经典谱估计在短突发信号频谱细化中的特性分析

为了比较经典谱估计中的栅栏效应和截断效应对谱估计的影响,利用补零FFT、移位DFT、线性调频Z变换（CZT）和Zoom-FFT等经典谱估计算法,利用Matlab进行了性能仿真,最后得出了它们在短突发信号频谱细化分析中存在的极限问题。尤其是对含有多频率分量的突发信号,当相邻频率间隔小且信号长度短时,即使采用很大的细分倍数也无法区分两个相邻的频率。为减少截断效应带来的能量泄漏和谱间干扰,对有限长数据需要利用现代谱估计算法进行频谱细化分析。实验结果验证了理论推导的正确性。

Analysis of classical spectral estimation used in spectral zooming of short burst signals

To compare the impacts of barrier effect and truncation effect existing in classical spectral estimation on spectral estimation,the performance of such classical spectral estimation algorithms as zero padding FFT,shifting DFT,CZT and Zoom-FFT were studied and simulated by Matlab.Finally,the paper analyzed the limit of these algorithms in spectral zooming of short burst signals.Especially in the spectral zooming of multi-frequency-component signals,even the large subdivision multiples can not differentiate two adjacent frequencies,when the isolation between frequencies is small and the length of the signal is short.To reduce energy leakage and spectral interference caused by truncation effect,a spectral zooming estimation for finite length data is given by means of modern spectral estimation algorithm.The experimental results proved the correctness of the theoretical inference.