The National Academies Logo
Research Associateship Programs
Fellowships Office
Policy and Global Affairs

Participating Agencies - NIST

  Sign InPrintable View

Opportunity at National Institute of Standards and Technology (NIST)

Statistical baseline spectrum estimation and uncertainty analysis

Location

Information Technology Laboratory, Statistical Engineering Division

RO# Location
50.77.62.C0081 Boulder, CO

Please note: This Agency only participates in the February and August reviews.

Advisers

Name E-mail Phone
Frey, Michael michael.frey@nist.gov 303.497.5690

Description

Statistical estimation of physical spectra is central to many laboratory processes, including, prominently, atom probe tomography, electromagnetic radiation characterization, gas phase spectroscopy, vibrational spectroscopy, chromatography, and NMR experiments. The baselines of measured spectra often include spectra noise, drift, and distortion, all of which obscure spectral lines, peaks, shoulders, and other spectral features of main experimental interest. Indeed, baseline distortions can be greater than and potentially fully obscure even peak spectral intensities.

In common practice, spectral baseline noise and distortion are identified and removed manually by a laboratory analyst. This is time-intensive and judgment-dependent, and the analyst’s subjective intervention greatly complicates the possibility of a subsequent rigorous uncertainty analysis. Ideally, an effective, robust automatic baseline detection/correction procedure would be available that permits statistical uncertainty estimation. A variety of such (semi-)automatic procedures have been proposed for baseline spectrum estimation, based variously on wavelets, penalized least squares, smoothers, and iteratively reweighted quantile regression. The challenge remains to adapt one or a combination of these procedures or to devise new procedures, perhaps machine learning-based, that (1) applies broadly to different physical spectral data without user intervention, (2) capably recognizes different forms of baseline corruption, and (3) allows a well-founded uncertainty analysis. Progress on these problems is an exciting technical challenge with, potentially, significant benefit to many laboratory sciences.

 

References

Bacher R, Chatelain F, Michel O: An adaptive robust regression method: Application to galaxy spectrum baseline estimation. In Acoustics, Speech and Signal Processing (ICASSP), 2016 IEEE International Conference: 4423-4427, March 2016

Guo S, Bocklitz T, Popp J: Optimization of Raman-spectrum baseline correction in biological application. Analyst 141(8): 2396-2404, 2016

Liu X, Zhang Z, Sousa PF, Chen C, Ouyang M, Wei Y, Liang Y, Chen Y, Zhang C: Selective iteratively reweighted quantile regression for baseline correction. Analytical and bioanalytical chemistry 406(7): 1985-1998, 2014

Mani-Varnosfaderani A, Kanginejad A, Gilany K, Valadkhani A: Estimating complicated baselines in analytical signals using the iterative training of Bayesian regularized artificial neural networks. Analytica chimica acta 940: 56-64, 2016

 

Keywords:
Spectral estimation; Baseline estimation; Uncertainty analysis; Automatic statistical procedure; Machine learning;

Eligibility

Citizenship:  Open to U.S. citizens
Level:  Open to Postdoctoral applicants
Copyright © 2014. National Academy of Sciences. All rights reserved. 500 Fifth St. N.W., Washington, D.C. 20001.
Terms of Use and Privacy Statement.