顾文涛编著的《语音韵律的实验分析与建模》语音韵律在言语交际中不仅传递语法语义等信息，而且传递副语言信息以及说话人的非语言信息。语音韵律不仅是语言学的重要课题，而且有效的韵律信息处理方法对语音技术系统性能的提升十分关键。《语音韵律的实验分析与建模》选自作者近年来关于语音韵律研究的论文，系统考察了字调、声调协同发音、时长与节奏、句调、韵律结构、焦点重音、情感表达等各个层面的韵律特征，特别突出了定量建模的研究方法。研究对象以普通话和粤语为主，涉及多种汉语方言，包含跨语言对比及语言接触的研究。

语音韵律在言语交际中不仅传递了语法、语义等信息，而且传递了副语言信息以及说话人的特征信息。不仅语音韵律是语言学研究的重要课题，而且韵律信息处理也是语音信息工程中的关键技术。顾文涛编著的《语音韵律的实验分析与建模》节选了作者关于语音韵律研究的部分论文，系统考察了时长、字调、声调协同、句调、韵律结构、焦点重音、情感表达、说话人风格等各个层面的韵律特征，特别突出了定量建模的研究方法。《语音韵律的实验分析与建模》研究对象以普通话和粤语为主，涉及多种汉语方言，同时包括跨语言对比及语言接触的研究。

PART I SPEAKER ADAPTATION FOR DURTION MODEL IN MANDARIN TEXT-TO-SPEECH SYNTHESIS

 1 Introduction

 1.1 Introduction to Duration Modeling in TTS Systems

 1.1.1 Text-to-Speech Synthesis and Segmental Duration

 1.1.2 Duration Model

 1.2 Speaker Adaptation for Duration Model--Goal and Basic

 Assumption

 1.3 The Source Model for Mandarin Duration

 1.3.1 Phone Categorization

 1.3.2 Multiplicative Model

 1.3.3 Duration Factors

 Model-Based Optimal Text Selection

 2.1 Introduction

 2.2 Coverage and Statistical Model

 2.3 Model-Based Greedy Text Selection

 2.3.1 Analysis-of-Variance Model

 2.3.2 Design Matrix and Parameter Estimability

 2.3.3 Matroid Cover Problem

 2.3.4 Model-Based Greedy Algorithm

 2.4 Multi-Model Based Greedy Algorithm

 2.4.1 Modified Algorithm for Multi-Model Cases

 2.4.2 Experimental Result

 2.4.3 Analysis of Computational Complexity

 2.5 Further Generalization of the Algorithm

 2.6 Experimental Result and Discussion

 2.7 Conclusion

 3 Speech Data

 3.1 Speech Recording

 3.2 Segmentation and Labeling

 3.3 Data Analysis

 4 Analysis of Multi-Speaker Mandarin Duration Models

 4.1 Statistical Analysis

 4.2 Muhiplicative Model Fitting

 4.3 Effects of Factors in Duration Models

 4.3.1 Vowel

 4.3.2 Plosive Burst and Aspiration

 4.3.3 Plosive Closure

 4.3.4 Nasal Coda

 4.3.5 Fricative

 4.3.6 Sonorant Consonant

 4.3.7 Common Effects across Phone Categories

 4.4 Compensatory Effects

 4.4.1 Burst/Aspiration and Closure of Plosives

 4.4.2 Vowel and Nasal Coda

 4.4.3 Obstruent and Vowel

 4.4.4 Obstruent and Glide

 4.4.5 Syllabic Compensatory Effects

 4.5 Syllable Duration

 5 Speaker Adaptation for Duration Modeling

 5.1 An Efficient Speaker Adaptation Model

 5.1.1 Target Model Assumption

 5.1.2 Validity of Scalable Hypothesis

 5.1.3 Theoretic Analysis of Model Estimation

 5.1.4 Model Fitting by Linear Regression

 5.1.5 Sentence Effect on Model Estimation

 5.1.6 Analysis of Model Robustness

PART II Quantitative Analysis and Modeling of Tonal and Intonational Variations on Various Layers

References