摘要

A new phase difference correction method, based on asymmetric windows, is proposed to correct the errors of frequency, phase, and amplitude in discrete spectra. Classic windows show the constant time delay and the same linear phase response due to their symmetric properties. Removal of the symmetry constraint on windows can result in variable time delays and different spectral phase responses. Based on the phenomenon of alterable spectral phases in asymmetric windows, we present a new approach, which could overcome inherent defects in traditional correction methods. The simulation results show that the new method can correct the errors of neighboring frequency components with high precision through the selection of proper asymmetric windows. Moreover, the new method can avoid the errors induced by a mistaken location of the spectral line, which is one of the major problems in traditional correction methods. When the symmetric windows in traditional methods are replaced by asymmetric windows, the improved algorithm also exhibits stronger robustness against additive noise. The comparative study reveals that each algorithm has its own advantages. In general, when the SNR is high, the interpolated method is a good choice for its simplicity; while the phase difference based methods provide better performance when the SNR is low. The new approach is recommended for closely spaced components or when high noise is involved.