【作者】 杨锋；

【导师】 陈海伦；

【作者基本信息】 广西大学 ， 汉语言文字学， 2008， 硕士

【摘要】 本文通过对标准壮语1863个单音节和200组双音节的声学分析,根据舒声韵和促声韵的不同,调类和声母清浊的不同,对基频曲线进行自动分段和统计分析,建立了标准壮语单音节、双音节基频曲线模型,并进行了初步的语音合成实验。文中提出一种新的自动去“除弯头降尾”和对基频进行自动分段描述的方法,并对清、浊辅音声母和塞尾对基频的影响情况进行了深入的分析。本文得出以下四个方面的结论:一、用多项式曲线拟合和求零点的方法能够有效的去除“弯头降尾”。二、根据零点的个数和位置能够对基频曲线进行自动分段,同时可以计算出每一段的时长、频率、斜率等参数,用这些参数能够对壮语基频曲线变化进行准确的描述。三、清、浊辅音声母对基频产生一定的影响,浊辅音对基频有明显的降低作用。清、浊辅音声母段具有固定的基频起始值,逐渐向后面的元音段基频目标值过渡,在不同的情况下表现为上升或下降,也就是对基频的提高和降低。塞尾对基频曲线有明显的降低作用,降低幅度的大小与元音段基频值和塞尾目标基频值的差值大小有关,高平调55调降低的幅度最大。四、通过初步的语音合成实验证明,本文所建立的标准壮语单音节、双音节基频曲线模型是有效的、可行的。更多还原

【Abstract】 This paprer is concerned with the study on F0 contour model of monosyllables and disyllables in Standard Zhuang Language. Based on the acoustic analysis of 1863 Monosyllables and 200 disyllables in Standard Zhuang Language, according to the different categories of Shushengyun and Cushengyun, the different tone, the voiceless-initial and voiced-initial, a F0 contour database for monosyllables and disyllables is built, based on the statistical analysis of F0 contour. Otherwise, this paper introduces a new method of removing the "offset-section and onset-section" and an automatic description method to describe the dynamic variation of the F0 contour.In this paper, four conclusions have been drawn : First, the segmentation of "offset-section and onset-section" can be removed effectively depending on the number and the positions of extreme points in the polynomial fitting expression of the F0 contour. Second, according to the number and the positions of extreme points, parameters of the length, frequency and slope can be calculated for each section, these parameters can be used in the dynamic F0 contour description. Third, the voiceless-initial and voiced-initial have a certain impact on the F0 contour, the voiced-initial have a lowering effect on F0. the voiceless-initial and voiced-initial have the fixed-frequency Starting values, and gradually slide to the vowel-frequency target value, in different circumstances to increase or decrease performance, which is based on the frequency of raising and lowering.Ending stop has a lowering effect obviously on F0. Four, the experiments of parameter synthesis show that this F0 contour model can be used to synthesize monosyllables and disyllables effectively.更多还原