A three-phase, computational method for the baseline correction of vibrational circular dichroism (VCD) spectra has been proposed. In the first phase the raw spectrum is subdivided into m segments (or regions) resulting in p rough estimates of the baseline. A second phase uses gradient characteristics to discriminate between baseline and band response for each baseline, in turn. In the final phase all baselines are interrogated simultaneously by assigning the median estimate of each differential response's distribution to the true baseline. Using VCD spectra of (R)-camphor as test cases, this work demonstrated that the accurate removal of baseline components is readily achievable with minimal user intervention. Baseline correction also demonstrated flexibility in that prior information, such as the symmetry of a baseline-free VCD spectrum, is readily used during the correction protocol. Although three adjustable parameters are present in the base algorithm, optimal performance and full automation were attainable following the use of analysis of variance (ANOVA) to analyze simulated bipolar spectra. These ANOVAs suggested that band point discrimination could be discarded and the remaining two default parameters adopted.
Andrew T. Weakley, D. Eric. Aston, and Peter R. Griffiths, "Automatic Baseline Correction of Vibrational Circular Dichroism Spectra," Appl. Spectrosc. 67, 1117-1126 (2013)