Online detection and estimation of parameter jumps

Abstract
The problem of fast online detection and estimation of parameter jumps is discussed. The proposed solution uses a modified test of hypothesis to select the appropriate parameter estimation algorithm. The most significant feature of the test is the use of different thresholds for acceptance and for rejection of the null hypothesis. This feature creates a maybe region which can be used to advantage. The resulting combined filter algorithm can minimize the upper bound of missed-jump and false-alarm probabilities. Computer simulations using a single sample testing show good detection speed.

This publication has 4 references indexed in Scilit: