模仿人耳听觉频率特性用于改进语音识别的方法

THE IMITATION OF THE HUMAN HEARING FREQUENCE USING THE IMPROVED METHOD OF SPEECH RECOGNITION

针对现有计算机语音识别的缺陷，提出一种模仿人耳听觉特性的Ｍｅｌ频率倒频谱特征，对它与现有的线性预测倒频谱特征进行实验对比研究，表明将人耳听觉特性应用于计算机语音识别，能产生更高的识别精度。

The cepstral feature of Mel frequence imitating the human hearing is put forward owing tothe existing shortcoming of computer speech recognition. As compared with the linear predictioncepstral feature,it indicates that the higher recognition accuracy is produced when the humanhearing feature is applied to the computer speech recognition.