Abstract:

The signal processing with Gabor wavelets is improved in this paper. When a s ignal with single frequency is selected, the interferences in both time and freq uency domains must be suppressed to minimum. Under these restraints, parameters  of the selected signals are optimized.  Moreover, it must be sure every time th at  the selected peaks are the strongest one in the difference spectrum. After all F ID with single frequency are taken out from the original FID, some huge and wide  peaks are given up for the other treatments. The remaining small peaks are reco nstructed. For liquid samples, the traditional Lorentzian function can be used g ene rally. Here, a modified Lorentzian function is given and the projections of all  sel ected signals are used for construction of a so called `projection spectrum',   or only two parameters (amplitude and frequency) are used to construct  a `lines spectrum'. It is proved by simulation and experimental data processing th at the detail information of small signals covered in a huge peak can be exposed  by means of the so called `microscope property' of  wavelets.

Key words: gabor wavelets, extraction of small signals, NMR, parameters optimization