Cyclostationary Feature based Modulation Classification with Convolutional Neural Network in Multipath Fading Channels

Modulation classification has been widely studied in recent years. However, few studies focus on the performance degradation multipath fading channels, whose impact is non-negligible. In this paper, a convolutional neural network (CNN) employing cyclostationary (CS) feature, which maintain essential characteristics proposed for robust modulation classification. Our method can be implemented two approaches, referred as CASE1 and CASE2. CASE1, single-structured CNN designed learning hybrid CS features to perform And CASE2, we present model based hierarchical structure two-stage Specifically, coarse performed by second-order with first-level CNN. Next, another selectively activated learn from high-order fine within subclass. way, our uses provide favorable guidance process of CNN, thus improving channels. The experimental results demonstrate advantages terms accuracy computational complexity.