应用小波变换实现光谱的噪声去除和基线校正

为消除实测光谱信号中的噪声和基线干扰,给出了一种基于小波变换实现两者同时去除及其参数选择的新方法。该方法通过对光谱信号在小波域内的低频段小波系数置零来实现基线校正,通过对较高频段小波系数阈值处理来实现噪声去除;并利用纯光谱和常见基线、噪声的仿真信号,通过兼顾重构信号整体逼近和特征峰处局部逼近的评估系数η来实现小波基、分解层数、阈值估计方式等参数的选取。仿真实验表明,仿真信号采用sym5、db5或db9等小波基进行5次分解,然后低频成分置0及所有高频成分利用单层Heursure阈值估计算法进行硬阈值处理较为合适。进一步的实验表明,该方法对实测光谱中噪声和基线的消除是行之有效的。

Denoising and baseline correction of spectrum by wavelet transform

A denoising and baseline correction method based on wavelet transform and their parameters choosing method were developed to remove noise and correct baseline in the spectrum effectively. Baseline wander was removed by setting approximation coefficients to zero, and noise was removed by the method of threshold detail coefficients of higher-frequency period in wavelet filed, and with simulated signals of pure spectrum and familiar baseline and noise, the parameters such as wavelet basis, decomposed level, threshold selection rule and so on were chosen by the evaluation coefficient η which considered approaches of integral signal and every characteristic peak simultaneously. The simulated experiment shows that it is more suitable that the simulated signal is decomposed at level 5 on the wavelet basis of sym5、db5、db9 and so on, and then the approximation coefficients are set to zero and threshold detail coefficients by hard threshold with single layer Heursure threshold selection rule. Further experiment shows that this method provides a novel and to correct effective way to remove noise and to correct baseline in the real spectrum.