CNN-based modulation classification in the complicated communication channel
Yongshi Wang, Jie Guo, Hao Li, Li Li, Wang Zhi-gang, Wang Houjun · 2017
Machine Learning are gradually applied in Automatic Modulation Classification (AMC) domain these years. However, in the past researches, the experiment conditions are mostly idealized or simplified. This paper proposes a modified structure of Convolutional Neural Network (CNN), which is proved to have strong capability of classification. The wavelet denoising technology is exploited to restrain the high frequency noise for input signal. As an analysis of the feasibility of this method for complicated channel, we compare the proposed method with the commonly used AMC methods through a large number of signal sequences, taking consider of various distractors. The result shows that the method proposed outperforms on most rugged conditions.