论文部分内容阅读
本文首先描述一种神经网络的拓扑结构——多层感知器神经网.其次推导出针对该神经网的整体学习算法,给出算法在一般微机上运行时所遇到的一些具体问题处理原则.最后将双隐层感知器用于语声识别.实验表明:多层感知器神经网络技术用于小词汇量、低信噪比、机载条件语声识别系统,是一种有益的尝试.
This paper first describes a neural network topology - multilayer perceptron neural network.Secondly, the overall learning algorithm for this neural network is deduced, and the principle of some specific problems encountered when the algorithm runs on the general microcomputer is given. Finally, the double-hidden layer perceptrons are used for speech recognition.The experiments show that the multi-layer perceptron neural network technology is a useful attempt for small vocabulary, low signal to noise ratio and airborne conditional voice recognition system.