Electromagnetic signal deep learning robust classification method based on complementary adversarial training

• Main Authors: REN JINYANG, QI PEIHAN, ZHOU XIAOYU, ZHANG WEILIN, LI ZAN, JIANG TAO, MENG YONGCHAO, LIANG LINLIN, YIN KAI, LIU MALIANG
• Format: Patent
• Language: Chinese; English
• Published: 28.05.2024
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; ELECTRIC DIGITAL DATA PROCESSING; PHYSICS

The invention provides an electromagnetic signal deep learning robust classification method based on complementary adversarial training. The method comprises the following steps: acquiring an open source modulation signal data set; performing normalization preprocessing on the open source modulation signal data set to obtain a training set and a test set; processing the training set based on an adversarial attack method and a disturbance sample generation algorithm to obtain an adversarial sample set and a disturbance sample set; combining the confrontation sample set, the disturbance sample set and the training set to obtain a new training set; training based on a complementary adversarial training algorithm and the new training set to obtain a target classifier; and verifying the performance of the target classifier through the test set. According to the electromagnetic signal deep learning robust classification method based on complementary adversarial training, the robustness of the model is improved with