| Fourier transform infrared (FTIR) chemical sensors have potential for use in remote sensing environments. In addition to remote sensing applications, FTIR-based chemical sensors also have potential use in noninvasive clinical applications. However, the applicability of this technique to quantitative analysis of target analytes has been limited by two fundamental problems: (1) the lack of a stable representative background spectrum, and (2) the requirement that the sensing apparatus be rugged, reliable, compact, and inexpensive.; To address these requirements, our laboratory is developing data analysis methodologies based on the application of narrow bandpass digital filters directly to short FTIR interferograms. With the design of digital filters that pass only the frequencies corresponding to a specific analyte band, spectral information can be isolated directly from a short interferogram. The coupling of digital filtering and partial least-squares regression was shown to provide further selectivity in matrices where filtering alone could not guarantee sufficient selectivity for the target analyte. Direct use of the interferogram segment eliminates the need for a separate spectral background measurement, while reducing the length of travel of the moving mirror of the interferometer increases the ruggedness and reliability of the spectrometer by decreasing both the data acquisition and data processing requirements for the measurement. These two potential benefits could lead to a compact, simpler, and less expensive sensor.; In this work, the proposed data processing algorithm has been shown to overcome problems that hinder quantitative analysis of FTIR remote sensing data. This was demonstrated with the use of laboratory absorption spectra of varying concentrations of benzene and nitrobenzene, and controlled emission spectra of varying concentrations of sulfur dioxide in field data collected with a cell at 50, 80, 120, and 150{dollar}spcirc{dollar}C. Success was also reported in the quantitative analysis of physiological levels of glucose in matrices where water, triacetin, and bovine serum albumin (BSA) were also present in high levels. Triacetin and BSA were used to model the total triglycerides and protein in blood. The above work was performed using both Fourier and time-domain filtering techniques, and these interferogram-based results were very comparable to those obtained through the use of spectral-based methods. |
