Search Results (66)

Automatic modulation classification (AMC) suffers from performance degradation under low signal-to-noise ratio (SNR) conditions, where modulation characteristics are affected by noise and signals belonging to the same modulation family exhibit similar feature representations. To address these challenges, this paper proposes a Transformer-based spatiotemporal fusion network that jointly exploits local spatial waveform characteristics and temporal dependency information while leveraging the global context modeling capability of the Transformer to integrate complementary multi-dimensional features. The proposed architecture improves feature representation capability under different SNR conditions. Experimental results demonstrate that the proposed method achieves improved classification performance compared with comparative approaches under different SNR conditions. In particular, it achieves an overall classification accuracy of 82.45% over the SNR range from $-10$ to 18 dB, and an average accuracy of 95.55% at SNRs of 2 dB and above, showing improved classification performance in the low-to-medium SNR transition region. Full article

The electromagnetic spectrum grows increasingly crowded with the rapid expansion of mobile, satellite and Internet of Things communications, making intelligent spectrum sensing and efficient management an urgent priority. Automatic modulation classification (AMC) serves as the core of cognitive radio and intelligent communication. Existing deep models often suffer from a large number of parameters and low storage efficiency. To overcome these limitations, we propose MDSCNet, a multi-scale depth-wise separable complex network. Built upon complex depth-wise separable convolution, the network makes full use of the phase information in in-phase and quadrature signals while naturally preserving the symmetric relationship between the in-phase and quadrature components (IQ). The asymmetric multi-scale structure combined with the embedded lightweight attention module jointly forms the overall feature extraction process. The overall parameter count is kept extremely low, at only 47.739 k. Experiments on the RML2016.10a and RML2016.10b datasets show that MDSCNet delivers recognition performance under low signal-to-noise ratios (SNR), reaching 63.42% and 66.71% respectively. More importantly, it outperforms mainstream methods in both parameter count and storage efficiency. Full article

Automatic Modulation Classification (AMC) aims to automatically identify modulation types based on the features of received signals, and acts as a vital technique for spectrum sensing, cognitive radio and electronic countermeasures. However, existing methods generally overlook the issue that modulation signals with similar characteristics are prone to confusion. In particular, under strong interference conditions, feature distributions become blurred and class boundaries tend to overlap, further exacerbating misclassification among modulation types with similar characteristics, thereby limiting improvements in classification accuracy and model robustness. To address this challenge, a modulation classification method based on fuzzy sample feature enhancement is proposed. Specifically, a modulation class entropy constraint and a fuzzy sample feature enhancement constraint are introduced to establish a Multi-scale Fuzzy Sample Feature Enhancement Framework (MTFSFEF). Through fuzzy sample selection and feature representation refinement, the proposed framework effectively mitigates feature overlap among fuzzy samples and enhances inter-class separability and discriminability. For $\mathrm{SNR}\ge 0\mathrm{dB}$, MTFSFEF delivers superior average classification accuracies across all three benchmark datasets: $92.82\%$ on RML2016.10a, over $93.43\%$ on RML2016.10b, and exceeding $92.78\%$ on RML2018.01a, outperforming existing methods by up to $2.68\%$, $2.89\%$, and $2.73\%$, respectively. Full article

Automatic modulation classification (AMC) becomes challenging in multipath fading channels, particularly for high-order quadrature amplitude modulation (QAM) signals whose constellation points are strongly distorted by inter-symbol interference, phase rotation, and fading. This paper presents a channel-type-guided hierarchical AMC framework that combines blind preprocessing with deep learning. In the first stage, the received in-phase and quadrature (IQ) signal is downsampled and preprocessed using blind signal processing techniques. Blind source separation (BSS) is used for additive white Gaussian noise (AWGN) and flat fading channels, whereas the constant modulus algorithm (CMA) followed by BSS is used for multipath fading channels. A convolutional neural network (CNN) then performs first-stage modulation classification and generates a QAM-family flag. If the first-stage output corresponds to a QAM-family signal, a second-stage refinement path is activated. In this path, a convolutional denoising autoencoder (CDAE) is applied to the original received signal to mitigate multipath-induced distortion, followed by BSS preprocessing and a dedicated CNN classifier for 16-QAM, 64-QAM, and 256-QAM. Simulation results over AWGN, flat fading, and multipath Rician fading channels show that the proposed hierarchy improves high-order QAM classification in the considered settings, especially for 64-QAM and 256-QAM under multipath fading with stronger time variation. Multi-frame Softmax averaging further improves decision stability. The results support the use of classical blind preprocessing and selective CDAE-based refinement as a practical, complementary front end for AMC in controlled multipath simulation scenarios, while real over-the-air validation and automatic channel-category detection remain future work. Full article

Automatic modulation classification (AMC) is a critical enabler of cognitive radio (CR) systems, allowing secondary users to identify primary user modulation schemes and adapt transmission parameters in real time. Traditional AMC approaches, based on likelihood functions or hand-crafted features, suffer from degraded performance under low signal-to-noise ratio (SNR) conditions and realistic channel impairments. In this paper, we propose MS-SENet (Multi-Scale Squeeze–Excitation Network), a novel deep-learning architecture that integrates multi-scale convolutional feature extraction, squeeze-and-excitation channel attention, residual learning, bidirectional long short-term memory (BiLSTM) temporal modelling, and global attention pooling into a unified framework for robust AMC. The multi-scale convolution module employs parallel branches with kernel sizes of 3, 5, and 7 to capture both fine-grained phase transitions and coarse envelope patterns from raw in-phase/quadrature (I/Q) signal samples. Squeeze–excitation residual blocks perform channel-wise feature recalibration, enabling the network to emphasize informative feature maps while suppressing less relevant ones. A bidirectional LSTM layer models temporal dependencies across the signal sequence, and a global attention pooling mechanism performs weighted temporal aggregation prior to classification. We present a comprehensive taxonomy of deep-learning architectures for AMC organised along five axes—input representation, feature extraction, temporal modelling, regularization strategy, and architectural complexity—and conduct a rigorous comparative evaluation against ten baseline architectures on a RadioML-style synthetic dataset (110,000 samples, 11 modulation classes, and 20 SNR levels from −20 to +18 dB). The experimental results demonstrate that MS-SENet achieves a mean classification accuracy of 87.9% at SNR ≥ 0 dB (the average of the medium and high SNR regime averages: 86.06% for 0 ≤ SNR < 10 dB and 89.68% for SNR ≥ 10 dB) while maintaining a compact footprint of approximately 406 K parameters, making it suitable for deployment on resource-constrained edge devices. We further analyze the robustness of the proposed architecture to multipath fading, carrier frequency offset, and sample rate offset, confirming its resilience under practical operating conditions. MS-SENet is an architecture designed for automatic modulation classification of I/Q signals and is not related to the homonymous architecture for speech emotion recognition. Full article

Automatic Modulation Classification (AMC) is essential for waveform-level signal characterization. It supports spectrum sensing, signal identification, and adaptive resource allocation in cognitive radio and next-generation wireless systems. However, channel impairments such as multipath propagation, frequency offset, fast fading, and noise degrade modulation signatures, making reliable AMC challenging. Existing deep learning-based approaches often rely on purely data-driven learning, leading to insufficient modeling of modulation-relevant features, loss of transient characteristics, and limited exploitation of hierarchical relationships among modulation types. To address these issues, this paper proposes PHM-Net, a physics-informed hierarchical multi-scale network for robust AMC. The model employs a hierarchical backbone with residual encoder blocks. A Transient Feature Gating (TFG) module enhances modulation-relevant representations, a Cross-Resolution Signal Aggregation (CRSA) module fuses multi-stage features, and a Physics-Informed Hierarchical Loss (PI-HL) enforces consistency between coarse- and fine-grained predictions. Experimental results on three benchmark datasets (RML2016.10a, RML2016.10b, and RML2018.01a) show that PHM-Net consistently achieves the highest average accuracy among all compared models. On RML2018.01a, which contains 1024-sample sequences and 24 classes, PHM-Net achieves an average accuracy of 64.59% and a best-case accuracy of 98.42%, surpassing AMC_Net by 11.14 and 17.09 percentage points and CNN-Transformer by 9.43 and 11.15 percentage points, respectively. PHM-Net provides a robust and interpretable solution for AMC under complex channel conditions. Full article

