The designs of optimal linear quadratic estimator and tracker for the linear continuous-time systems with unknown disturbances have been improved in this paper. Compared with the advanced robust control design methodologies developed in the literature, the proposed method significantly improves both the unknown input estimation and the servo performance for the systems subject to the unknown disturbances. The aforementioned design objective can be achieved due to the fact that the proposed approach ensures the relative stability in both the unknown input estimation and the servo performance from the theoretical point of view. Besides, based on the equivalent input disturbance principle developed in the literature, the proposed approach is applicable to the system with a class of mismatched input disturbances. Furthermore, the newly proposed unknown input estimation technique is able to extend the recently developed generalized optimal linear quadratic tracker for continuous-time systems with known disturbances to those with unknown disturbances.