Deep Learning and Rayleigh Channel Based Signal Modulation Pattern Recognition System

This paper presents a novel deep learning framework for signal modulation recognition in wireless communication systems, addressing the challenge of accurately identifying diverse modulation types under varying signal-to-noise ratios (SNRs). Modulation recognition is essential for optimizing communication efficiency and reliability, but existing methods often struggle with complex signal environments and low SNRs. To overcome these limitations, a hybrid model combining convolutional neural networks (CNNs) and Gated Recurrent Units (GRUs) is proposed. The framework first applies circular shifting to the channel data to enhance robustness against noise and interference, ensuring that the input data is invariant to cyclic shifts. It then uses CNNs to extract hierarchical features from the signal, capturing spatial characteristics, followed by GRUs to model temporal dependencies, which are crucial for distinguishing between different modulation types. Experiments demonstrate significant improvements in recognition accuracy across multiple modulation types, including BPSK, QPSK, 8PSK, 16QAM, and 64QAM, under various SNRs compared to existing methods. The integration of CNNs and GRUs effectively captures both spatial and temporal features, highlighting the innovation of our approach. This framework provides a robust solution for modulation recognition in modern wireless communication systems, paving the way for more efficient and reliable signal processing.