Automatic modulation classification (AMC) is a core capability for spectrum monitoring, adaptive receivers, and electronic support. Most radio-frequency machine learning (RFML) studies train multi-class classifiers on benchmark datasets that contain a single modulation per recording at baseband. In operational settings, however, the objective is often to detect only a small set of signals of interest, making large multi-class models unnecessarily expensive to train and deploy. In addition, multi-class formulations can increase false-alarm risk due to confusion among non-essential classes and may allocate model capacity inefficiently to distinctions that are irrelevant for the operational objective. This paper investigates an alternative workflow based on targeted binary transformer detectors and evaluates their robustness under practical RF complications. Using the RadioML 2018.01A dataset, we construct binary detection tasks with BPSK as the signal of interest and introduce three increasingly realistic conditions: (i) center-frequency shifts away from baseband, (ii) sampling-rate mismatches via decimation and interpolation, and (iii) multi-signal mixtures where modulations co-occur either in frequency (simultaneous transmissions) or in time (temporal concatenation). The results show that baseband-trained detectors do not generalize to center-frequency-shifted signals, and multi-signal interference can cause complete detection failure unless explicitly modeled during training. We investigate early-exit transformer inference to reduce computation on high-confidence examples, showing it maintains (and occasionally improves) detection performance. We also evaluate inter-modulation transfer learning and intra-modulation adaptation from baseband to mixed- and multi-signal scenarios. Full article

The growing diversity of wireless standards and complex real-world channel effects render automatic modulation classification (AMC) increasingly challenging for spectrum monitoring and edge intelligence. However, most competitive deep-learning-based AMC networks still require ${10}^5$–${10}^6$ parameters, exceeding the memory available on resource-constrained edge platforms. We propose LGP-Net, a lightweight gated-fusion network that pairs a physics-informed expert branch with a compact temporal encoder built from depthwise separable convolution (DSConv), squeeze-and-excitation (SE) attention, and a single-layer gated recurrent unit (GRU). Specifically, unlike other dual-branch structures that directly concatenate the outputs of both pathways, this work designs a lightweight gating unit that requires no external signal-to-noise ratio (SNR) labels and adaptively reweights the two pathways according to signal-quality degradation. With fewer than 40 K parameters, a peak activation footprint of 26.00 KB and an amortised inference latency of 9.7 $\mathsf{\mu }$s per sample under GPU acceleration, LGP-Net attains 65.00% overall accuracy on RadioML 2016.10B (91.48% at 0 dB) and 62.76% on RadioML 2016.10A, placing it in a competitive accuracy–efficiency regime relative to architectures consuming 5× to 500× more parameters. These characteristics support deployment-oriented feasibility under memory-constrained edge settings and high-throughput spectrum-monitoring pipelines. Full article

Automatic modulation classification (AMC) is a key component of spectrum awareness, cognitive radio, and signal intelligence, enabling receivers to identify modulation schemes from noisy in-phase and quadrature (IQ) observations. Traditional approaches rely on likelihood-based methods or handcrafted feature extraction, which often struggle under channel impairments and real-world variability. Recent advances in deep learning enable models to learn directly from multiple signal representations, including raw IQ samples, engineered features, and time–frequency or constellation-based encodings, improving adaptability across diverse signal conditions. This paper presents a structured review of deep learning approaches for AMC, including CNNs, RNN/LSTM models, and transformer-based architectures, with a focus on performance, robustness, and system-level trade-offs. We analyze how representation choices, dataset design, and evaluation protocols influence reported results, and highlight key challenges such as domain shift, low-SNR environments, and multi-signal interference. Finally, we outline future directions focused on improving generalization, integrating classical signal processing with learning-based methods, and enabling efficient deployment in real-world and resource-constrained systems. Full article

