语音增强：理论与实践（第二版）

5星 · 超过95%的资源需积分: 49 107 浏览量更新于2024-07-22 7 收藏 17.51MB PDF 举报

"《语音增强：理论与实践》第二版由Philipos C. Loizou撰写，是一本深入探讨如何在噪声环境中提升语音清晰度和可理解性的书籍。该书全面介绍了语音增强算法，适用于移动设备、听力辅助设备等领域的应用。书中包含711页内容，涵盖了基础理论、主要的增强算法、性能评估方法以及未来发展方向。第二版新增了关于客观语音可理解性度量的新章节和改善语音可理解性的算法。附带的CD提供了MATLAB代码实现和实际噪声录音库，便于读者实践和评估算法。" 《语音增强：理论与实践》第二版详细阐述了以下几个关键知识点： 1. **语音增强基础**：作者首先介绍了数字信号处理和短时傅立叶分析的基本概念，这些是理解和设计语音增强算法的基础。 2. **语音生产与感知**：讲解了人类如何在噪声环境中补偿语音，帮助读者理解人类听觉系统的特性。 3. **噪声补偿**：详细讨论了人类听众如何在噪声中提高语音理解能力，为后续的算法设计提供了理论依据。 4. **增强算法**：涵盖了频谱减法算法、维纳滤波、统计模型方法、子空间算法等多种经典和现代的语音增强技术，以及噪声估计算法。 5. **性能评估**：介绍了评估语音增强方法效果的各种指标，包括质量与可理解性的客观度量，同时对比了不同算法的表现。 6. **MATLAB和C/C++代码**：随书附带的CD提供了实现这些算法的代码，使得读者可以直接应用到实践中。 7. **目标掩蔽算法**：专门讨论了在理想条件和现实条件下提高语音可理解性的二进制掩蔽算法。 8. **未来趋势**：探讨了进一步提升这些算法在实际环境中的潜力和可能步骤，为研究者和工程师指明了方向。本书不仅适合于对语音增强和噪声抑制感兴趣的音频处理工程师、研究人员，也适合作为相关专业研究生或高年级本科生的教学参考书。通过阅读此书，读者可以了解到当前最先进的语音增强技术和评估方法，以及如何将这些理论应用于实际的噪声环境。

xvContents

12.2.5 Comparisons in Reference to Noisy Speech ............593

12.2.6 Contribution of Speech and Noise Distortion

toJudgment of Overall Quality ................................ 597

12.2.7 Summary of Findings ...............................................598

12.3 Comparison of Enhancement Algorithms:

SpeechIntelligibility ............................................................. 598

12.3.1 Listening Tests: Procedure ....................................... 599

12.3.2 Intelligibility Evaluation: Results ............................. 600

12.3.3 Intelligibility Comparison among Algorithms .........602

12.3.4 Intelligibility Comparison against Noisy

Speech .............................................................. 603

12.3.5 Summary of Findings ............................................... 604

12.5 Summary ...............................................................................605

References ........................................................................................605

PART IV Future Steps

Chapter 13 Algorithms That CanImprove Speech Intelligibility ......................609

13.1 Reasons for the Absence of Intelligibility Improvement

with ExistingNoise-Reduction Algorithms ..........................609

13.1.1 Inuence of Speech Distortions ............................... 610

13.1.2 Lack of Effective SNR Increase ............................... 612

13.2 Algorithms Based on Channel Selection: ADifferent

Paradigm for Noise Reduction .............................................. 613

13.3 Channel-Selection Criteria .................................................... 619

13.3.1 SNR Criterion ...........................................................620

13.3.2 SNR

ESI

Selection Criterion ....................................... 623

13.3.3 Other Selection Criteria ............................................628

13.3.4 Channel-Selection Criteria for Reverberation .......... 629

13.3.5 Universal Selection Criterion for All Types

ofBackground Interferers ........................................ 631

13.4 Intelligibility Evaluation of Channel- Selection-Based

Algorithms: Ideal Conditions ................................................ 632

13.4.1 Broadband Noise Conditions .................................... 633

13.4.2 Competing-Talker Conditions ..................................634

13.4.3 Reverberant Conditions ............................................ 639

13.5 Implementation of Channel-Selection-Based Algorithms

in Realistic Conditions .......................................................... 639

13.5.1 Simple, But Ineffective, Algorithms for Binary

Mask Estimation .......................................................640

13.5.2 Effective Channel-Selection Algorithms Based

on Binary Classiers ................................................ 641

13.5.3 Adapting to New Noise Environments ..................... 645

xvii

Preface to the Second Edition

The second edition of this text not only revises the rst edition but also expands it.

In particular, it includes two new chapters. Chapter 11 provides a thorough cover-

age of objective intelligibility measures and Chapter 13 covers algorithms that can

improve speech intelligibility. The feedback I received from most readers about the

rst edition is that much of the focus of the text has been placed on algorithms that

can improve speech quality rather than speech intelligibility. The second edition

comes in response to those readers. With the proliferation of mobile devices and hear-

ing devices (hearing aids and cochlear implants), there is a growing and pressing need

to design algorithms that can improve speech intelligibility without sacricing qual-

ity. The inclusion of a chapter on intelligibility measures, which can predict reliably

speech intelligibility, is absolutely necessary in order to understand why the algo-

rithms described in Chapter 13 do improve speech intelligibility. Secondly, having a

good understanding of some of the commonly used intelligibility metrics can assist us

in the design of novel noise reduction algorithms that derive statistical estimators that

maximize/minimize such metrics. This stands in contrast with the conventional esti-

mators that aim to minimize the mean-square error (MSE), a metric that is “speech

ignorant” and not motivated by any known facts about how human listeners are able

to perceive speech in complex listening situations. It is the opinion of this author that

our obsession with the MSE metric delayed progress in the eld of noise reduction.

