The goal of this book is to describe speech mechanisms and phenomena that form the basis for applications such as speech recognition and synthesis and the application of neural networks in these fields. Its intended audience consists mainly of engineers, and a certain background in signal processing is necessary.
After a brief introduction, the second chapter summarizes the fundamentals of computational phonetics and speech technology, and of neural networks. The subsequent two chapters deal with speech production and perception in more detail, while the last two chapters describe the application of neural networks in modeling speech perception and in performing speech synthesis and speech recognition.
The most problematic aspect of the book is the early date of its original publication in Japanese, 1990. Since both fields are very active, much relevant information and many important references are missing. The description of speech research, the authors’ first goal, is brief and contains most of the basics known in the late 1980s. However, the excellent book by O’Shaughnessy [1] from 1987 covers the same field better and more thoroughly. The second part, on the application of neural networks, lists the most important approaches to this field in the late 1980s. Though the individual sections are well written, threads are missing, especially in chapter 6.
Some of the stylistic errors made in the book’s translation from Japanese are not severe, but some consistent misspellings might cause readers confusion (for instance, “multilayer perception” instead of “multilayer perceptron” and “format synthesis” for “formant synthesis”). Many of the references cite articles by Japanese researchers, a substantial number of them written in Japanese; they are inaccessible, unless translated, to most researchers who do not read Japanese.
At the time of its original publication in 1990, the book presented a terse but reasonably complete overview of those aspects of speech research relevant for engineers. In addition, it described all applications of neural networks important to speech research and technology up to the late 1980s. The English version, published in 1995, would have benefited greatly from a thorough revision and inclusion of newer research results. Also, the references should have been updated and adapted to an international audience. Thus, it is difficult to recommend this book in its current form.