语言发音模型研究综述

Overview of speech production models

对各种语言发音模型进行了综述,分别讨论了言语声音模型和言语动作模型。言语声音模型研究语言发音的声学原理,利用声音信号处理技术重构语音信号波形,由于对声源和共鸣之间的关系的认识不同,以及对共鸣的分析方法的不同,产生了3种不同的语言发音模型,第一种是频谱分析模型,第二种是共振峰模型,第三种是生理发音模型。言语动作模型研究发音器官的运动过程,利用图像信号处理技术重构发音器官的发音动作,根据建模方法的不同,言语动作模型可以分为3类:生理机能模型、几何特征模型、统计参数模型。

This paper studies all kinds of speech production models including speech sound models and speech gesture models. Speech sound models deal with acoustic theory of speech production and reconstruct speech waveform by audio signal processing techniques. Owing to different understanding of the relationship between source and resonator, and different method of resonation analysis, there exist three different speech sound models：Spectrum analysis model, formant model and articulatory model. Speech gesture models focus on the physiological process of speech production, and rebuild speech organ gestures by visual signal processing techniques. According to different method of modeling, there are three speech gesture models：Physiological mechanism model, geometrical feature model and statistical parameter model.