A Novel Automatic Modulation Classification Method Using Attention Mechanism and Hybrid Parallel Neural Network

Automatic Modulation Classification (AMC) is of paramount importance in wireless communication systems. Existing methods usually adopt a single category neural network or stack different categories networks series, and rarely extract types features simultaneously proper way. When it comes to the output layer, softmax function applied for classification expand inter-class distance. In this paper, we propose hybrid parallel AMC problem. Our proposed method designs structure which utilizes Convolution Neural Network (CNN) Gate Rate Unit (GRU) spatial temporal respectively. Instead superposing these two directly, three attention mechanisms are assign weights features. Finally, cosine similarity metric named Additive Margin function, can distance compress intra-class simultaneously, adopted output. Simulation results demonstrate that achieve remarkable performance on an open access dataset.