Digital Signal Processing Reference
In-Depth Information
severe dynamic range problems. We have shown that ICA methods can
be useful to separate these water artifacts out and obtain largely undis-
torted, pure protein spectra. Generalized eigenvalue decompositions us-
ing a matrix pencil are an exact and easily applied second-order tech-
nique to effect such arefact removal from the spectra. We have tested
this method with simple EDTA spectra where no solute resonances ap-
pear close to the water resonance. Application of the method to protein
spectra with resonances hidden in part by the water resonance showed
a good separation quality with few remaining spectral distortions in
the frequency range of the removed water resonance. It is important
to note that no noticeable spectral distortions were introduced farther
away from the water artifact, in contrast to the FastICA algorithm,
which introduced distortions in other parts of the spectrum. Further,
baseline artifacts due to the intense water resonance can also be cured
to a large extent with this procedure. Further investigations will have to
improve the separation quality even further and to determine whether
solute resonances hidden underneath the water resonance can be made
visible with these or related methods.
Search WWH ::
Custom Search