摘要
在综合分析五种离散频谱校正方法特点的基础上，通过仿真研究，得出在无噪声情况下比值校正法的校正精度高于其它几种方法：在小噪声情况下，各种校正方法的校正精度大致相当；在大噪情况下，第二种相位差校正法（构造新序列，即将原时域序列前Ｎ／２点个移Ｎ／４点，将序列的前后Ｎ／４点置零，分别对原序列和新序列进行ＦＦＴ分析，利用对应峰值谱线的相位差进行频谱校正的通用方法）的校正精度要高于其它几种方法。对于相隔较近的频率成分（不小于４个频率分辨率），两种相位差校正法的校正精度要高于其它几种方法。在小噪声的工程实际中，推荐简单方便的加 Ｈａｎｎｉｎｇ 窗的能量重心法（ｎ＝１或 ｎ＝２）进行校正，幅值误差小于 １％，频率误差小于０.０２个频率分辨率，相位误差小于５度。
On the basis of synthetically analyzing the characteristic of five correcting methods on discrete spectrum, simulation study indicates: When there is no noise the precision of ratio correcting method is higher than that of any other method. When the signal contains small noise, all the methods' precision almost is same. Under the condition of big noise, the precision of the second Phase Difference Correction (constructing a new sequence by translating the former N/2 points of original time-domain sequence at the scale of N/4 points, setting both the former and latter N/4 points to zero, performing FFT analysis on the original sequence and new one, the spectrum can be corrected by taking advantage of the phase difference of two corresponding peak spectral lines.) is higher than others'. As to the signals whose frequency intervals are small (no less than 4 resolutions), the precision of two Phase Difference Correcting methods is higher than others. In engineering application, we recommend to correct the spectrum containing small noise by Energy Centrobaric Correcting Method with Hanning window (n=1 or n=2), which is simple and convenient. The errors of amplitude, frequency and phase are under 1% 0.011 resolution and 5 degree respectively.
出处
《信号处理》
CSCD
2001年第1期91-97,共7页
Journal of Signal Processing
