摘要
本文研究运用矢量码书和动态内插限制的方法解决语音生成逆向解的非唯一性问题 .表征声道的对数截面积函数的限带傅立叶余弦展开包括了偶次项 ,采用可变声道长度 ;频域声学目标增加了前N个零点频率 ;结合腔包形声学———几何形态映射矢量码书 ,为声学目标匹配得到过渡音的起始和结尾端点的零点与声道长度 ,并用内插方法来限定过渡点的零点频率和声道长度 ,从而解决非唯一性问题 .
This paper presents a method of using codebook and dynamic interpolation constraint to solve the inverse speech problem.The vocal tract is modeled by coefficients of band limited Fourier cosine expansion of logarithm area function,together with a variable vocal tract length.The lowest N zero frequencies are added to the acoustic target information.By using a root cell acoustic to geometric mapping codebook,the zeros and vocal tract lengths of the endpoint of vowel to vowel transition are determined.Their transitions are then constrained by interpolation method.The result of computer simulation shows effectiveness of the method.
出处
《电子学报》
EI
CAS
CSCD
北大核心
2000年第4期52-56,共5页
Acta Electronica Sinica
基金
国家自然科学基金!(No.69972 0 4 6)资助课题
浙江省自然科学基金!(No.6980 76)
关键词
语音生成逆向解
动态内插限制
矢量码书
inverse speech solution
speech analysis and synthesis
speech processing