Method of correcting spectral data for background
AssigneeThe Perkin-Elmer Corporation
InventorsJohn B. Collins, Marc L. Salit
Life of this patent
Estimating the term…
Abstract
A method of producing spectral data corrected for background includes generating spectral data, ascertaining background and subtracting background points to produce the corrected data. To ascertain background the spectral data are masked initially at the spectral peak of interest, and a least squares curve is fitted to the masked data. A runs test is performed to detect whether the number of runs exceeds a selected significance level. If so, the data are further masked and a curve is refitted. These steps are iterated until no further significant number of runs is detected. Similar testing for outliers is then effected with similar masking, a final curve represents the background level.
Claims28
A method of producing spectra corrected for background level, with a spectrographic instrument including means for effecting a spectrum having a background level and a spectral peak at a characteristic wavelength, means receptive of the spectrum for generating and storing spectral data as discrete data points, and means for processing and displaying the spectral data, the method comprising operating the instrument so as to generate and store a set of spectral data, ascertaining background level in the spectral data, subtracting points in the background level from corresponding discrete data points to produce corrected spectral data, and displaying the corrected spectral data, wherein the step of ascertaining background level comprises:
- initially masking the spectral data at the characteristic wavelength over a preselected initial masking width, thereby modifying the spectral data into a set of initially diminished data;
- initially fitting a smooth curve to the initially diminished data;
- processing the spectral data by selecting a statistical significance level, performing a runs test on residuals of the diminished data with respect to the smooth curve so as to detect a largest run, determining a wavelength for a largest residual of the largest run or a median wavelength for the largest run, masking the diminished data at the determined wavelength over a selected subsequent masking width, and refitting a smooth curve to the masked diminished data;
- iterating, with residuals for cumulatively masked diminished data, the steps of processing until the number of runs is not significant, thereby further modifying the spectral data into a set of runs diminished data;
- and
- employing runs diminished data in fitting a final smooth curve representing the background level for producing the corrected spectral data.
Dependencies are inferred from the claim text.