摘要
论文对适合唇读研究的连续音节双模态语料库及其语料切分算法的设计和研究工作进行了讨论。介绍了基于句子级的双模态语料库HITBi-CAVDatabaseII的设计和建立,形式化地讨论了该库的主要特点及基于语音能量的语料切分算法的可行性。该切分算法在基于能量的语音切分算法基础上,结合了双模态语料库的一些特征,实现了对语料的自动切分。
The topic of this paper is about the design of Bimodal Database for continuous Lip-Reading and the research on its material segmentation.First,it describes the design and foundation of a new Bimodal Database HIT Bi-CAVDatabaseII which is for Lip-Reading on sentence.Then,its key characters and the feasibility of the material segmentation algorithm based on the speech energy are analyzed formally.This segmentation algorithm combines the speech segmentation approach based on energy with the characters of the database.Now,the automatic segmentation in the Bimodal Database can be realized.
出处
《计算机工程与应用》
CSCD
北大核心
2005年第3期174-177,190,共5页
Computer Engineering and Applications
基金
国家863高技术研究发展计划(编号:2001AA114160)
哈尔滨工业大学校基金(编号:HIT2002.72)资助
关键词
唇读
双模态语料库
语料切分
Lip-Reading,Bimodal Database,material segmentation
