THE WAVELET TRANSFORM - A NEW PREPROCESSING METHOD FOR PEAK RECOGNITION OF INFRARED-SPECTRA

The wavelet transform, also called wavelet decomposition, recently int roduced into the applied sciences and available as software packages, is a powerful method for smoothing experimental data. The wavelet tran sform is a mathematical transform for hierarchically decomposing funct ions. It leads to a description of a function, including discrete data vectors or matrices, in terms of a coarse overall shape and details o f a graded sequence. This decomposition is the basis for noise reducti on. At the various levels of decomposition the coarse coefficients are due to the characteristic signals and part of the details may be inte rpreted as noise. The method will be discussed on examples of peak rec ognition in infrared spectroscopy. We will show that some of the wavel et bases lead to a very good compromise between signal/noise ratio enh ancement and preservation of the real data structures. Subsequently it enables a 'Peak Picker' to find the local maxima of the curve corresp onding to real data structures.