Feature Domain Transform Filter for the Removal of Inherent Noise Bound to the Absorption Signal.

Abstract

We propose to replace the traditional time-frequency domain filtering with feature domain filtering to realize an innovation of filtering algorithm. A feature domain transform filter (FDTF) is composed of the feature domain transform layer based on principal component analysis (PCA) algorithm, the feature domain information extractor based on deep learning and the time domain transform layer. It is established to filter out the noise with the same frequency and phase as the signal and is verified on methane gas. Although FDTF is established based on the simulated data set, the filtering effects of the simulation test set and the experimental data set show that the proposed FDTF outperforms other widely used time-frequency filtering algorithms. The FDTF-assisted methane sensor has good linearity at different concentrations of methane gas. With the FDTF enhancement, the optimized methane sensor performs excellent precision and stability in real-time measurements and achieves the minimum detectable column density of 2.50 ppm·m. This is undoubtedly a successful attempt to move the signal to a new domain for parsing and separation.