The contents of the DVD-ROM have also been updated to include (1)MATLAB

code with the implementation of some of the intelligibility measures described in

Chapter 11, (2) MATLAB code and C/C++ code with the implementation of the

algorithms described in Chapter 13, and (3) real-world noise recordings. In addition,

it includes the implementation of the wideband-version of the PESQ measure for

assessing the quality of speech sampled at rates higher than 8 kHz.

I wish to express my sincere thanks to all my graduate students who contributed

in some way to the writing of the second edition. In particular, I would like to

thank my postdoctorate students, Drs. Kamil Wójcicki, Jianfen Ma, and Fei Chen,

who wrote many of the MATLAB algorithms included in the updated DVD-ROM.

Many thanks also go to Oldooz Hazrati and Dr. Kostas Kokkinakis. Thanks also

go to Jacky Gao, Yi Gao, Li Xiao, and Fang Deng who, in the course of translat-

ing the rst edition to Chinese, found many errors and typos. I am also thankful

to Nora Konopka, editor at Taylor & Francis Group, for providing the support and

encouragement for this book.

Finally, I would like to express my deepest gratitude to my wife Demetria for her

understanding and undying support throughout this project.

Philipos C. Loizou

University of Texas at Dallas

Dallas, Texas

xix

Preface to the First Edition

This text is, in part, an outgrowth of my graduate course on speech signal processing,

which I have been teaching at the University of Texas at Dallas since the fall of 1999.

It is also, in part, a product of my own research in the area. The fact that no textbook

existed at the time on speech enhancement, other than a few edited books suitable for

the experts, made it difcult to teach the fundamental principles of speech enhance-

ment in a graduate-level course. It must be equally frustrating for new students or

speech scientists interested in getting into the eld of speech enhancement without

having access to a tutorial review or introductory paper (the last review paper was

published in the Proceedings of IEEE in 1979 by Lim and Oppenheim). That work

provided the initial motivation to write this book. My interest in this area stems from

my research to develop noise reduction algorithms that can be used to help hearing-

impaired listeners (cochlear implant listeners) better communicate in noisy environ-

ments.* Crucial to the development of such noise reduction algorithms is the basic

understanding of the limitations and potential of existing enhancement algorithms,

which I believe this book provides.

The textbook consists of 11 chapters, which are outlined in detail in Chapter 1

(Introduction). It is divided into three main parts. Part I presents the digital-signal

processing and speech-signal fundamentals needed to understand speech enhance-

ment algorithms. Part II presents the various classes of speech enhancement algo-

rithms proposed over the past two decades, and Part III presents the methods and

measures used to evaluate the performance of speech enhancement algorithms.

The text body is supplemented with examples and gures designed to help the

reader understand the theory. The book is accompanied by a DVD-ROM, which

contains a speech corpus appropriate for quality and intelligibility evaluation of

processed speech, and MATLAB

code with the implementation of major speech

enhancement algorithms. It is my strong belief that having access to MATLAB code

and a common speech database against which to evaluate new speech enhancement

algorithms is crucial and necessary in order to move the eld forward. Appendix C

provides a detailed description of the contents of the DVD-ROM.

The book can be used as a textbook for a one-semester graduate-level course on

speech enhancement. Necessary prerequisites for such a course would be a course

on digital signal processing and fundamental knowledge of probability theory, ran-

dom variables, and linear algebra. This book can also be used as a supplement to an

introductory course on speech processing. In this case, Chapters 4 through 8 could

be covered along with a select set of sections from Chapters 9 and 10.

I wish to express my sincere thanks to the many colleagues and graduate stu-

dents who contributed in some way to the writing of this book. I would like to

thank Professors Patrick Wolfe, Kuldip Paliwal, Peter Assmann, John Hansen, and

This work is supported by the National Institutes on Deafness and Other Communication Disorders,

NIH.

剩余704页未读，继续阅读

ramissue

粉丝: 354
资源: 1487

语音增强：理论与实践（第二版）

Speech Enhancement Theory and Practice（语音增强-理论与实践）DVD数据

speech enhancement 语音增强

语音增强理论与实践

语音编码相关的参考书籍

将这个代码做成gui界面，分别有打开语音，语音增强以按钮

语音降噪联合语音识别训练的相关主流文献和开源代码有哪些

生成基于模糊集理论的图像增强算法的MATLAB代码

J. K. Kim, S. H. Park, "Adaptive noise cancellation using a Kalman filter for speech enhancement," IEEE Transactions on Consumer Electronics, vol. 47, no. 3, pp. 564-570, 2001.概括文献内容

Fine-Grained Feature Enhancement for Object Detection in Remote Sensing Images

最新资源