论文部分内容阅读
根据连续语音中不同类语音段之间内在的相异性 ,提出了段间距离的概念 ,并基于特征矢量距离给出了段间距离的一种测度 .在此基础上 ,提出一种新的语音自动分割方法 ,它根据语音段间距离曲线的峰值点对语音音素进行分割 ,而不需要关于语音单元的先验知识或参考模型 .在该方法中 ,语音特征矢量按帧计算 .通过采用适当的帧重迭度 ,可以使语音分割具有较高的时间分辨率 .实验结果表明 ,该方法具有较高的分割精度和速度
Based on the inherent dissimilarity between different speech segments in continuous speech, this paper proposes the concept of the distance between segments and gives a measure of the distance between segments based on the distance of eigenvectors. On this basis, a new speech Automatic segmentation method that divides the speech phoneme from the peak points of the distance curve between speech segments without the prior knowledge or reference model about the speech unit In this method the speech feature vector is calculated on a frame basis by using appropriate Frame overlap degree can make speech segmentation with high temporal resolution.The experimental results show that this method has higher segmentation accuracy and speed