Explore the vast range of titles available.

It is a challenge for automatic modulation recognition (AMR) methods for radiation source signals to work in environments with low signal-to-noise ratios (SNRs). This paper proposes a modulation feature extraction method based on data rearrangement and the 2D fast Fourier transform (FFT) (DR2D), and a DenseNet feature extraction network with early fusion is constructed to recognize the extracted modulation features. First, the input signal is preprocessed by DR2D to obtain three types of joint frequency feature bins with multiple time scales. Second, the feature fusion operation is performed on the inputs of the different layers of the proposed network. Finally, feature recognition is completed in the subsequent layers. The theoretical analysis and simulation results show that DR2D is a fast and robust preprocessing method for extracting the features of modulated radiation source signals with less computational complexity. The proposed DenseNet feature extraction network with early fusion can identify the extracted modulation features with less spatial complexity than other types of convolutional neural networks (CNNs) and performs well in low-SNR environments.

Automatic modulation recognition has regained attention as a critical application for cognitive radio, combining artificial intelligence with physical layer monitoring of wireless transmissions. This paper formalizes signal surface augmentation (SSA), a process that decomposes signals into informative subcomponents to enhance AI-based analysis. We employ Bivariate Empirical Mode Decomposition (BEMD) to break signals into intrinsic modes while addressing challenges like adjacent trends in long sample decompositions and introducing the concept of data overdispersion. Using a modern, publicly available dataset of synthetic modulated signals under realistic conditions, we validate that the presentation of BEMD-derived components improves recognition accuracy by 13% compared to raw IQ inputs. For extended signal lengths, gains reach up to 36%. These results demonstrate the value of signal surface augmentation for improving the robustness of modulation recognition, with potential applications in real-world scenarios.

Deep learning (DL) has brought new perspectives and methods to automatic modulation recognition (AMR), enabling AMR systems to operate more efficiently and reliably in modern wireless communication environments through its powerful feature learning and complex pattern recognition capabilities. However, convolutional neural networks (CNNs) and recurrent neural networks (RNNs), which are used for sequence recognition tasks, face two main challenges, respectively: the ineffective utilization of global information and slow processing speeds due to sequential operations. To address these issues, this paper introduces CTRNet, a novel automatic modulation recognition network that combines a CNN with Transformer. This combination leverages Transformer’s ability to adequately capture the long-distance dependencies between global sequences and its advantages in sequence modeling, along with the CNN’s capability to extract features from local feature regions of signals. During the data preprocessing stage, the original IQ-modulated signals undergo sliding-window processing. By selecting the appropriate window sizes and strides, multiple subsequences are formed, enabling the network to effectively handle complex modulation patterns. In the embedding module, token vectors are designed to integrate information from multiple samples within each window, enhancing the model’s understanding and modeling ability of global information. In the feedforward neural network, a more effective Bilinear layer is employed for processing to capture the higher-order relationship between input features, thereby enhancing the ability of the model to capture complex patterns. Experiments conducted on the RML2016.10A public dataset demonstrate that compared with the existing algorithms, the proposed algorithm not only exhibits significant advantages in terms of parameter efficiency but also achieves higher recognition accuracy under various signal-to-noise ratio (SNR) conditions. In particular, it performs relatively well in terms of accuracy, precision, recall, and F1-score, with clearer classification of higher-order modulations and notable overall accuracy improvement.

Multimodal automatic modulation recognition (MAMR) has emerged as a prominent research area. The effective fusion of features from different modalities is crucial for MAMR tasks. An effective multimodal fusion mechanism should maximize the extraction and integration of complementary information. Recently, fusion methods based on cross-modal attention have shown high performance. However, they overlook the differences in information intensity between different modalities, suffering from quadratic complexity. To this end, we propose an efficient Alignment before Fusion Transformer Network (AbFTNet) based on an in-phase quadrature (I/Q) and Fractional Fourier Transform (FRFT). Specifically, we first align and correlate the feature representations of different single modalities to achieve mutual information maximization. The single modality feature representations are obtained using the self-attention mechanism of the Transformer. Then, we design an efficient cross-modal aggregation promoting (CAP) module. By designing the aggregation center, we integrate two modalities to achieve the adaptive complementary learning of modal features. This operation bridges the gap in information intensity between different modalities, enabling fair interaction. To verify the effectiveness of the proposed methods, we conduct experiments on the RML2016.10a dataset. The experimental results show that multimodal fusion features significantly outperform single-modal features in classification accuracy across different signal-to-noise ratios (SNRs). Compared to other methods, AbFTNet achieves an average accuracy of 64.59%, with a 1.36% improvement over the TLDNN method, reaching the state of the art.

