Multi-Level Wavelet Transform Network Based Automatic Modulation Classification

Automatic Modulation Classification (AMC) is a crucial technology in wireless communication for spectrum management, signal monitoring, and control. Deep learning neural networks have been widely applied in AMC. However, in low signal-to-noise ratio (SNR) environments, noise severely degrades the effective features extracted by neural networks, leading to suboptimal overall modulation classification performance. To address this challenge, this paper proposes a modulation classification algorithm based on a multi-level wavelet transform network. The algorithm employs multi-level wavelet transforms to decompose signals into multi-scale components, separating them into low-frequency and high-frequency parts. Since the low-frequency components typically contain the primary signal information while the high-frequency parts include noise and details, removing the high-frequency components effectively suppresses noise. Experiment results demonstrate that the proposed algorithm achieves a maximum modulation recognition accuracy of 94.4%, and outperforming other state-of-the-art baseline network models under low-SNR conditions.