论文部分内容阅读
本文提出在宽带相关噪声中,提取窄带信号谱分量的一种高分辨率谱分析方法.基于宽带噪声的相关半径远小于窄带信号分量的相关半径,在输入时间过程的相关函数的大延时部份,宽带噪声的影响已相当弱.因而可利用测量到的大延时部份的相关函数来估计窄带信号的谱分量.本文首先用复指数近似方法,将大延时相关函数内插,得到窄带信号分量的小延时部份的相关函数的估计值.利用大延时相关函数的测量值,加上小延时相关函数的估计值,构成完整的相关函数。其次,按最大熵方法外推相关函数,得到高阶的最大熵预测系数,由此得到窄带信号分量的频率、功率和带宽的精确估计.为了达到高分辨率而采用高阶相关矩阵,谱估计的运算量十分大.本文提出节省运算量的有效方法,可避免同时解出全部谱分量,而仅选出功率较大的谱分量来对其频率作精确估计.最后,以宽带相关噪声中的多个正弦信号为例,给出了线谱分量的各参量估计的计算机模拟结果.本文提出方法的谱分辨率较常规谱分析法的谱分辨率提高很多倍.
In this paper, we present a high-resolution spectral analysis method for extracting the spectral components of narrow-band signals in broadband-related noise.Based on the correlation radius of wideband noise, which is much smaller than the correlation radius of narrowband signal components, The impact of broadband noise is quite weak, so the correlation function of the measured large delay can be used to estimate the spectral components of the narrow-band signal.Firstly, the complex exponential approximation method is used to interpolate the large delay correlation function to obtain The estimated value of the correlation function for the small delay component of the narrow-band signal component, using the measurement of the large delay correlation function plus the estimate of the small delay correlation function, constitutes the complete correlation function. Secondly, the maximum entropy prediction coefficient is obtained by extrapolating the correlation function according to the maximum entropy method, and the accurate estimation of the frequency, power and bandwidth of the narrowband signal component is obtained. In order to achieve high resolution, the high-order correlation matrix The computational complexity is very large.This paper presents an effective method to save the computational complexity, which can avoid the simultaneous solution of all the spectral components, and only select the larger spectral components to make accurate estimates of their frequencies.Finally, As an example, a number of sinusoidal signals are given, and computer simulation results of each component of the line spectral components are given.The spectral resolution of the proposed method is many times higher than that of the conventional spectral analysis.