在对星座图聚类的启发下,构造了一种新的高阶交叉累量.将截获信号的直观几何特征(星座图)映射到变换域中,可区别不同调制内部的同相正交分量的相对关系,避免了聚类算法受噪声干扰和其他随机因素影响.数值化后的星座图特征精度较高,更适合干扰较大的自动智能识别流程.
Inspired by the fuzzy clustering approach of constellation, we present a novel type of feature-high order cross cumulant (HOCC), which map the intuitionistic geometry feature onto the normalized feature in transform domain. The novel feature distinguishes relative distribution between in-phase and quadrature components and lightens the great influence caused by noise and other random factor in clustering algorithm. The numerical feature with high precision is well suited for automatic modulation recognition flow without human operator in great interference.