摘要
先讨论由语音信号反演声门激励函数,发现波沿非均匀声管传导既能产生新的频率(共振作用),也可抑制原有频率(滤波作用).这说明目前用于语音处理基础的全极点LPC模型并不能充分反映这些特征.然后研究如何直接从语音生成的偏微分方程模型出发,用反演方法提取一种新的语音特征参数代替LPC作为研究各种语音处理的基础.
The inversion of glottis excitation from measuremnts of sound presure at lip is studied;showed that acoustic wave propagating through nonhomogeneous tracts not only yields,but also filters some frequencies,and LPC parameters based on only extreme points model can not characterize enough various voices.A new system of feature parameters are abstracted directly from partial differential equations generating voice.These parameters instead of LPC can be a basis for following study of speech signal processing.A fast algorithm for computing these parameters is given.
出处
《山东大学学报(自然科学版)》
CAS
CSCD
1998年第1期44-51,共8页
Journal of Shandong University(Natural Science Edition)
基金
国家自然科学基金
关键词
语音信号处理
语音模型
数理方程反演
语音特征
speach signal processing
speech model
math-physics equation inversion
speech charaiteristic