论文部分内容阅读
为分析拉萨方言语音差异性,通过采集语音样本,获取其语音特征向量,并将特征向量代入AP聚类算法,结果认为拉萨方言语音存在较大的个体差异,可将覆盖率较高聚类簇作为拉萨藏语方言规范语音.运用现代科学定量研究藏语方言语音特征,以较高能量的语音帧作为语音特征量,帧能量比作为语音特征片段的起止点,数据选取上具有一定的代表性,数据代入传统的AP聚类算法,进而分析程序的运算结果,实验过程一定程度上体现了定性与定量相结合的原则,验证了算法的收敛性和鲁棒性.
In order to analyze the speech difference of Lhasa dialects, the speech samples were collected and their speech eigenvectors were obtained. The eigenvectors were substituted into the AP clustering algorithm. The results showed that Lhasa dialects have larger individual differences, As the standard pronunciation of Lhasa Tibetan dialect.Using modern science to quantitatively study the speech characteristics of Tibetan dialects, the higher energy speech frames as speech features, the frame energy ratio as the starting and ending points of speech feature segments, the data selection has a certain representation , The data is substituted into the traditional AP clustering algorithm to analyze the operation results of the program. The experimental process reflects the principle of combining qualitative and quantitative to some extent, and verifies the convergence and robustness of the algorithm.