A Bayesian Approach to Binary Classification of Mid-Infrared Spectral Data With Noisy Sensors

The problem of classifying substances using MIR laser and sensors with low signal-to-noise ratio remains challenging. The existing methods rely largely on using lasers at multiple wavelengths and expensive high quality sensors. We propose and demonstrate a statistical method that classifies spectral data generated from MIR imaging spectroscopy experiments using few wavelengths and inexpensive detector arrays while still achieving high accuracy. Results with quantifiable analytic performance are obtained by attributing probability distribution functions to the images obtained and implementing a binary decision process. Our method can provide a solution with as few as a single measurement and allows the use of low SNR sensors. This can increase throughput and lower costs on security checkpoints, pharmaceutical production monitoring, industrial quality control, and similar applications.