This paper is concerned with the state estimation of nonlinear systems subject to matched/mismatched system uncertainties. An adaptive control is applied in the sliding mode observer (SMO) so that the approaching condition can be guaranteed in an efficient way. Moreover, due to the existence of mismatched uncertainties, the sliding motion is no longer insensitive and the state observation performance may be deteriorated with an inappropriate selection of observer gain. To guarantee estimation stability and achieve better estimate performance simultaneously, a compromised way for the determination of the control gains is presented. Existence of the solution is characterized by linear matrix inequalities (LMIs).