基于数据驱动的语音库的分析与构建  

The Structure and Analysis of Speech Sound Bank Based on Data Driving

作　　者：朱玉玺[1] 胡学骏[1] 赵慧民[1] 

机构地区：[1]佛山大学计算机科学系,广东佛山528000

出　　处：《电声技术》2006年第5期50-52,共3页Audio Engineering

摘　　要：语音数据库和规则库是转换系统中最基本和重要的部分,在进行转换时,根据韵律建模的结果,从原始语音库中取出相应的语音基元,利用特定的语音合成技术对基元进行韵律调整合成。针对语音数据库和规则库的建立技术,在GPS嵌入式系统的语音库构建过程中,在与传统方法比较的基础上,给出了基于数据驱动的构建方法,并在特定平台RAM10上,对“文-语”转换技术中的瓶颈之一进行了探讨和实践。The speech sound bank and the regulation bank are the most essential and important parts in the translation system. According to the results of cadence modeling, the speech basic elements are taken from the original data pool and then are regulated by specifical speech synthesizng system. The approaches to configuration are given on the base of data driving. Finally, on the special platform RAMIO, one of the bottle-necks in the ＂language-speech＂ translating technology is discussed and practised.

关 键 词：语音数据库 嵌入式系统 计算机信息处理 语音合成 PSOLA算法 

分 类 号：TP392[自动化与计算机技术—计算机应用技术]