摘要
自动调制识别作为确保通信安全的关键技术之一,有着重要的军用和民用价值。针对低信噪比情况下,综合识别率较低的问题,提出一种基于高阶累积特征,结合稀疏自编码器与特征阈值判决的二阶调制识别模型。零均值高斯白噪声的高阶累积量理论值等于0,因此引入高阶累积量作特征,可使系统免受高斯白噪声的影响。高阶累积量组合得到高阶累积特征,充分利用其所携带的信息。稀疏自编码器作为分类器有利于微弱特征的表征。增加高阶累积特征阈值判决提高了MFSK与MQAM信号的类内识别率。对2ASK、4ASK、2FSK和2PSK等十类调制信号的仿真结果表明,所提算法综合识别效果较对比算法更好,且算法复杂度较低,为高阶累积量与深度学习在调制识别上的结合应用提供了新思路。
As one of the key technologies to ensure the communication security, automatic modulation identification has important military and civilian value. Aiming at the problem of low comprehensive recognition rate under low signal-to-noise ratio, this paper proposes a second-order modulation recognition model based on high-order cumulative feature, sparse self-encoder and feature threshold judgment. The theoretical value of the high-order cumulant of zero-mean gaussian white noise was equal to 0. The high-order cumulants were combined to obtain the high-order cumulants and make full use of the information they carry. As a classifier, sparse self-encoder was beneficial to the characterization of weak features. The addition of higher order cumulative feature threshold decision improves the inclass recognition rate of MFSK and MQAM signals. The simulation experiments of 10 kinds of modulation signals such as 2 ASK, 4 ASK, 2 FSK and 2 PSK show that the proposed algorithm has better comprehensive recognition effect and lower algorithm complexity than the comparison algorithm, which provides a new idea for the combination of high order cumulant and deep learning in modulation recognition.
作者
张秦
龚晓峰
雒瑞森
杜淼
ZHANG Qin;GONG Xiao-feng;LUO Rui-sen;DU Miao(College of Electrical,Sichuan University,Chengdu Sichuan 610065,China)
出处
《计算机仿真》
北大核心
2021年第12期454-459,464,共7页
Computer Simulation
基金
四川省重点研发计划项目(20ZDYF3113)
国家自然科学基金(61876114)
企合作项目(17H1199,19H0355)。
关键词
自动调制识别
低信噪比
高阶累积特征
稀疏自编码器
特征阈值判决
Modulation pattern recognition
Low SNR
High order accumulation characteristics
Sparse auto-encoder
Characteristic threshold decision
