Two-scale Sequence Generation Method Using Machine Learning for Discrete Wavelet Transform

Wavelet Transform is a method that can analyze non-stationary and non-linear signals. In particular, by Discrete Wavelet Transform(DWT), we can perform signal analysis at high-speed with low computation. Analyzing a signal by Wavelet Transform, the Mother Wavelet(MW) used must be appropriate for the signal. However, MWs capable of generating two-scale sequences for DWT are even less than MWs in Continuous Wavelet Transform(CWT). Therefore, analyzing a signal at high-speed using DWT, we may not be able to select the appropriate MW for analysis. In this study, we propose a method to generate a two-scale sequence corresponding to an arbitrary MW by using machine learning.