Research of spectrum preprocessing based on wavelet transform and principal component analysis

The accuracy of spectrum preprocessing, as the precondition of precise qualitative and quantitative spectrum analysis, performs importantly. This paper presents a new approach of combining wavelet transform (WT) with principal component analysis (PCA) in the preprocessing of spectrum analysis. The multi-scale characteristic of WT in time and frequency domain is utilized to extract useful signals and eliminate the noise and the baseline excursion from the spectrum; meanwhile, PCA method which can lower data dimension effectively and rapidly is adopted to deal with the overlapped, crossed and redundant information that exists in the spectrum. Application of the new approach in the analysis of samples is mainly discussed in this paper. The result of experiment shows that this method is suitable and advanced.