On-line identification for low frequency oscillation needs to measure signals which couple with white Gaussian noise from wide area monitoring systems (WAMSs). Processed by low pass filter, white Gaussian noise can turn into colored Gaussian noise, so the accuracy of oscillatory indentification would be reduced. To solve the problem of colored Gaussian noise, in this paper we propose a cross-correlation-function (CCF) method that could reduce the influence of colored Gaussian noise. Combined with TLS-ESPRIT algorithm, CCF-TLS-ESPRIT could identify oscillatory modes in the environment of colored Gaussian noise rapidly. The simulation results show the effectiveness of the proposed method.