Parameter estimation of chirp signals has been studied for a long time. The maximum likelihood estimator has been given and applied widely. However, in many applications, we have known the distributions of the parameters to be estimated and the Bayes estimator can be used. In this paper, by incorporating a priori knowledge, the linear minimum mean square error (LMMSE) estimation of the parameters is given. This method outperforms the traditional maximum likelihood estimator. Besides, the LMMSE estimation is realized based on two linear phase models. One is the traditional linear phase model and the other is the optimal linear phase model which is presented recently. The result shows that the estimator based on the optimal linear model has a better performance.