Analysis of a low-complexity change detection scheme

• In many applications, for example in fault detection, it is important to discriminate between changes in system dynamics and abrupt changes in the disturbance level. A new low-complexity change detection method based on the average behaviour of the estimated impulse response parameters of the normalized least mean-square (NLMS) algorithm is presented. The solution includes second-order Kalman filters based on exponential transient models for parameter convergence. Explicit formulas for time-varying state covariances and Kalman gains are given. The receiver operating characteristics (ROC) is also computed and used for performance evaluation. The effects of the approximations in the averaging analysis that occur for high adaptation gains are handled with an experimental ROC analysis.

Ämnesord

TEKNIK OCH TEKNOLOGIER  -- Elektroteknik och elektronik -- Reglerteknik (hsv//swe)

ENGINEERING AND TECHNOLOGY  -- Electrical Engineering, Electronic Engineering, Information Engineering -- Control Engineering (hsv//eng)

Nyckelord

Adaptive algorithms

Approximation theory

Kalman filtering

Mathematical models

Matrix algebra

Parameter estimation

Riccati equations

Signal receivers

Time varying systems

Low complexity change detection

Normalized least mean square

Receiver operating characteristics

Transient models

Signal detection

Publikations- och innehållstyp

ref (ämneskategori)

art (ämneskategori)