Detection and diagnosis of model parameter and noise variance changes with application in seismic signal processing

The change detection and diagnosis methods have gained considerable attention in scientific research and appears to be the central issue in various application areas. These applications need some robust change detection schemes to work well and separate the changes in the experimental conditions from the real changes in the system, especially for systems with arbitrary and non-stationary known or unknown inputs. The objective of the paper is to develop such kind of change detection and diagnosis scheme. In the first part of the paper we give the conceptual description of some classical change detection schemes based on sliding windows and likelihood techniques. Then, starting from these classical change detection schemes, a new algorithm able to discriminate between the model parameter and noise variance changes is presented. Finally, we include some Monte Carlo simulations for change detection in a second order FIR model and experimental results obtained in analysis of seismic signals, using the proposed approach.