In this paper, we investigate the deep learning applications of radio automatic modulation recognition (AMR) applications in unmanned aerial vehicle (UAV)-to-ground AMR systems. The integration of deep learning in a UAV-aided signal processing terminal can recognize the modulation mode without the provision of parameters. However, the layers used in current models have a small data processing range, and their low noise resistance is another disadvantage. Most importantly, large numbers of parameters and high amounts of computation will burden terminals in the system. We propose a multi-subsampling self-attention (MSSA) network for UAV-to-ground AMR systems, for which we devise a residual dilated module containing ordinary and dilated convolution to expand the data processing range, followed by a self-attention module to improve the classification, even in the presence of noise interference. We subsample the signals to reduce the number of parameters and amount of calculation. We also propose three model sizes, namely large, medium, and small, and the smaller the model, the more suitable it will be for UAV-to-ground AMR systems. We conduct ablation experiments with state-of-the-art and baseline models on the common AMR and radio machine learning (RML) 2018.01a datasets. The proposed method achieves the highest accuracy of 97.00% at a 30 dB signal-to-noise ratio (SNR). The weight file of the small MSSA is only 642 KB.

The automatic modulation recognition (AMR) of optical signals in non-cooperative conditions has recently become one of the most important research projects. The feasibility and effectiveness of using deep learning techniques for automatic modulation format recognition of received wireless optical communication signals is investigated in this framework. Convolutional neural network (CNN) architecture has been developed and refined in recent work to attain performance that outperforms expert-based approaches. An AMR technique based on fan-beam (FB) projection was proposed in this study. The constellation diagram is created by preprocessing the received signal. The constellation diagram is then projected in the FB domain. The FB is used as a tool for extracting features. The (CNN) uses this projected constellation diagram as input to identify different types of modulation formats. For determining the modulation format of an optical signal, three types of classifiers (Alexnet, VGG16, and VGG19) are investigated. For our investigations, we looked at the constellation diagrams of eight common signals, including (2/4/8/16- phase-shift keying (PSK) and 8/16/32/64- quadrature amplitude modulation (QAM). Typical classification accuracy values versus optical signal-to-noise ratio (OSNR) are calculated for evaluation over an OSNR range of (5–30 dB). A study of the effect of increasing the number of samples on the accuracy of the classifiers is also presented for each modulation format. Finally, the phase noise resistance of the (FB) projected diagram is investigated. When compared to traditional modulation format recognition algorithms, simulation results suggest that the proposed algorithm presented in this paper has a greater accuracy rate for signal modulation recognition at low OSNR.

Higher-order cyclic cumulants (CCs) have been widely adopted for automatic modulation recognition (AMR) in cognitive radio. However, the CC-based AMR suffers greatly from the requirement of high-rate sampling. To overcome this limit, we resort to the theory of compressive sensing (CS). By exploiting the sparsity of CCs, recognition features can be extracted from a small amount of compressive measurements via a rough CS reconstruction algorithm. Accordingly, a CS-based AMR scheme is formulated. Simulation results demonstrate the availability and robustness of the proposed approach.

Aiming at the problem of automatic modulation recognition based on deep learning in non-cooperative communication, the migration problem of adversarial sample attacks in both the model and the representation domain is studied. The experimental results show that both the model transfer and the representation domain transfer of the modulation recognition data after the adversarial sample attack will cause the attack effect to decline. The model transfer has a better effect and is more stable than the representation domain transfer. This research provides a new perspective for understanding the robustness and generalization laws of adversarial attacks in non-cooperative communication systems, and lays a theoretical foundation for designing more robust defense mechanisms for deep learning-driven signal processing tasks.

Automatic Modulation Recognition (AMR) in Underwater acoustic (UWA) channels is challenging due to severe noise and strong multipath propagation. The recognition of signals becomes particularly difficult when orthogonal frequency division multiplexing coexists with several recently presented types including orthogonal time frequency space, orthogonal chirp division multiplexing, and orthogonal time sequency multiplexing, because their characteristics are similar. To address these challenges, this study considers the WATERMARK UWA channel under non-Gaussian noise and proposes a lightweight model for AMR, capable of recognizing both conventional and emerging modulation types. The proposed model reorganizes each one-dimensional signal into a matrix, where each row represents a short-time segment of the signal. Then, it captures both local features within segments and dependencies across segments through parallel convolutional branches, each followed by gated recurrent units to integrate information, compressed via a 1×1 convolution, fused by a gated fusion unit, and finally fed into two fully connected layers for modulation recognition. Simulations show that compared with benchmark models, the proposed model achieves better recognition performance while maintaining low computational cost.

Automatic modulation recognition of radio signals is a crucial research direction in non-cooperative communication and satellite communication domains. In recent years, deep learning has achieved remarkable results in the field of automatic modulation recognition. However, adversarial machine learning techniques have introduced threats of adversarial attacks to deep learning-based automatic modulation recognition systems. To enhance the adversarial security of automatic modulation recognition systems, this paper proposes a cross-representation domain ensemble adversarial training defense method. Experiments conducted on the public dataset RML2016.10a demonstrate that by integrating robust models generated through adversarial training in two representation domains, the proposed approach significantly improves the defense capability of automatic modulation recognition models against adversarial example attacks.