Identifying of dynamic parameters such as modes, damping and stiffness is a key issue to predict the stability of dynamic system. There are two groups of techniques to find out the dynamic parameters: experimental modal analysis (EMA) and analytical approach. EMA is generally done by impact test or a shaker test that calculate frequency response function or impulse response function from measurements of both input excitations and corresponding responses. If input excitations are unknown, the modal parameters can be obtained by operational modal analysis (OMA). At present work, OMA is carried out to estimate natural frequencies of a dynamic system in milling. Responses (vibrational accelerations) are stored via three-axial accelerometer. Furthermore, power spectral density matrix of output response is estimated. Using singular value decomposition method the dependence of singular values from frequencies is obtained and represented as a graph. Modal parameters are estimated via peak picking method. Previously, R. Brinker suggested this approach at first time in 2000 as Frequency domain decomposition method. To check and demonstrate approach described above, simulation model of milling had been created and analyzed. After, output-only modal analysis was adopted to real system that includes milling machine tool, workpiece, fixture and cutting tool. Three-axial accelerometer is mounted on the base of spindle. A number of cutting test have been done and output responses is stored on PC. After response acquisition, OMA is done and natural frequencies are obtained using peak piking method. Results are presented in article as well.