There is rich information of phase in MPSK. Different wavelet coefficient peak collection probability density function and different maximum likelihood function of different MPSK can be obtained from its phase information. These parameters can be used to do modulation recognition for MPSK. In this paper, Morlet wavelet is used to extract the different probability density function and maximum likelihood function from three modulation signals which are BPSK, QPSK, 8PSK. The value of M is got by mode recognition. Simulation shows that the method is perfect for modulation recognition under the condition that SNR is low (greater than or equal to 4dB).