Abstract: For the purpose of the technology of modulation recognition of MPSK and MQAM modulation in burst adaptive modulation signals, a new blind identification algorithm is proposed. This algorithm can recognize eight kinds of modulation signals, including BPSK signals, QPSK signals,8PSK signals, 16QAM signals, 32QAM signals, 64QAM signals, 128QAM signals and 256QAM signals. Firstly the cyclostationarity of the signals is analyzed and discussed and the theoretical basis of the signal identification based on the features of cycle High-order Cumulation is given. Secondly three features are proposed based on cyclic cumulation of the signals to classify PSK signals and QAM signals, MPSK signals, square QAM signals and cross QAM signals respectively. Finally, through research and analysis of the instantaneous amplitude distribution of MQAM signals, a new feature based on the variance of instantaneous envelope square is proposed to classify the MQAM signals. The algorithm uses the binary tree support vector machine as classifier and designs a new identification process to classify the signals mentioned above. It does not require precise synchronization and is robust to the carrier phase. Moreover, it is suitable for the IF signals recognition. Simulation results show that the algorithm can achieve identification of burst signals under low SNR.