| 摘要: |
| 在航天发射活动中，高效快速地检测干扰信号是保障电磁环境安全的关键环节。针对航天发射场内干扰信号检测效率低、可靠性差的问题，提出了一种新颖的干扰信号检测方法，该方法结合了对抗性自编码器(Adversarial Autoencoder，AAE)与时频注意力机制(Time Frequency Attention Mechanism，TFAM)。首先，通过编码器与时频注意力机制，提取输入频谱数据的潜在特征、时间特征、频率特征，并通过对抗性的训练，使用鉴别器引导潜在特征分布在特定特征空间；其次，通过解码器利用三重特征进行频谱重构；最后，基于重构结构与输入数据之间的均方误差检测干扰信号。通过与经典异常检测算法的对比，所提的方法具备更优越的检测性能。 |
| 关键词: 对抗性自编码器 时频注意力机制 干扰信号检测 |
| DOI： |
|
| 基金项目:国家自然科学基金（62371463） |
|
| Research on Interference Signal Detection Methods for Aerospace Launch Sites |
| LI Wenhan,JI Haipeng,YU Peng |
| (1.College of Electronic Science, National University of Defence Technology, Changsha 410073, China;2.The 63rd Research Institute of National University of Defense Technology, Nanjing 210007, China;3.Xichang Satellite Launch Center,Wenchang 571300, China) |
| Abstract: |
| In aerospace launch activities, detecting interference signals efficiently and quickly is a key aspect of ensuring the safety of the electromagnetic environment in critical areas. In response to the problems of low efficiency and poor reliability in detecting interfering signals in space launch sites, this paper proposes a novel interference signal detection method. This method combines the Adversarial Autoencoder (AAE) with the Time Frequency Attention Mechanism (TFAM). First, the encoder and TFAM extract latent features, time features, and frequency features from the input spectrum data. Through an adversarial training process, a discriminator guides the latent feature distribution within a specific feature space. Second, the decoder reconstructs the spectrum using the triple features. Finally, the mean squared error between the reconstructed structure and the input data is used to detect interference signals. Simulation results show that the proposed method outperforms existing interference detection methods. |
| Key words: Adversarial autoencoder Time frequency attention mechanism Interference signal detection |