Customer Reviews

Discrete-Time Processing of Speech Signals (Ieee Press Classic Reissue)

5 star:		(4)
4 star:		(1)
3 star:		(1)
2 star:		(1)
1 star:		(0)

 

 

I've seen a lot of books on Speech processing, but this one's the best by quite some margin. Rabiner & Schafer's classic could've edged this one out, but it's slightly dated now(1978 was a long time ago!) - it remains a great source for the fundamentals, though. However, this book is uptodate, as well as quite exhaustive in its coverage of the basics. One can get a solid grasp of basic speech processing(eg pitch analysis), synthesis, compression and recognition. The chapters on recognition, especially, are quite impressive,with detailed treatment of all the standard techniques in the field,including Dynamic time warping,Hidden Markov models & Neural networks. Everything from isolated word recognition to large vocabulary continuous speech recognition is touched upon.

This book is an ideal starting point for exploring the world of speech processing,which is a very rich area of research and has virtually limitless commercial applications,too.

A word of caution though - this is not for the layman.One requires a bit of a background in mathematics,and a little knowledge of digital signal processing won't hurt either.

I have only nice things to say about this book.The only possible crib could be the hefty price tag,but then this is typical of books in this field.

All in all,it's an excellent textbook,which ought to be read by everyone working in speech processing.

This book covers many aspects of speech signal processing. In doing so, it is a great conceptual primer, but people with specific interests in one aspect or other of speech signal processing may look into other books. I have found it very valuable in clarifying certain concepts and an essential reference for my thesis. Some readers may find this book frustrating in some aspects because they are not familiar with the background of certain topics. It is often best to reread the conceptual portion of those chapters and look for other resources that explain the details of those topics in a more tangible language.

This is an attempt to cover many aspects of speech processing, including modeling, coding, synthesis, enhancement, and recognition. Since each of the mentioned areas is a thick book on its own, the present book does not and cannot provide in depth coverage. I like some stories told in the book, like the development of the mechanical vocoder. But I find it to be wordy and in many instances, somehow impractical, like the exagerated mathematical treatment of the acoustic tube model. Also, the mathematical notations tend to be awkward. Due to technological advances, many topics contained in this book are already outdated.

This book is very thorough, but at times it seems like the authors go out of their way to keep their discussion on a very theoretical level.

Chapter 1: A highly theoretical review of DSP. You need good knowledge of DSP to understand it.

Chapter 2: Goes over the human speech production and recognition systems. Here you get some practical info on the spectral and time-domain properties that distinguish speech sounds.

Chapter 3: Describes a model of the speech production based on a series of pulses passed through filters that correspond to features of the human speech production system. Practical issues such as which zero and pole values work best are left as an exercise to the reader.

Chapter 4: A lot of mathematics relating long-term statistical properties to those of a short frame of speech data. Contains good info on how to find recursive formulas for statistical properties of speech frames. It is a great shame that the authors don't include examples in MATLAB or pseudo code.

Chapter 5: Linear Prediction. Discusses a mathematical algorithm for creating a prediction filter that could be used to predict the next value in a series of data. In speech processing we are interested in using the coefficients of this prediction filter to encapsulate the properties of a speech frame. Examples of 1st,, 2nd, and 3rd order filters would have gone along way to illustrate how to implement this. There are some good formulas to measure the degree of similarities between speech frames based on their LP filter coefficients.

Chapter 6: Introduces the concept of the cepstrum. Cepstral analysis allows you to de-convolve speech data to separate the excitation source from the vocal tract filter. That way you can lifter out (a play on filter out) the excitation (responsible for pitch) and focus only on the vowel sounds. As always, how to make it work is left as an exercise.

The rest of the book is about speech coding and speech recognition. My class did not cover the later chapters, so I only read the parts that applied to projects.

The book helped me understand speech processing journal articles which tend to assume a lot of background knowledge.

The theoretical background that this book provides is necessary; good engineers need to understand the underpinnings of their bag of tricks. It is painful, however, at times. To apply the information in this book, you will certainly need a good mentor and a lot of tinkering in MATLAB.

It is a classic and a necessary book for researchers in the speech processing area. Undergraduate students might find some sections of the book too theoretical. In which case, it is highly recommended as a reference book. It ranks among the best signal processing books ever written.

I was skeptical when I started on this book. But this book changed the way I looked a speech processing for ever. Wonderful classic on Speech Processing. I spent many moments with this book and came out with awe and respect for the authors. It is broad in its scope. Thanks IEEE for reissuing this classic.

A good book should make you think loud. This one does that job splendidly well.

The book is probably the best reference on the field. The authors take their utmost effort to explain even some of the most subtle details but in a unified theoretical fashion. Must buy for anyone interested in the Speech processing Area