Automatic Segmentation of Nonstationary EM Emission of Electronics Product

Abstract

Characterizing complex and multi-functional devices is a very challenging task. One problem in statistical near field analysis on complex electronic products is the emergence of nonstationary electromagnetic (EM) signals. Such emergence will lead to an incorrect decision if the signal is used as an input to propagation analysis. The most appropriate approach to this problem seems to be one based on the segmentation of the nonstationary time series obtained from measurements into an ensemble of piecewise stationary signals. In this paper, we propose three approaches for automatic segmentation of nonstationary EM emission signals: short-time energy (STE), short-time zero-crossing rate (STZCR), and short-time kurtosis (STK). Test results show that STE is the best in terms of success in segmenting the nonstationary signals to achieve piecewise stationary time series and being less computationally intensive.