基于softmax回归的通信信号循环谱的多分类识别方法
2018-01-31刘亚冲唐智灵
刘亚冲+唐智灵
摘 要: 通信信号调制方式的自动识别在通信对抗领域中具有重要作用,同时也是未来认知无线电系统的重要组成部分,如何在日趋密集的信号环境中快速准确地识别多个混合通信信号是实现通信信号调制方式自动识别的重点。针对这种情况,以数字通信信号的循环谱为特征,通过构建softmax回归多分类识别器,提出一种基于softmax回归的通信信号循环谱的多分类识别方法。通过计算机验证不同条件下的算法性能,证明了该方法无需知道典型的数字调制信号(如ASK,BPSK,QPSK,16QAM,64QAM)的符号率、载频以及同步定时等先验信息,对它们组成的混合信号可以正确识别其中包含的每个调制信号的调制方式,并且识别速度较快。
关键词: softmax; 多分类识别; 循环谱; 调制方式识别; 神经网络; 电子对抗
中图分类号: TN911.6?34 文献标识码: A 文章编号: 1004?373X(2018)03?0001?05
Abstract: The automatic identification of communication signal modulation mode has important application in the field of communication countermeasures, and is an important component of the future cognitive radio system, so how to quickly and accurately recognize the multiple mixed communication signals in the increasingly?intensive signal environment is the key to realize the automatic identification of communication signal modulation mode. By taking the cyclic spectrum of the digital communication signals as the feature, and building the softmax regression multi?classification recognizer, a ssoftmax regression based multi?classification recognition method of communication signal cyclic spectrum is put forward. The algorithm performance was verified with computer under different conditions, which proves that the method needn′t know the symbol rate, carrier frequency, synchronization timing and priori information of the typical digital modulation signals (such as ASK, BPSK, QPSK, 16 QAM and 64 QAM). The mixed signals can identify the modulation mode including each modulation signal correctly, and has fast identification speed.
Keywords: softmax; multi?classification recognition; cyclic spectrum; modulation mode recognition; neural network; electronic countermeasure
0 引 言
随着通信技术的快速发展,调制方式也在不断变化和发展,使得电磁环境日趋复杂。通信接收方想对接收的信号解调,必须先知道该信号的调制类型和信号参数,调制方式的自动识别技术成为认知无线电的必备功能之一[1],具有很强的发展潜力和研究价值[2?3]。
目前调制方式的自动识别方法主要是将小波变换技术[4?5]、信号波形统计特征[6?7]、高阶累积量[8?9]、循环谱技术[10?12]和调制识别技术相结合,形成新的调制方法。但现有方法都还有各自的不足,例如,基于小波变换的方法只能对不连续的相位或频率调制进行识别,同时选择适合分析的小波母波也很困难;……
