ENHANCEMENT OF 2D NMR-SPECTRA USING SINGULAR VALUE DECOMPOSITION

An enhanced 2D NMR spectrum can be obtained by zeroing some of the singular values prior to reconstruction of the products of a singular value decomposition (SVD) of the spectrum. Large artifacts can be removed by zeroing the largest singular values, and noise can be reduced by zeroing the smallest singular values. To accomplish this, several reconstructions of the SVD products are created with individual or small ranges of singular values to determine the correspondence between the singular values and the eigencomponents. Undesired large components are then removed by zeroing a few of the largest singular values, and noise is removed by zeroing a large number of the smallest singular values. Typically, zeroing from one to five of the largest singular values and the smallest half of the singular values will result in a reconstruction with large artifacts removed, an improvement in signal-to-noise, and no loss of actual signal. © 1990.