Wavelet-based burst system model change detection

A new scheme for detecting system burst changes is developed based on the continuous wavelet transform (CWT). The system is concisely represented by its time-frequency representation (TFR), the ratio of the CWTs of the output and the input. The TFR is first estimated, assuming that no system changes occur during a time period. A chi-square test is then executed to test the TFR estimate\u27s match to the data. System model changes are detected at the time samples when the assumption that the system is time invariant is broken. Simulations verify the capability of the proposed algorithm to detect burst system changes