基于语谱能量的音素边界检测  

A phone classification method based on spectrum

作　　者：李立永[1] 张连海[1] 冯志远[1] 

机构地区：[1]信息工程大学信息系统工程学院,河南郑州450002

出　　处：《太赫兹科学与电子信息学报》2013年第6期936-941,共6页Journal of Terahertz Science and Electronic Information Technology

基　　金：国家自然科学基金资助项目(61175017)

摘　　要：根据音素发音时语谱结构的变化提出了一种基于语谱能量的音素边界检测方法。该方法首先根据语谱结构变化特点将信号频域划分为高频、中频、低频3个区域,并以语音帧间语谱能量向量的欧氏距离为判别依据分别对3个区域进行音素边界检测,然后对3个区域检测的边界分别进行二次筛选,最后将3个区域的边界信息融合,得到音素边界检测结果,相对于基于音素属性的边界检测方法,计算复杂性大大降低,边界检测率提高了3.95%。This article proposes a phone boundary detection method based on the spectrum. The spectrum is divided into three frequency regions according to the structure of the spectrum. The potential boundaries of every region are detected through the Euclidean distance between the neighbors of two vector frames of spectrums. The false phone boundaries are removed by a second examination. The phone boundaries are obtained via the fusion of the boundaries of three regions. Comparing with the method using phonological attributes, the computational complexity is significantly reduced, and a higher score of 3.95% is achieved.

关 键 词：音素边界检测 语谱能量 语音事件 

分 类 号：TN912.34[电子电信—通信与信息系统]