期刊:IEEE International Conference on Electronic Measurement & Instruments日期:2025-08-22卷期号:: 514-518
标识
DOI:10.1109/icemi66537.2025.11306801
摘要
Noise in communication systems is interference factor that constrains radio modulation classification. To address the impact of interference on automatic modulation classification (AMC), we propose a novel AMC method: Feature Density Contrastive Learning (FDCL). The proposed FDCL encodes modulation signal into image. And then, this paper employs two-dimensional kernel density estimation to enhance the image information in image perspective. In this work, contrastive learning is utilized to extract feature, thereby improving the generalization ability of the model. Simulation results shows that the FDCL shows advantage under low signal-to-noise (SNR).