Communication spectral behavior analysis is critical to the improvement of communication situation awareness and electromagnetic reconnaissance capability in an electromagnetic countermeasure environment. With the development of artificial intelligence technology， communication spectral behavior analysis techniques have been gradually transferred from traditional methods based on feature extraction to intelligent methods based on deep learning technology. However， the insufficient and incomplete spectrum monitoring data in the electromagnetic countermeasure environment will hinder the deep network from feature learning. Moreover， the dynamic battlefield makes it even more challenging for real-time analysis. This paper categorizes the communication spectral behavior analysis technologies into three groups： Frequency behavior analysis， network topology recognition， and communication intention inference from researching objectives in the electromagnetic countermeasure environment. Furthermore， the inner relationship between the three categories is illustrated. Finally， the existing research and development venation are reviewed and prospected considering challenges.