In order to solve the problem of low rate in radar emitter signal recognition under low SNR, a new approach based on the feature of wavelet ridge for recognition of intra-pulse modulation was proposed. The proposed modified Morlet wavelet atom was used to extract the instantaneous frequency. Then the instantaneous frequency was transformed to extract the characteristic vector for classification. The results of computer simulation showed that the characteristic vector of this method had good recognition ability, and the average recognition rate could reach 90% or more than 90% when the SNR was as low as 2 dB. The superiority of the proposed algorithm under low SNR was proved by simulation results compared with the existing approach.