提出一种新的非高斯信号激励的非最小相位ＭＡ模型参数辨识方法。该方法基于三阶矩和自适应算法，并利用ＦＫＡ算法极小化输出信号的三阶矩估计方差。计算机模拟结果表明，木文方法的收敛速度比最大梯度下降法（ＮＰＧ）快，与ＮＰＯ算法同数量级，但每次递归的运算量比ＮＰＯ算法少，占内存量也小。 A new non-minimum phase MA model parameter identification method based on non-Gaussian signal excitation is proposed. The method is based on the third moment and the adaptive algorithm, and uses the FKA algorithm to minimize the variance of the third moment of the output signal. Computer simulation results show that the convergence speed of the woody method is faster than the maximum gradient descent method (NPG) and same as that of the NPO algorithm, but each recursive computation consumes less than the NPO algorithm and accounts for a small amount of memory.