Burg谱的阶数依赖性与最佳正则化参数的选择

利用Burg法分析了被白噪声污染的功率谱,并将结果与传统周期图法进行了比较。当阶数比较小时,Burg谱比较平坦,信号被噪声掩盖;随着阶数的增加,分辨率增加;当阶数进一步增加时,虚假峰出现,并逐渐增多,以至于信号被噪声淹没。这一结果表明,系统一定存在某个阶数,在这个阶数附近Burg谱质量最好。指出了这个阶数就是最佳正则参数;而且还指出了,只需进一步考虑残差向量的归一化累积误差,并同K-S判据相结合就可以确定它。

The Order-Number Dependence of Burg Spectrum and Choice of Optimum Regularization Parameter

The power spectrum of the signal polluted with the white noise is analysed by Burg method,and its result is compared with the periodogram method.When the ratio of the order-number is little,Burg spectrum is over-smooth and the signal is covered by the noise;with the increase of the order-number,the resolution increases too. When the order-number further increases,the false peak occurs and gradually increases,so that the signal is submerged by the noise.The results show that there must be an order-number and...