Research on Intelligent Recognition of Communication Signal Modulation Patterns Based on Convolutional Neural Network

This paper presents a novel deep learning framework designed to automatically recognize digital modulation schemes, even under adverse low signal-to-noise conditions. The proposed C-BILSTM-A model amalgamates convolutional neural networks for extracting spatial features with bidirectional long short-term memory networks to capture temporal dependencies, while an integrated attention mechanism selectively emphasizes the most informative segments of the signal. Digital signal datasets encompassing 2ASK, 4ASK, 4FSK, BPSK, QPSK, 8PSK, and 64QAM were synthesized through realistic modulation and noise simulation techniques, ensuring a rigorous evaluation environment. Comparative analyses against conventional CNN, BiLSTM, support vector machines, random forests, and k-nearest neighbors reveal that the C-BILSTM-A approach achieves superior recognition accuracies, particularly in challenging scenarios characterized by low SNR. The architecture's ability to autonomously distill and dynamically weight critical features marks a significant stride in modulation recognition technology, offering a robust and scalable solution for modern wireless communication systems.