Abstract: To solve the problem of limited convergence in Least Square (LS) estimation method, an improved scheme is proposed. At first, the scheme uses the idea of matrix diagonalization to improve the Minimum Mean Square Error (MMSE) estimation method, making it to reduce greatly the complexity while maintain the estimate convergence. Then, a new scheme is proposed according to the adaptive theory, the estimate scheme can select the corresponding algorithm to estimate adaptively according to the state of current channel, which shows its high flexibility. The simulation result shows that using this scheme can further improve the Bit Error Rate (BER) and Mean Square Error (MSE) with low computational complexity, which is promising for practical applications.