structural health monitoring (shm) plays a key role towards condition assessment of large-scale civil structures using modern sensing technology. once the rich vibration data is collected, important system information is extracted from the data and sub- sequently such information is used for necessary decision making including adopting maintenance, retro tting or control strategies. system identi cation is one of the key steps in shm where unknown system information of the structures is estimated based on the response measurements. however, depending on excitation characteristics or system behavior, vibration measurements become complicated where traditional methods are unable to accurately analyze the data. in this thesis, multivariate empirical mode decomposition (memd) method is ex- plored to undertake ambient system identi cation of structures using the multi-sensor vibration data. due to inherent sifting operation of emd, the traditional memd re- sults into mode-mixing that causes signi cant inaccuracy in structural modal identi - cation. in this research, independent component analysis (ica) method is integrated with the memd to alleviate mode mixing in the resulting modal responses. the pro- posed hybrid memd method is veri ed using a suite of numerical, experimental and full-scale studies (e.g., a high-rise tower in china and a long-span bridge in canada) considering several practical applications including low energy modes, closely spaced frequencies and measurement noise in real-life buildings and bridges. the results show signi cantly improved performance of the proposed method compared to the standard emd method and therefore, the proposed method can be considered as a robust ambient modal identi cation method for exible structures.