低频成分的频谱校正  

Correction of frequency spectrum for low frequency components

作　　者：陈奎孚[1] 王建立[2] 张森文[3] 

机构地区：[1]中国农业大学理学院,北京100083 [2]邢台职业技术学院,河北邢台054035 [3]暨南大学应用力学研究所,广东广州510632

出　　处：《振动工程学报》2008年第1期38-42,共5页Journal of Vibration Engineering

摘　　要：目前已经提出了多种频谱校正方法,但是将它们应用于低频成分时,其精度显著下降,这是由于模型忽略了实信号所包含的负频率的贡献。给出了一种考虑负频率贡献的显式校正新方法,它利用局部谱峰附近的3条谱线。采用351×180个仿真样本对提出的方法进行了考核。不同仿真样本所含低频成分的周期数(CiR)从1变化到8,步长为0.02;相位从0°变化到179°,步长为1°。考察了每个CiR的180个不同相位算例的最大误差。结果表明:1)就所考查的参数范围,校正频率的误差小于0.000 6Δω(Δω为快速傅立叶频率分辨率);2)幅值和相位误差上限不超过0.3%和0.3°;3)误差的大体趋势随CiR增加而下降,精度最高的条件仍然是整周期采样。Several spectrum correction approaches have been developed, but their improvements deteriorate significantly when they are applied to the corr（ection of the low frequency component（LFC）.It is due to that the parameter model underlying the conventional approaches neglects the influence of the negative frequency in the real signal. A new approach is proposed to obtain the LFC parameters from fast Fourier transform FFT） spectrum lines, which can take into account the contribution from the negative frequency. This explicit approach works only for the rectangular window, and makes use of three spectrum lines around the spectrum main lobe. The proposed approach has been validated by 351×180 simulation signal samples. The parameters in these simulated samples distribute widespread, The oscillating cycles covered in record （CiR） range from 1 to 8 by 0.02, and each CiR case included series samples with the initial phases increased from 0^o to 179^o by 1^o.The maximum error by the proposed approach among 180 phase samples for each CiR is inspected. The simulation results show that so far as the aforementioned widespread parameters range is concerned, the frequency error of the proposed approach is no greater than0.0006△, where △ is the FFT resolution. The errors of the amplitude and phase are not greater than 0.3% and 0.3^o, respectively. The general trend of error decreases as the CiR increases, but fluctuation indeed exists with the local minimum er- ror aligning with the coherent sampling where the CiR is an integer.

关 键 词：低频 频谱 快速傅里叶变换(FFT) 窗函数°° 

分 类 号：TN911.6[电子电信—通信与信息系统] O329[电子电信—信息与通信工程]