Low frequency oscillation modal analysis provides the most basic elements of information for the safe and stable operation of power grid. To identify low frequency oscillation modal parameters based on ambient excited data from a wide area monitoring system (PMU), this paper investigates the use of the natural excitation technique (NExT) in conjunction with the eigensystem realization algorithm (ERA) for the modal analysis of power systems. At the same time, it introduces the fuzzy C clustering algorithm to automatically pick up the results, to better identify the authenticity of a modal. The method is capable of utilizing synchronous measured data from WAMS as well as unsynchronous measurements by the truncation approach. By modal identification of system for the IEEE four machine and IEEE sixteen machine simulation signals to verify the validity and high efficiency of this method to extract the dominant mode, which can also meet the needs of online and offline applications. This work is supported by National Natural Science Foundation of China (No. U1134205) and China Railway Corporation Major S&T Projects (No. 2015J005-A).