Application of Machine Learning Methods for Signal Processing in Piecewise-Polynomial Bases

This article is devoted to digital processing of radiation signals in the field of geophysics based on spectrum analysis. With the help of these signals, it was studied whether it is possible to determine the layers of underground mineral wealth. A machine learning method was proposed to determine the layer where the ores are located. Haar’s piecewise-quadratic basis was chosen as the mathematical model of the machine learning method due to the small number of calculations. The purpose of choosing this model is that in the digital processing of signals, the number of near-zero values of spectral coefficients is large, and these values can be discarded as signal noise. This process helps us reduce the amount of data. As a result of comparing the values of spectral coefficients that are not close to zero, it gives an effective result in determining the location of the ore layer.