric convolution to extract intra-symbol and inter-symbol correlation with reduced computational complexity and leverage selective kernel technology to adaptively learn the receptive field. The proposed framework can identify four different OFDM modulation formats and enhance performance by IQ imbalance and CFO compensation with a fraction of additional computation. Simulation results demonstrate that the proposed Cross-SKNet outperforms the state-of-the-art method across a range of signal-to-noise ratio (SNR) levels and exhibits superior generalization capabilities at low SNR conditions. Based on the simulation results of the generated dataset, the Cross-SKNet achieves a classification accuracy of over $\bm {95\%}$ at SNR > $\bm {8}$ dB. Moreover, the incorporation of IQ imbalance and CFO compensation modules enhances the performance of the Cross-SKNet, which can obtain an accuracy gain of $\bm {8\%}$ at low SNR condition of $\bm {-10}$ dB.
20 February 2025 in IEEE Transactions on Vehicular Technology
Orthogonal frequency division multiplexing (OFDM) technology, robust against frequency-selective fading channels, is widely applied in the fifth-generation (5G) communication systems, which makes automatic modulation classification (AMC) tasks critical for OFDM communication systems in military and civilian applications. This paper proposes a self-supervised learning framework, named cross-view prediction with automatic augmentation (CPAA). Unlike existing supervised methods requiring substantial labeled data, CPAA leverages massive unlabeled samples and minimal labeled data. Initially, an automatic data augmentation module is incorporated with traditional augmentation methods. Subsequently, the unlabeled pairwise signals are utilized for self-supervised pertaining through the cross-view prediction (CP) module. Finally, supervised fine-tuning using well-trained encoder is conducted on a small amount of labeled datasets. The proposed framework is capable of identifying five different OFDM modulation formats and enhancing performance through automatic data augmentation, which effectively retains data identity and generates diverse views. Simulation results indicate that CPAA can achieve robust classification performance with massive unlabeled samples and extremely limited labeled samples. Moreover, the proposed framework outperforms state-of-the-art methods across a range of signal-to-noise ratio (SNR) levels and demonstrates better generalization capabilities under low SNR conditions. Based on the simulation results, the CPAA framework achieves a classification accuracy of over $\bm {85\%}$ at SNR $\bm {>}$ 8dB with only one labeled sample per SNR for each modulation scheme. Moreover, the automatic data augmentation module enhances the performance of self-supervised learning when limited labeled data are available, yielding an approximate average performance gain of $\bm {48.78\%}$ and demonstrating good transferability on the existing self-supervised learning algorithms.
28 November 2025 in IEEE Transactions on Vehicular Technology
Orthogonal Frequency Division Multiplexing (OFDM) has become a fundamental implementation technology for fifth-generation (5G) communication systems. The OFDM modulation classification tasks are crucial for ensuring the reliable performance of OFDM-based communications in military and civilian applications. In this paper, we propose a temporal-spatial convolutional transformer (TSConvT), which utilizes asymmetric convolutional (AsyConv) projections and separated temporal and spatial attention mechanisms to enhance the intra-symbol and inter-symbol feature extraction. Based on the simulation results, the TSConvT network demonstrates superior classification performance compared to state-of-the-art methods for OFDM modulation classification, which achieves an average test accuracy of 84.05% across signal-to-noise ratio (SNR) levels ranging from -10dB. to 20dbFurthermore, the proposed model demonstrates effective generalization capability under varying 5G communication channel scenarios.