基于语音参数的神经网络识别研究

Research on the Application of Neural Networks for Speech Recognition

本文通过对BP和SOM神经网络的理论学习研究,分别设计了基于语音参数为输入样本的BP三层和SOM两层网络模型及相应算法实现,并通过对比两种神经网络模型在相同输入语音样本参数情况下的不同运行机理,最终确认基于LPCC语音参数的BP网络更适合应用于语音识别。

From the theoretical study and research of BP and SOM neural network, this paper proposed BP and SOM neural network models and algorithms based on the input voice sample parameters and then made a conclusion that BP network is more suitable for voice recognition by comparing the two neural network models running under a different mechanism with the same input sample parameters.