Adaptive Mra Digital Filter Using Wavelet, and Narx Neural Network for Ground-Based Transmission Spectroscopy

The aim of this work is to remove systematic noise from transmission spectroscopy observations of exoplanets observed from the Earth s surface using unsupervised clustering and multiresolution analysis (MRA) based on a Daubechies 12 as mother Wavelet. This will train a neural network of NARX topology (Nonlinear Autoregressive Model with Exogenous Inputs) which learns through the proposed method to remove the noise present in the observations. Within the method, it is proposed to remove the noise using the MRA in two different ways: applied for each wavelength observed to then recompose the Light Curve and directly on the Light Curve. From these, the best method is chosen based on the root mean square error (RMSE) metric, which was of the order of 10-7 for the best neural network in this research. © 2024 IEEE.