Abstract: Combining the advantage of order analysis and blind identification of time series model, a new blind identification algorithm of AR model on order domain is proposed. The proposed algorithm includes two parts: (1) Preprocessing, the non-stationary signal is resampled by order sampling, i.e. constant angular interval, and transformed into stationary signal at the angular domain. (2) Blind identification, the stationary signal at the angular domain is identified by the blind identification of AR model. The proposed algorithm not only inherits the advantages of traditional blind identification of AR model, but also overcomes the deficiency of raditional blind identification of AR model. The simulation result shows that the proposed algorithm is very effective, and can be a powerful tool for processing non-stationary signal.