A novel narrow band digital filter and its application to multivariable system identification

Abstract A novel narrow band time-varying digital filter is proposed, which has desirable properties such as global asymptotic stability, asymptotic noise annihilation and asymptotic signal tracking. It is shown that the proposed filter is comparable to the Kalman filter in performance, but with substantial computational simplicity; no Ricatti equation is involved. It is basically a Fourier analysis method but the Fourier coefficients are found recursively. The application of the proposed filter for on-line identification of a linear multivariable system subject to both deterministic and stochastic disturbances is presented; simulation results are given.