<p>Modern measurement systems combined with measurement-based methods enable modal identification to monitor power system stability and reliability. However, less attention has been given to ambient mode shape estimation techniques. In this paper, a complex blind source separation (CBSS) technique is proposed, integrating the Hilbert transform (HT) with the second-order blind identification (SOBI) algorithm to estimate the dynamic behavior (oscillation components and mode shapes) from ambient data. The HT provides the complex representation, including phase information, and the SOBI algorithm extracts the complex mixing matrix and source signals based on their temporal structure. The proposed method is applied to simulated ambient data from the 16-generator test system to identify dominant oscillatory components and their related electromechanical mode shapes. The results demonstrate that the extracted complex mixing vectors exhibit dynamic patterns closely matching the theoretical mode shapes, as indicated by the modal assurance criterion. Finally, the spectral method is employed to estimate mode shapes, showing good agreement with the obtained results.</p>