摘要
针对对平稳随机信号建立参数模型的常用信号处理方法，研究了信号建模的本质，给出了“准确模型”的定义，指出，对给定的平稳随机信号，以其不同阶次统计量建立的“准确模型”的输出信号与原信号并不相等，二者只是在相应的统计特征上相匹配。从这一概念出发，在总结了二阶统计意义下建模的一些主要结论后指出，平稳信号在二阶统计意义下总可以准确建模，证明了在三阶统计意义下并非所有的平稳过程均可准确建模，并给出了可准确建模的必要条件。
Because setting up a parametric model for stationary random signals is a common signal processing method, so it is important to explain explicitly the meaning of modeling, i.e. the relationship between the signal and its model. The concept of “perfect model” is brought forward. According to this concept, it is emphasized that the meaning of modeling is only to approximate certain order statistical property of the original signal but not to reveal its generating mechanism. Subsequently, it is proved that an arbitrary stationary random signal has the perfect model at the second statistical order and not each stationary random signal has a perfect model at third statistical order.
出处
《清华大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
1997年第4期68-71,共4页
Journal of Tsinghua University(Science and Technology)
关键词
平稳随机信号
参数模型
信号处理
平稳过程
stationary random signals
perfect model
high order cumulant
bispectrum
