SNR-Aware Adaptive Feature Fusion for Robust Automatic Modulation Classification in Low-SNR Environments

Automatic modulation classification (AMC) in lowSNR environments remains challenging because single-domain features are easily degraded by noise and channel uncertainty. To address this issue, this paper proposes an SNR-adaptive feature fusion framework that combines frequency-domain spectral descriptors with higher-order statistical features and dynamically adjusts their relative contributions according to the estimated SNR. The fused representation is then classified using a support vector machine (SVM) with a radial basis function kernel. Experiments on six digital modulation types over a wide SNR range show that the proposed method consistently improves over single-domain and conventional FFT-based baselines, especially in the low-SNR regime, while retaining moderate computational cost. These results indicate that the proposed framework provides a practical and robust AMC solution for cognitive radio systems.