Abstract: The paper introduces a new dual-mode blind equalization algorithm for the MPSK/MQAM signals without modulation types identified under the time-varying channel. At the beginning, the constant modulus algorithm (CMA) is adopted in equalization for its robust performance. After a while, we get the cluster values of the equalized signal by subtraction clustering method. Notice that nearly all of the MPSK/MQAM signals' constellations have geometrical rotary symmetry, so the fuzzy c-means algorithm is introduced to modify the former cluster values for the signals' constellation rebuilding. Therefore we can get the signal's precise alphabet information from the modified constellation. Finally, the equalization switches to alphabet-matched algorithm (AMA) to obtain better performance, what's more the switching time depends on the cluster validity index. The new dual-mode algorithm can effectively overcome the difficulty that the modulation types are unknown. Compared with the existing methods, this new algorithm converges fast with few inter symbol interference and enable equalization switch smoothly from CMA to AMA. Simulation results have proved the efficiency of the proposed algorithm for 4PSK and 16QAM inputs.