Large Solvent and Noise Peak Suppression by Combined SVD-Harr Wavelet Transform

Abstract

By utilizing singular value decomposition (SVD) and shift averaged Harr wavelet transform (WT) with a set of Daubechies wavelet coefficients (1/2, - 1/2), a method that can simultaneously eliminate an unwanted large solvent peak and noise peaks from NMR data has been developed. Noise elimination was accomplished by shift-averaging the time domain NMR data after a large solvent peak was suppressed by SVD. The algorithms took advantage of the WT, giving excellent results for the noise elimination in the Gaussian type NMR spectral lines of NMR data pretreated with SVD, providing superb results in the adjustment of phase and magnitude of the spectrum. SVD and shift averaged Haar wavelet methods were quantitatively evaluated in terms of threshold values and signal to noise (S/N) ratio values.