Modulated Signal Open-Set Identification of Complex Convolutional Neural Networks Combined with Incremental Learning
Jinwei Wang, Xiaoqiang Qiao, Jiang Zhang, Chengyuan Sun, Fenghui Liu · 2023
This paper proposes a modulated signal open set recognition model based on CVCNN (Complex Value Convolutional Neural Network) combined with incremental learning to address the limitations of traditional modulated signal open set recognition in distinguishing unknown classes. The study utilized CVCNN as a feature extractor to extract semantic features of samples and identify unknown signals based on the distance center measure between the semantic signals. The findings indicate that the CVCNN model can accurately identify trained samples, reject untrained unknown classes, and distinguish between different unknown classes. Additionally, the complex convolutional neural network proposed in this study is well-suited for feature extraction and recognition of IQ (Inphase and quadrature) signals.