MICNet: A Hybrid Model Based on Inception Network and CNN for Automatic Modulation Classification
Ming Liang Xue, Ming Huang, Jing Jing Yang, Jun Chang Chen · 2021 International Conference on Wireless Communications and Smart Grid (ICWCSG) · 2021
In recent years, the increasingly complex electromagnetic environment requires more accurate and rapid modulation methods. With the continuous development and application of deep learning, more and more methods and structures are proposed to solve the problem. In this letter, we propose an efficient architecture for an automatic modulation classification (AMC) based on inception network and CNN. We combine the improved inception modules with convolutional neural network (CNN) to obtain a mixed network structure (MICNet), which has fewer parameters and higher accuracy. In the experiments, MICNet reaches the overall 24-modulation classification rate of 97.54% at 28 dB SNR on the well-known DeepSig dataset. In addition, MICNet can be used for classification modulation recognition on computers with less hardware configuration.