Robust Identification and Recovery of Obscured CO2 Absorption Signals Using Deep Machine Learning

Abstract

We present the efficacy of deep learning (DL) in identifying and recovering the CO2 absorption line from a noisy spectrum. Following a simulation-based assessment of the DL method's capabilities, it was applied in an experiment utilizing a lock-in amplifier and mechanical chopper as a phase-sensitive detection unit. The results demonstrate that the DL method is fully qualified to replace traditional noise-reduction systems, simplifying spectroscopy while ensuring reliability and intelligence. © 1989-2012 IEEE.