Abstract
Spectrum awareness gained attention to solve the emerging communication problems, e.g., overcoming spectrum shortage and performing precise interference management. Phase noise and fading are well-known concerns in wireless communication, which could result from the channel or thermal noise that worsens the communication systems. Deep learning showed outstanding results in solving communication system problems compared to traditional methods. Our work focuses on automatic modulation classification with different phase noise levels in the fading channel using constellation diagrams as input. Our neural network demonstrates its capability to capture a different modulation format under different phase noise levels with excellent classification accuracy.