Assurance of integrity of structures is the main task of structural health monitoring. The condition (health) of a structure may be revealed from its dynamic characteristics in response to natural or man-made loads. As the measurement of the natural or operating forces on actual real-life structures is generally formidable, dynamic characteristics of the structures have to be extracted from the available output signals only. The stochastic subspace identification (SSI) technique is adopted in this study to identify the equivalent system parameters of the discrete-time state equation using covariance functions of the measured output signals. With the system parameters realized, the method of damage locating vector (DLV) is then considered for further assessment. Members with nearly zero stress under the loadings of DLVs are considered potentially damaged, whereas the DLVs are derived from singular value decomposition of the change in flexibility matrix of the structure before and after the damage state. In this study, the feasibility of DLV method for damage detection of a planar structure is explored based on its seismic responses. To comply with the desired output-only scenario, the information of ground motion is discarded in the stage of SSI analysis. Despite the non-stationary nature of earthquakes, the proposed scheme has been proved sufficient for damage localization of structures from global responses (floor accelerations). The damaged locations can be identified when the structure is fully observed, regardless of single or multiple damages. In the case of partial but co-located observation (damaged floors observed), the damaged locations can still be identified with acceptable accuracy and reliability.