Abstarct: Nowadays, there are many important existing structures which a comprehensive knowledge about their dynamical properties, i.e. natural frequencies, damping ratios, and mode shapes, is necessary for any structural evaluation, control or rehabilitation purpose. In recent decades, as a result of developments in the signal processing techniques, several methods have been proposed in this concept and each method has some benefits and shortcomings. In this thesis, we have investigated two recently developed algorithms in signal processing context and try to fix their restriction when using in modal parameters identification purpose. In the first proposed method, Local Mean Decomposition (LMD) algorithm has been utilized in modal parameters identification context for the first time. To prevent the mode mixing phenomenon, a preprocessing step including filtering of the responses has been used. The ability of the proposed method has been investigated due to some numerical and experimental simulations and the results compared with those obtained from Hilbert-Huang Transform (HHT). As a major advantage, the proposed method can extract natural frequencies and damping ratios of all active modes from only one suitable measurement. However, the responses at all DOFs are required for estimation of the structural mode shapes. The identification results from numerical simulations in the free vibration test showed that the maximum estimation error of the natural frequencies for the first proposed method and the HHT method were 1% and 0.5%, respectively. Also, the maximum estimation error of the damping ratios for the first proposed method and the HHT method were 4.1% and 6.2%, respectively. The mode shape vectors identified by both methods had MAC values greater than 99.8%, which means a high level of accuracy. It should be mentioned that the estimation errors in both methods increase for poorly-excited modes. In addition, owing to the filtering step, the existence of noise in the measurements has no considerable effect on the identification results of both methods. Investigating the plots of instantaneous phase angles and the natural logarithm of instantaneous amplitudes showed that the first proposed method in comparison with the HHT method is less affected by the end effect issue. Also, the proposed method showed fewer deviations from the linear form in a longer time interval. The identification results of the experimental simulation with ambient vibrations showed that the natural frequencies are accurately identified by either the first proposed method or the HHT ones. Moreover, the maximum estimation error of the damping ratios in the case of 10% noise is restricted to 7% and 6%, respectively, for the first proposed method and the HHT ones. The second proposed method is baxsed on Second-Order Blind Source Separation (SOBSS) algorithm. A major restriction of the standard SOBSS algorithm in the identification context is that the responses in all DOFs should be recorded. In the current research, this problem has been resolved by reducing the number of active modes in the measurements using the filtering technique so that the number of active modes is equal to the number of measurements. Then the standard SOBSS algorithm can be applied on the filtered responses to extract the modal parameters of the structure. The identification results of a simulated 5-story shear frxame for different noise levels showed that the noises in the measurements cause to increase the estimation errors of the damping ratios in the second proposed method (the maximum error of 23.5%). However, the natural frequencies and the mode shape vectors identified by this method have not been considerably affected by the noise level. Although in both cases of free or ambient vibrations, the proposed method can identify the active modes using only two measurements, but more accuracy can be achieved for poorly excited modes, by increasing the number of measured DOFs. The identification results of the experimental simulation indicate the disability of the second proposed method in estimation of the damping ratios of the poorly-excited modes (with maximum errors of 41%). Nevertheless, the estimation errors of the natural frequencies for all of the identified modes are less than 1% and the corresponding MAC values are approximately 100%.