Spectral Preprocessing
Spectrum transformation centers and scales each individual spectrum to remove the influence of baseline shifts and variations in light intensity. Standard normal variate is used in near-infrared spectroscopy to ensure that the chemical information in the signal is not overwhelmed by physical differences between samples. Each data point is transformed by subtracting the mean of the spectrum and dividing by its standard deviation.
Signal Isolation
Primary purpose of this transformation is to account for the effects of varying particle size and path length in solid samples like paper and pulp. When standard normal variate is applied, the spectra are brought to a common scale, which makes it easier to compare them directly. This step is performed before any regression analysis to improve the stability of the resulting models.
It does not require a reference spectrum, making it an efficient way to process large amounts of data.
Accuracy Enhancement
Removing the noise caused by light scattering allows the sensors to detect smaller changes in the concentration of additives. Using standard normal variate reduces the error in predicting the physical properties of the moving web on a paper machine. It is effective for analyzing materials with rough surfaces or inconsistent densities.
This mathematical correction ensures that the reported values are based on the actual chemistry of the substrate.
