Machine learning for quantitative gas sensing using sparsely sampled time-domain spectroscopy

Thumbnail Image

URL

Journal Title

Journal ISSN

Volume Title

School of Electrical Engineering | Master's thesis

Authors

Date

2025-09-16

Department

Major/Subject

Autonomous Systems

Mcode

Degree programme

Master's Programme in ICT Innovation

Language

en

Pages

45

Series

Abstract

Fourier transform infrared spectroscopy (FTIR) is widely used in volatile organic compound (VOC) detection, but traditional uniform high-rate acquisition is time-consuming and results in significant data redundancy. This study employed a compressed sensing method to sparsely reconstruct the interferogram under non-uniform sampling conditions, significantly reducing the sampling rate while preserving the spectral structure. This method, combined with a machine learning model, was used to predict the absolute concentration of a ternary VOC mixture consisting of acetone, isopropyl alcohol, and ethanol. The results showed that even when the sampling rate was reduced to 20%, the prediction accuracy of the reconstructed spectrum remained essentially consistent with that of the full sampling method. Furthermore, linear regression models outperformed deep learning methods on both simulated and reconstructed datasets, with ridge regression being the most accurate. To conclude, this study validated the effectiveness of combining compressed sensing spectral re-construction with machine learning prediction and demonstrated its potential for application in air quality early warning systems for enclosed environments such as cockpits, cargo holds, and warehouses.

Description

Supervisor

Savin, Hele

Thesis advisor

Koppens, Frank

Keywords

machine learning, compressed sensing, Fourier Transform Infrared Spectroscopy (FTIR), Volatile Organic Compounds (VOCs), spectral reconstruction, concentration prediction

Other note

Citation

Permanent link to this item

https://urn.fi/URN:NBN:fi:aalto-202510208323

Collections

[dipl] Sähkötekniikan korkeakoulu / ELEC