An adaptive Kalman filtering algorithm based on back-propagation (BP) neural network applied for simultaneously detection of exhaled CO and N2O.

Sheng Zhou,Ningwu Liu,Chongyang Shen,Lei Zhang,Tianbo He,Benli Yu,Jingsong Li
DOI: https://doi.org/10.1016/j.saa.2019.117332
2019-01-01
Abstract:A compact high-resolution spectroscopic sensor using a thermoelectrically (TE) cooled continuous-wave (CW) room temperature (RT) quantum cascade laser (QCL) was demonstrated for simultaneous measurements of exhaled carbon monoxide (CO) and nitrous oxide (N2O). The sampling pressure was optimized to improve the sensitivity, the optimal pressure was determined to be 150 mbar based on an optical density analysis of simulated and measured absorption spectra. An adaptive Kalman filtering algorithm based on back-propagation (BP) neural network was developed and proposed for real-time exhaled breath analysis in order to perform fast and high precision on-line measurements. The detection limits (1σ) of 1.14 ppb and 1.12 ppb were experimentally achieved for CO and N2O detection, respectively. Typical concentrations of exhaled CO and N2O from smokers and non-smokers were analyzed. The experimental results indicated that the state-of-the-art CW-QCL based sensor has a great potential for non-invasive, on-line identification and quantification of biomarkers in human breath.
What problem does this paper attempt to address?