New recursive identification algorithm for time-varying systems

A new recursive identification algorithm for time-varying systems is developed in this paper. It is demonstrated that the parameters of the time-varying model can be expressed by an explicit approximation using cosine series. By using the orthogonal characteristics of basis functions, efficient updating of matrices and recursion of parameters are implemented in the algorithm. In contrast with the conventional basis function based methods, RLS and LMS algorithms, the proposed algorithm has less sensitivity to selection of basis functions and data window, and has higher tracking performance even in rapid varying processes.