AMC-Transformer: Enhanced Automatic Modulation Classification based on Attention Model

Wireless signal recognition plays a critical role in both military and civilian domains. Developing high-precision Automatic Modulation Classification (AMC) algorithms is essential for effective spectrum management in today's complex wireless communication environment. Deep learning-based AMC techniques have shown promising results in extracting effective classification rules from complex modulation signals. In this work, we propose a novel transformer-based model for AMC that leverages an attention mechanism to enhance recognition accuracy. Our model divides In-phase and Quadrature (IQ) signal sequences into fixed-length patches and utilizes multiple transformer layers to capture complex temporal and spatial feature correlations. Extensive experiments have been conducted to demonstrate that our proposed model outperforms the typical CNN and ResNet models, achieving a 25% and 15% improvement in recognition accuracy at a signal-to-noise ratio of 10 dB, respectively. The proposed attention-based transformer model provides a powerful and efficient approach for AMC in complex wireless communication environments.