Lately a great deal of research in the field of identification of the autoregressive with exogenous noise (ARX) model has been conducted. In general, these methods are considered by the time-domain estimate or the frequency-domain estimate. Combining the time-domain estimate and the frequency-domain estimate to identify the ARX model interfered by noise is discussed in this paper. The concept of general prediction error (GPE) criterion is introduced for the time-domain estimate with optimal frequency estimation introduced for the frequency-domain estimate. A new identification method, which is called the empirical frequency-domain optimal parameter (EFOP) estimate, is proposed for the ARX models interfered by noise. The algorithm theoretically provides the globally optimum frequency-domain estimate of the model. Some simulations are included to illustrate the new identification method.