Infinite impulse response filters for direct analysis of interferogram data from airborne passive Fourier transform infrared spectrometry

Infinite impulse response (IIR) filters and support vector machines (SVMs) are applied to interferogram data to implement the automated detection of heated ethanol plumes by airborne passive Fourier transform infrared (FT-IR) spectrometry. To simulate monitoring of industrial emissions, the spectrometer is mounted in a downward-looking mode on a fixed-wing aircraft and used to detect plumes released from a portable emission stack. Both IIR and finite impulse response (FIR) filters having similar frequency responses are designed and applied to the interferogram data to extract the information associated with the ethanol C-O stretching band at 1066 cm−1. This study investigates the effects of distortions in the interferogram introduced by the nonlinear phase responses of the IIR filters on the ability of the SVMs to recognize ethanol. In addition, three ways to apply the filters to the interferogram are explored. These studies reveal that the phase distortion introduced by the IIR filters causes no measurable degradation in classifier performance. This allows the intrinsically lower filter order of the IIR filters to be used to reduce the length of the interferogram segment required to implement the filtering. With an independent prediction set of interferograms collected during overflights of the stack, the best SVMs miss detecting ethanol in less than 4% of the cases while achieving false detection (false positive) rates of less than 0.5%. This level of performance is accomplished with an interferogram segment of only 24 points, including those points required to implement the filtering. The ability to detect analyte vapor plumes with such a short interferogram segment is particularly advantageous in the airborne measurement because of the potential to increase the interferogram acquisition rate, thereby improving the ability to resolve ground-based targets.