前馈神经网已经被大量用于非线性信号处理 .经典反向传播算法是一种标准的前馈网络学习算法 ,但是 ,对许多应用 ,反向传播算法的收敛速度却很慢 .本文根据对网络的非线性单元进行线性化而提出一种新的算法 ,该算法在非线性信号处理中在精度和收敛速度方面都优于传统的反向传播算法 . Feedforward neural networks have been extensively used for nonlinear signal processing. The classical backpropagation algorithm is a standard feedforward network learning algorithm, but for many applications the backpropagation algorithm converges very slowly.In this paper, This paper presents a new algorithm for linearization of nonlinear elements in network, which is superior to the traditional back propagation algorithm in accuracy and convergence speed in nonlinear signal processing.