In order to recognize complicated PRI modulation modes, a classify algorithm based on autocorrelation and normalization is proposed. The algorithm defines three characteristic quantities, i.e. density of peak value, intension of monotone and energy of sequence, and the paper sets up feature space after making autocorrelation and normalization to four PRI impulse sequences. By comparing the characteristic quality with the threshold value, PRI modulation mode can recognize automatically. The result shows that the recognition accuracy can reach to 95% when the numbers of pulse are 100 and the SNR is 4 dB, and the recognition accuracy can reach 95.5% when the numbers of pulse are only 80. The algorithm can maintain more than 93% of accuracy when the ratio of the pulselosing reaches 25% and the numbers of impulse are 100.