This paper presents a design method for multi-input single-output nonlinear adaptive digital filters using neural networks. By applying neural networks in the field of adaptive signal processing, we propose in this paper a design method for parallel nonlinear adaptive digital filters using back-propagation training of parallel connected several small-scale feedforward multilayered neural networks, in order to design the nonlinear adaptive digital filter required for real time processing with many parameters. Furthermore, in comparison with this method and the method based on conventional linear theory, if proposed method is used, a good result can be obtained. and, it is possible that the learning efficiency is improved, because the parallel learning is carried out in this method. Finally, the results of computer simulation are presented to illustrate the effectiveness of the proposed method.