X射线荧光光谱的交叉验证傅里叶变换平滑

Use cross validation Fourier transform to smooth X-ray fluorescence spectra

针对当前谱光滑处理方法大多需要大量的实验以获取先验知识的问题,提出了一种基于交叉验证的傅里叶变换去噪法。在傅里叶变换去噪的基础上,该方法利用交叉验证确定最优阈值,实现X射线荧光光谱的自动去噪。实验结果表明:不同信噪比下,交叉验证确定的最优阈值和实际最优阈值的相对误差小于30%,该方法取得良好的光滑效果。和小波阈值去噪法相比,该方法具有使用简单的优点;和Savitzky-Golay多项式滤波器法相比,该方法具有更高的准确度。

[Background] Spectral smoothing is the basis of qualitative and quantitative analysis in X-ray fluorescence （XRF） spectra. However, most of the current spectral smoothing methods require a large number of experiments to obtain prior knowledge. [Purpose] This study aims to propose a novel method to automatically denoise XRF spectra. [Methods] Based on Fourier transform denoising, the optimal threshold is determined by cross validation to achieve automatic noise elimination. [Results] Simulation results show that the relative error between the optimal threshold determined by cross validation and the actual optimal threshold is less than 30% under different signal-to-noise ratios. [Conclusion] Proposed cross validation Fourier transform method has the advantages of convenience compared with wavelet threshold denoising method, and preserve higher precision than that of Savitzky-Golay polynomial filter method.