DTV radio spectrum anomaly detection based on an improved GAN
DTV radio spectrum anomaly detection based on an improved GAN
复制标题
DOI：
10.23919/ursigass49373.2020.9232396
复制
发表时间：
2020-08
期刊：
影响因子：
--
通讯作者：
Liu Zhang;C. Huang;Hao Tang;J. Yang;Ming Huang
中科院分区：
文献类型：
--
作者：
Liu Zhang;C. Huang;Hao Tang;J. Yang;Ming Huang
由于无线应用的普及，检测无线电频谱中的异常行为是一项紧迫的任务。无线电频谱异常包括授权频段和非授权频段的异常。本文主要研究数字电视（DTV）频段（474 MHz）无线电频谱的无监督异常检测。首先，提出了一种改进的GAN模型，并在MNIST数据集上证明了其有效性。然后，通过采集DTV频段的频谱数据集对模型进行训练。通过模拟不同的干扰信号，验证了模型的性能。结果表明，当干扰脉冲的宽度和功率为1.5MHz，比信号大5dB时，检测精度达到97.86%。对功率为-15 dBm的真实的FM干扰信号，检测率接近90%。
Detecting anomalous behavior in radio spectrum is an urgent task due to the popularity of wireless applications. Radio spectrum anomalies include both the anomalies in authorized and unauthorized frequency band. In this paper, we focus on unsupervised radio spectrum Anomaly detection of Digital Television (DTV) frequency band (474MHz). Firstly, an improved GAN model was proposed and its efficacy was demonstrated on a MNIST dataset. Then, the model is trained by collecting the spectrum data sets at DTV band. The performance of the model was tested by simulating different interference signals. Results show that when the width and power of the interference pulse is 1.5MHz and 5dB greater than the signal, the detection accuracy achieves 97.86%. Besides, for a real FM interference signal with power of -15dBm, the detection rate is nearly 90%.