In this paper we are presented primarily on-line algorithms for the estimation of parameters within a given model structure. The on-line estimation algorithms deals with sequential data, which requires that the parameter estimates be recursively updated within the time limit imposed by the sampling period. In particular, we emphasize on-line parameter estimation scheme and study in some detail. We shall call these algorithms the projection algorithm and the least-squares algorithm. It can apply effectively to adaptive filtering, prediction, and control.