论文部分内容阅读
本文在微机上实现了语音特征参数的提取以及用分级结构神经网络进行识别的计算机模拟实验,改进了通用的神经网络反向传播(BP)学习算法,对BP算法的收敛过程进行了深入的分析并提出了一些有助于神经网络快速、平稳收敛的具体措施。
In this paper, the extraction of speech feature parameters and the computer simulation experiments using hierarchical neural network are implemented on the computer, and the general BP neural network back propagation (BP) learning algorithm is improved. The convergence of BP algorithm is analyzed in depth And put forward some concrete measures that contribute to the rapid and smooth convergence of neural network.