To address the growing congestion of the radio frequency (RF) spectrum, Cognitive Radio (CR) systems employ Automatic Modulation Classification (AMC) to dynamically optimize spectrum utilization without introducing protocol overhead. In modern Orthogonal Frequency Division Multiplexing (OFDM) standards, effective AMC requires advanced signal-processing techniques capable of accurately identifying modulation schemes under dynamic channel conditions. Therefore, maintaining robust performance under realistic environments remains a fundamental challenge. This paper evaluates how dataset scale, synthetic impairments, and hardware-induced signal impairments affect the cross-domain generalization of a Convolutional Neural Network (CNN) architecture for OFDM Automatic Modulation Classification (AMC), using 2D amplitude-phase histograms for signal representation. To assess these effects, the CNN is trained on five distinct datasets, encompassing both synthetically generated signals with varying scales and synchronization impairments, as well as a conducted hardware dataset. The cross-domain generalization of the trained models is assessed by evaluating them on a completely unseen indoor Over-The-Air (OTA) dataset collected across 13 distinct positions. Statistical analysis demonstrates that the large-scale synchronization-impaired synthetic dataset achieves the best generalization performance, reaching a mean indoor OTA accuracy of 93.36% and outperforming the limited-size conducted hardware dataset. Overall, this study demonstrates the critical role of data-generation strategies and establishes a robust baseline for achieving reliable cross-domain generalization of CNN-based AMC. Full article

Timely and accurate identification of agricultural pests is imperative for precision crop protection. However, real-world pest recognition faces two critical challenges: the interference of complex field backgrounds, which introduces significant noise, and the severe large intra-class variance caused by pest metamorphosis, which confuses standard classifiers. To address these issues, this paper proposes a coarse-to-fine cascade framework that integrates object localization with fine-grained multi-modal classification. First, we deploy a YOLOv8-based detector to precisely localize and crop pest regions from cluttered environments, effectively eliminating background redundancy. Second, for the cropped targets, we design a fine-grained classification network based on ResNeXt50 integrated with the Convolutional Block Attention Module (CBAM) to extract discriminative features. Crucially, to tackle the challenge of multi-state pest morphologies, we propose a novel Adaptive Multi-Center Classification Head (AMC-Head). Unlike traditional methods that enforce a single feature center for each class, our approach dynamically allocates multiple latent sub-centers for each category, allowing the model to automatically disentangle and cluster distinct morphological representations within a single label. Extensive experiments on the large-scale benchmark dataset IP102 demonstrate that our method achieves an end-to-end accuracy of 91.4%, significantly outperforming single-stage baselines. The proposed framework effectively mitigates the impact of complex backgrounds and metamorphic variation, providing a robust solution for automated pest monitoring. Full article

The emergence of radio frequency machine learning has significantly propelled the application of deep learning (DL) methods in automatic modulation classification (AMC). However, under non-cooperative scenarios, the performance of DL-based AMC suffers severe performance degradation due to scarce labeled samples and noise interference. To enhance noise robustness in few-shot AMC, this paper proposes a complex-domain autoencoder-based method where a complex-valued noise reduction network (CNRN) is embedded into the AMC framework, jointly extracting complex-valued and temporal features from noisy signals to achieve signal–noise separation. Our framework executes four sequential operations: high-signal-to-noise-ratio (high-SNR) samples are first isolated from limited raw data via unsupervised classification; rotation and cyclic time-shifting operations then augment the sample space; the CNRN is subsequently trained on augmented data; and final AMC classification is implemented through DL-based classifiers. Experimental validation on RML 2016.10a dataset demonstrates: (1) for −20 dB signals, denoising achieves 20.18 dB SNR improvement with 87.74% mean squared error reduction; (2) across the −20 dB to 18 dB range, denoised signals exhibit accuracy improvements of 21.57% under DL-based classifiers. Physical validation further confirms that the proposed method exhibits enhanced noise robustness, demonstrating its practical utility in real-world scenarios. Full article

