A CNN-assisted mid-infrared high-sensitivity exhaled ammonia sensor based on cavity ring-down spectroscopy

Abstract

The cavity ring-down spectroscopy (CRDS) is becoming increasingly important in sensitive gas detection for atmospheric pollutant detection, greenhouse gas monitoring, human respiratory gas analysis, and measurement of specific gases in industrial processes. However, the accuracy and sensitivity of CRDS are greatly limited by noise and ring-down time extraction algorithms. To solve this problem, a CNN-assisted high-sensitivity exhaled ammonia sensor based on cavity ring-down spectroscopy was developed in this paper, which is the first time to introduce the convolutional neural network to cavity ring-down spectroscopy. A CNN filter and a neural network-based ring-down time extraction algorithm trained by two different datasets were applied to the sensor, solving the shortcomings of the traditional CRDS. Among several filters, the CNN filter achieves the best performance and the SNR is improved by 3.4 times. Compared to traditional nonlinear fitting algorithm, a sensitivity enhancement factor of 3.48 was obtained and a sub-ppb detection limit can be achieved by using Allan deviation analysis. The experimental results prove the feasibility of deep learning for improving the performance of cavity ring-down spectroscopy gas sensor.