Abrupt changes detection in the time-scale and in the time-frequency planes: a comparative study

We present in this paper a comparison between two segmentation approaches based on time-scale (t-s) and time-frequency (t-f) analysis. The purpose is to detect abrupt changes in the spectral characteristics of multicomponent nonstationary signals. After exposing the two procedures, we carry out different synthetic simulations in order to test and compare the performances of the methods for strictly and not strictly stepwise stationary signals. Finally, we apply the algorithms to musical signals which are not exactly piecewise stationary. This study demonstrates the potential segmentation capability of the proposed methods