This paper proposes a novel structural health monitoring framework for damage detection of smart structures. The framework is developed through the integration of the discrete wavelet transform, an autoregressive (AR) model, damage-sensitive features, and a support vector machine (SVM). The steps of the method are the following: (1) the wavelet-based AR (WAR) model estimates vibration signals obtained from both the undamaged and damaged smart structures under a variety of random signals; (2) a new damage-sensitive feature is formulated in terms of the AR parameters estimated from the structural velocity responses; and then (3) the SVM is applied to each group of damaged and undamaged data sets in order to optimally separate them into either damaged or healthy groups. To demonstrate the effectiveness of the proposed structural health monitoring framework, a three-story smart building equipped with a magnetorheological (MR) damper under artificial earthquake signals is studied. It is shown from the simulation that the proposed health monitoring scheme is effective in detecting damage of the smart structures in an efficient way.