he outcome of applying a moving average process on the peak spectrum derived

ving the estimated baseline using BWH for the spectra shown in Figure 5.12.

ne stands for the original peak spectrum shown in Figure 5.12. The dark line is

ectrum derived from the moving average process.

e generation of the merged and unique peaks

oothing a peak spectrum derived from a raw spectrum with an

d baseline removed, the other important benefit can be seen. It can

rom Figure 5.13 that the artifacts which have low intensities have

n smoothed, leading to less number of glitches. Therefore, the

difference between the true peaks and the artifacts has been

The next question is how to separate the true peaks from the

A statistical process is therefore required.

ng no a priori knowledge which peaks are the true peaks for

y, a density estimation approach can be used for the

ation between the true peaks of interest and artifacts. A density

n process as a classic statistical process can be used to deliver a

ve model to identify the significant difference between the true

d the artifacts in a smoothed peak spectrum.

e intensity heights (for short, the heights) in a smoothed peak

are positive, no matter they are the true peaks or the artifacts.

r, if the height difference between the true peaks and the artifacts

ently great and every peak has a trend that it moves to the peak

adually and moves down from the peak height gradually, the

f the heights will follow a sharp decay pattern. All peak heights