The need of effective knowledge of the status of ageing structures has increased markedly the interest in the development of structural monitoring techniques with the aim of identifying possible earthquake related damage. For this purpose, usually the changes in dynamic parameters (e.g., elastic moduli, shear wave velocity, fundamental frequency) during an earthquake are monitored. It is considered the simplest way to detect the onset of structural damage and its impact on the dynamic response of the building. Moreover, in presence of soil–structure interaction, the fundamental frequency of the system, and hence its variation, estimated from seismic data recorded in the building, can give a misleading evaluation of the dynamic properties of the structure. This is due to the fact that detected variations of the frequencies may be related as well to nonlinearities in the foundation, the structural elements, or both (e.g. Trifunac et al. 2008; Michel et al. 2011; Rahmani et al. 2015). For monitoring the status of a building, it is necessary to separate the response of the building from soil–structure coupling. In recent years, monitoring the status of buildings by installing sensors in different positions in the building and processing the recorded motions has been made easier by the rapid development of low-cost instruments and the increased capacity for data acquisition and transmission. Assessing damage evolution through monitoring the structural response allows one to better understand the dynamic behavior of structures. Since only a limited number of moderate to strong earthquakes occurs close to well-instrumented buildings, the nonlinear dynamic behavior is usually studied through numerical analysis. This kind of analysis presents several drawbacks, like simplified modelling assumptions, and the difficulty in the proper calibration of mechanical parameters for damaged structures and materials. Regarding methods based on empirical data, the seismic response of a system is classically determined by estimating the characteristics of the normal modes by using the vibrational approach (e.g., Chopra 1996). Moreover information on the building’s properties are obtained by studying seismic waves through the building by the waveform approach (e.g., Kanai 1965; Safak 1989; Snieder and Safak 2006; Kohler et al. 2005, 2007; Prieto et al. 2010; Todorovska 2009b; Rahmani and Todorovska 2013; Bindi et al. 2014; Wen and Kalkan 2017). Starting from the idea of Kanai (1965), the dynamic behavior is studied via wave propagation through the building by analyzing the shear wave velocity, which is a function of the structural material. The main advantage of deconvolution interferometry (when the station at roof is used as reference) is that it allows one to retrieve the structural response, irrespective of the amount of coupling with soil, manifested by soil–structure effects (e.g. Snieder and Safak 2006; Todorovska 2009a, b; Michel et al. 2011). Both vibrational and waveform approach are not reliable when the dynamic response evolves over time because of the loss of information about the local spectrum and the timevarying behavior. Several techniques for signal analysis and identification of structural dynamics have been proposed in the past with the aim to evaluate the dynamic characteristic of a soil– structure system over time.