介绍一种新颖的非线性自适应滤波器——自适应神经网络滤波器。由于神经网络具有学习非线性函数到任意的精度以及自适应能力，这种滤波器优于线性滤波器，能适应各种噪声环境。在自适应ＬＭＳ算法基础上，提出了在线ＢＰ训练算法、收敛速度快。最后以自适应噪声对消系统为例，进行了计算机仿真，结果显示了这种滤波器的良好性能 A novel nonlinear adaptive filter - adaptive neural network filter is introduced. Since neural networks have the ability to learn nonlinear functions to any degree of accuracy and self-adaptability, this filter outperforms linear filters and can adapt to a variety of noisy environments. Based on the adaptive LMS algorithm, an online BP training algorithm is proposed, which has fast convergence speed. Finally, taking adaptive noise canceling system as an example, a computer simulation is carried out and the result shows the good performance of this filter