语音识别的神经网络方法研究

An artificial neural network model study of speech-identifying

用前馈多层神经网络方法研究了计算机对于不同语音输入者的识别能力 .输入层为语音的平均频率、均方频率、频率宽度、平均振幅、均方振幅、振幅宽度 ;输出层为识别输入者 ;隐含层的节点数为 8.最后预测结果成功率为 82 .5 % .

This paper investigated speech-identifying abilities by using BP artificial neural network model.In this artificial neural network model,the input nodes were average frequency,mean-square frequency,breadth of frequency,average swing,mean-square swing,breadth of swing.The output node were speech,and the number of hidden node was 8.The accurate ratio was 82.5%.Artificial neural network model may provide a method to investigate speech-identifying.