中文摘要：

针对动态信号模式分类问题,提出了一种反馈过程神经元网络模型和基于该模型的分类方法。这种网络的输入可直接为时变函数,网络的信息传输既有与前馈神经元网络一样的前向流,也有后面各层节点到前层节点的反馈,且可对节点自身反馈输出信息,能直接用于动态信号的模式分类。由于反馈过程神经元网络在对输入样本的学习中增加了神经元输出信息的反馈,可提高网络的学习效率和稳定性。给出了具体学习算法,以时变函数样本集的分类问题为例,实验结果验证了模型和算法的有效性。

英文摘要：

To solve the classification of dynamic signal, this paper proposed a feedback process neural networks model and classification methods based on this model. The time-varying function could be directly used as input of this network. In addition to the existence of the feed-forward information flow like a normal neural network, there still existed the feedback information flow from output to input in this model, and the nodes could also form a self-feedback. The network could be directly used into pattern classification of dynamic signals. Improved the efficiency and stability of the network evidently with application of the feedback information from the neurons in output layer in the learning process of the feedback process neural networks. Taking the classification of time-varying functions as an example, the experimental results show that the model and the algorithm are efficient.