Modal analysis consists in the characterization of a system through its modal parameters. When the main excitation source is the system s environment, this characterization is defined as operational modal analysis (OMA). On those cases, the forces are unknown (not measured) and only the responses are monitored. Because of there random nature, those signals are incorporated into the numerical model as stochastic processes. The main goal of this dissertation is to describe the identification techniques in OMA. Therefore, two different approaches were created: a theoretical one and an experimental one. In the theoretical part, the required hypotheses for system s identification with OMA are presented. An analysis of the errors caused by noisy signals are also performed, allowing the method s sensibility to be evaluated. Besides the standard identification methods, two new ones are proposed. They both has been developed as extension of the Proper Orthogonal Decomposition (POD) and combine an efficient computational process with the possibility of quantify the parameters uncertainties. In the experimental approach, the goal is to illustrate and validate the identification of structures. Therefore, three different test bench were created: a two floor building, a wind turbine blade and a cable-stayed bridge. After their construction, those structures were proper instrumented with different sensors. A data acquisition system were built using commercial hardwares and analyzed through a graphic interface specially made for OMA and developed in the vibration laboratory.