Spectra often need baseline correction before data analysis such as peak picking or library search can be performed.

Automatic Baseline Correction

Many automatic baseline correction techniques have been proposed in the literature. They all rely on estimation of the baseline and then subtraction of the estimated baseline. Peak® implements these automatic baseline correction algorithms:

• 'Function Fit': fits a polynomial of through the spectrum and subtracts that curve from the spectrum.
• 'GIFTS auto-leveling': baseline slope and offset are removed by fitting a straight line to the data and iteratively discarding points that do not fit the line well.
• 'Adaptive Iteratively Reweighted Penalized Least Squares' (airPLS): iteratively changes the weights of sum squares errors (SSE) between the fitted baseline and original signals, and the weights of the SSE are obtained adaptively using the difference between the previously fitted baseline and the original signals.

Figure 1. A sample and the baseline-corrected result using Quintic Function Fit.

Manual Baseline Correction

Manual baseline correction often yields better results than automatic correction. The human eye is very good at seeing which points are peaks and which are baseline. In manual baseline correction, the user picks points that define a new baseline. The baseline can be composed of straight line segments between these points, or a smooth cubic spline curve can be fitted through those points. This baseline is subtracted from the spectrum to yield the baseline corrected spectrum. The points the user choses can be saved and applied to other similar spectra.

Figure 2. Points chosen by the user to establish a baseline.

Figure 3. The Applied Baseline Correction.

Using a Fitted Baseline as a Background

An advantage of Peak's manual baseline correction is that, in addition to baseline correction of a spectrum, the fitted (estimated) baseline can be saved as a separate file. This fitted baseline can be used as a background spectrum to ratio against sample spectra. That is, the fitted baseline can serve as a sample spectrum's own background. This has uses in applications where it is difficult to get a clean baseline spectrum, such as open-path spectroscopy, or samples taken in the presence of water vapor.