TIME-DOMAIN SEGMENTATION METHODS FOR COMPLEX SIGNAL ANALYSIS

Abstract

This article analyzes modern time-domain segmentation methods used in complex signal analysis. Fixed-scale, adaptive, and multiscale approaches, as well as structural analysis and change detection methods, are considered. It is shown that in existing approaches the segmentation scale is mainly treated as an auxiliary parameter rather than a controlled factor in signal ensemble formation.

A comparative experimental study is performed for different scenarios of signal structural nonstationarity, including stationary conditions, local degradations, gradual structural changes, and impulsive interference. The evaluation focuses on degradation localization accuracy, segmentation-scale stability, and over-segmentation effects.

The results demonstrate inherent limitations of existing time-domain segmentation methods in ensemble formation tasks and motivate the development of approaches with controlled segmentation-scale selection based on local signal structure and ensemble requirements.