In complex wireless communication environments, automatic modulation classification (AMC) faces two critical challenges: the lack of robustness under low-signal-to-noise ratio (SNR) conditions and the inefficiency of integrating multi-scale feature representations. To address these issues, this paper proposes CAIC-Net, a robust modulation classification network that integrates a dynamic cross-attention mechanism with a cross-SNR contrastive learning strategy. CAIC-Net employs a dual-stream feature extractor composed of ConvLSTM2D and Transformer blocks to capture local temporal dependencies and global contextual relationships, respectively. To enhance fusion effectiveness, we design a Dynamic Cross-Attention Unit (CAU) that enables deep bidirectional interaction between the two branches while incorporating an SNR-aware mechanism to adaptively adjust the fusion strategy under varying channel conditions. In addition, a Cross-SNR Contrastive Learning (CSCL) module is introduced as an auxiliary task, where positive and negative sample pairs are constructed across different SNR levels and optimized using InfoNCE loss. This design significantly strengthens the intrinsic noise-invariant properties of the learned representations. Extensive experiments conducted on two standard datasets demonstrate that CAIC-Net achieves competitive classification performance at moderate-to-high SNRs and exhibits clear advantages in extremely low-SNR scenarios, validating the effectiveness and strong generalization capability of the proposed approach. Full article

Automatic modulation classification (AMC) under low signal-to-noise ratio (SNR) and complex channel conditions remains a significant challenge due to the trade-off between robustness and efficiency. This study proposes a lightweight temporal convolutional network (TCN) and Mamba fusion architecture designed to enhance modulation recognition performance. In the modulation signal denoising stage, a non-local adaptive thresholding denoising module (NATM) is introduced to explicitly improve the effective signal-to-noise ratio. In the parallel feature extraction stage, TCN captures local symbol-level dependencies, while Mamba models long-range temporal relationships. In the output stage, their outputs are integrated through additive layer-wise fusion, which prevents parameter explosion. Experiments were conducted on the RadioML 2016.10A, 2016.10B, and 2018.01A datasets with leakage-controlled partitioning strategies including GroupKFold and Leave-One-SNR-Out cross-validation. The proposed method achieves up to a 3.8 dB gain in the required signal-to-noise ratio at 90 percent accuracy compared with state-of-the-art baselines, while maintaining a substantially lower parameter count and reduced inference latency. The denoising module provides clear robustness improvements under low signal-to-noise ratio conditions, particularly below −8 dB. The results show that the proposed network strikes a balance between accuracy and efficiency, highlighting its application potential in real-time wireless receivers under resource constraints. Full article

To overcome the limitations of existing automatic modulation classification (AMC) methods that mainly target single-signal scenarios in radio-over-fiber (RoF) system, a mixed-signal AMC scheme based on phase noise-insensitive high-order cumulants (PNI-HOC) and distribution characteristics is proposed. The approach enables accurate classification of mixed signals in RoF system. Specifically, a PNI-HOC algorithm is first introduced to mitigate the influence of laser linewidth-induced phase noise. Then, distribution characteristics derived from the signal amplitude histogram are extracted to construct a two-dimensional characteristics space. These characteristics are subsequently fed into decision tree and support vector machine (SVM) classifiers for signal identification. To validate the effectiveness of the scheme, a 10 GBaud RoF system with a 70 km fiber link is implemented. The simulation results show that, compared with the conventional high-order cumulant method, the approach solely based on amplitude histogram distribution characteristics and the scheme based on deep neural networks (DNN) classifier using histogram characteristics, the proposed scheme achieves significantly higher classification accuracy at low optical signal–noise ratios (OSNRs). In particular, when the fiber length is 70 km and the OSNR is ≥16 dB, the classification accuracy of mixed signals is consistently maintained at 100%. Furthermore, the robustness of the proposed method is verified under various system impairments, including laser phase noise, chromatic dispersion and nonlinear effects, amplified spontaneous emission noise, multipath fading, etc., confirming its superior and stable performance. Full article