An adaptive denoising method for Raman spectroscopy based on lifting wavelet transform. (4th June 2018)

• Record Type: Journal Article
• Authors: Chen, Hao; Xu, Weiliang; Broderick, Neil; Han, Jianda
• Is Part Of: Journal of Raman spectroscopy. Volume 49:Number 9(2018)
• Page Start: 1529
• Page End: 1539
• Publication Date: 2018-06-04
• Subjects: adaptive denoising -- lifting wavelet transform -- noise reduction -- Raman spectroscopy; Raman spectroscopy -- Periodicals; 535.846
• DOI: 10.1002/jrs.5399
• Languages: English
• ISSNs: 0377-0486

Abstract

Abstract: Noise, especially high‐level noise, is a severe problem in Raman spectral analysis. It smears informative Raman peaks, distorts spectral features, and therefore affects final analytical results, particularly in multivariate analysis, which is frequently used in Raman spectroscopy. This becomes even worse when it comes to optical Raman probe‐based biological applications due to limited acquisition time, laser power, and collection efficiency. Noise suppression is usually the first step in the preprocessing procedure of Raman spectral analysis. It is crucial to reduce noise effectively before performing further analysis. Discrete wavelet transform is a useful tool for noise reduction. However, it only provides limited and fixed filter banks, which may not be optimal for the data under investigation. In this paper, a novel adaptive denoising method based on lifting wavelet transform is presented for improving the signal‐to‐noise ratio for a Raman probe‐based system. It enables users to develop an infinite number of lifting schemes from a base wavelet, and with the help of genetic algorithm, the optimal one can be easily found. This method is examined by a set of simulated Raman spectra with various noise level and a set of experimental Raman spectra. Performance comparison with other commonly used denoising methods is made. The results indicate that the proposed method is able to remove noise effectively while retaining informative Raman peaks satisfactorily. Abstract : A novel method based on lifting wavelet transform for random noise reduction in Raman spectroscopy is presented. By applying genetic algorithm, an optimal elementary lifting step is obtained to form the second generation from a chosen base wavelet using lifting scheme. The newly generated wavelet is more suitable to the spectral data at hand than the base wavelet. Test results on single Raman spectra and a classification dataset show the effectiveness of this denoising method.