In the radar passive signal sorting, the extraction of the pulse repetition interval (PRI) value is critical. Aiming at the shortcomings of histogram evaluation and PRI transform method in the accuracy of pulse repetition period, a PRI value extraction method based on density clustering is proposed by this paper. By making three pulse arrival time differences, subjected to density-based clustering process, and then select the category as the mean PRI estimate. The simulation results show that the PRI value extracted by this method is significantly higher than that of the traditional method in terms of accuracy. This method has a better sorting effect on the radar signals with error jitter and spurious pulses, and improves the ability of sorting of pulsed signals of passive radar in complex environment.