Abstract: An improved multimodulus blind equalization algorithm for QAM signal is presented. The real part (imaginary part) information is used to construct cost function. Moreover, conjugate gradient algorithm is proposed to optimize the equalizer coefficients. With its equalization performance comparable to the conventional Constant Modulus Algorithm (CMA) and MultiModulus Algorithm (MMA), the proposed algorithm is quadratic convergent and has the advantages of better convergent performance and less computational complexity. Finally, the reliability and validity of this algorithm are analyzed by Bit Error Rate (BER) and rate of convergence. Its good performance is verified by simulation results.