1962 Book Reviews 331

An Alternative Code for Indian Languages-S. N. Narahari Pandit is introduced to get the mathematical models taking into account
(in English). (Dept. of Math., Indian Inst. Tech., Kharagpur.) our imperfect knowledge of some signal parameters and thus
(Printed in the J. Inst. Telecommun. Engrs., vol. 7, pp. 14548; giving more adequate description of the real signals. Reception of
May, 1961.) signals transmitting messages discretely or continuously varying,
one and multi-dimensional is considered. New, more accurate
A telegraphic code for Indian languages, expressed in ortho- formulas giving the minimum square error in modulation systems
graphic form, has been given by Ramakrishna and others. In the were obtained.
present paper, an alternative code employing the same three symbols
of dot, dash and long dash is presented. This proposed code in
general has a slightly larger rate of transmission, is structurally On the Application of a Preemphasis Device to an SSB Radio
more akin to the Indian alphabets and hence shows a preference to Telephone System Using the VODAS-Y. Tsuruoka and S. Adachi
‘favorable noise.’ (in Japanese). (Inst. Elec. Engrs. Japan at Tohoku University,
Japan; June 14, 1961.)

On the Construction of Minimally Redundant Reliable System This paper describes the test results of an SSB Radio Telephone
Designs-D. K. Ray-Chaudhuri (in English). (Bell Sys. Tech. J., System employing preemphasis/deemphasis devices.
vol. 40, no. 2; March, 1961.) It was found that a marked improvement is brought about in
the reduction of the cutting distortion caused by the VODAS at
Several authors have considered the possibility of increasing the the initial syllable of the talk spurt when the speech inputs to the
reliability of large and complex binary digital systems by introducing radio transmitter are processed to the flat spectrum by the pre-
some redundancy in the system. In a companion paper, Armstrong emphasis device having a tilting characteristic at frequencies
proposes a scheme for applying error correction to a synchronous higher than 800 cps.
digital system. In this paper we develop a general mathematical In the case of S/N = 20 db. the test system showed an advantage
theory for generating minimally redundant error-correcting codes of 10 per cent in syllable articulation and 3 per cent in sentence
for the scheme in question. This results in what are called “mini- intelligibility over an ordinary system. This advantage results from
mally redundant reliable systems.” The problem of constructing a decrease in the masking effect on the transmitted signals at the
minimally redundant reliable systems whose output is free of higher frequencies and also from a noise improvement of 3 db in
error when there is a fault in at most one block of the system is the R-VODAS sensitivity.
completely solved. An example is considered in detail showing how
the mathematical theory can be actually applied.
An Examination of Axiomatization and Information Theory Applica-
tion in Neurophysiology-G. Werner (in English). (J. ScientiJic and
Methods of Threshold Signal Reception in Presence of Noise- Industrial Res., vol. lSA, no. 10, pp. 467471; 1959.)
J. Seidler (in Polish). (Prace Przemsylowego Znstytutu Telekomuni-
kacji, vol. 9, pp. 1-51; 1961.) The axiomatization tendency and the information network theory
application in neurophyaiology have been critically examined, and
The paper presents an attempt to give a unified approach to the experiments and linquistic habits involved in the various
the optimalization problem of threshold signal reception. Analysis theories have been briefly discussed. From this analysis it is con-
is based on decision function theory and prospective as well as cluded that biological material can be handled only in a framework
retrospective methods are used. The concept of passive parameters of empirically workable concepts.

Book Reviews

An Introduction to the Principles of Communication Theory- suppressed carrier, and single sideband. Performance under large
John C. Hancock. (McGraw-Hill Book Co., Tnc., New York, signal-to-noise ratio conditions is discussed. The results of this
N. Y.; 1961. 253 pages + x. $9.50) and later sections of the chapter are derived by treating a single
infinitesimal noise spectral component plus signal passing through
This book is intended as a text for a senior level course in in- a nonlinear device. Superposition is then applied over the entire
formation systems. As prerequisites for the course, the student noise spectrum. Unfortunately, the author fails to mention that the
should have familiarity with Fourier analysis, a good knowledge of assumption of high signal-to-noise ratio allows him to ignore the
electronic circuits and some network theory. noise-times-noise interactions and hence to apply superposition in
The book contains a number of possible t,rouble spots and pitfalls the otherwise nonlinear element. On the subject of angle modula-
for the unwary student. At worst, these may result in his taking tion, the standard results on spectral characteristics and signal-to-
away with him misconceptions of the subject matter. At best, he noise ratio are derived. In a section on pulse modulation, the author
will understand the author’s intent after the expenditure of extra obtains signal-to-noine ratio relations for pulse position modulation
time and effort. In the following paragraphs, which summarize the and also covers pulse-amplitude and duration modulation. A
contents, the reviewer has called attention to some of the trouble- section on pulse-code modulation treats the signal-to-noise ratio
some points. due to quantization noise, ignoring the channel noise. The chapter
In Chapter I the author presents a brief review of Fourier analysis concludes with a comparison of the various modulation systems in
including a thorough discussion of the Dirac delta function, and regard to SNR and bandwidth requirements.
concludes with a discussion of sampling theory. In an otherwise The chapter on random signal theory contains an elementary
tightly written chapter, the author has included a heuristic dis- exposition of probability concepts. Starting from first principles,
cussion of the matched filter before any prior mention of noise or the author derives a number of important relations, including the
signal-to-noise ratio ideas. This material, which is treated in greater Poisson distribution (unnamed in this text), the probability density
depth later on, seems out of place in the first chapter. function for the sum of two random variables, and the densit.y
Chapter II discusses various types of modulation beginning with function for a possibly multivalued transformation of a random
three forms of AM modulation: double sideband, double sideband variable. Autocorrelation functions, ergodicity, and spectral density
332 IRE TRANSACTIONS ON TNFORMATION THEORY July
are treated next, but, although two-dimensional distributions are was missed by not introducing the ratio of signal energy to noise
discussed, the role of the autoccrrelation function in the Ga,ussian power density. This fundamental communication parameter is
distribution is not mentioned. The general properties of correlation relevant to the discussions on matched filters, pulse code modula-
fun&ions are discussed and numerous examples serve to illustrate tion, channel capacity, thermal noise, and radar detection.
the results. In the reviewer’s opinion the book might serve a useful purpose
Chapter IV on network analysis treats the passage of randam in a survey course on communication theory for nonspecialists in
signals through linear and nonlinear networks. The input-output the field, hut it cannot be recommended ai the basis for a first
relations for autocorrelation and spectral densities are obtained. acquaintance to anyone intending to delve more deeply into any
In a section on probability density input-output relationships, the of the major topics.
author mistakenly introduces as an approximation the fact that, if a STEVEN M. SUSSMAN
Gaussian random process is applied to a linear system, the output Applied Science Div.
is also Gaussian. This property is treated in the same context with Melpar, Inc.
the approximation that the output of a narrow-band system ap- Watertown, Mass.
proaches a Gaussian distribution as a consequence of the central
limit theorem. The combination of these two ideas may breed
misconceptions in the reader, notwithstanding a correct statement Cybernetics, or Control and Communication in the Animal and
of the first property several paragraphs later. The section on non- the Machine-Norbert Wiener. (John Wiley and Sons, Inc., New
linear systems develops the distribution of envelope and phase for York, N. Y., and London, 1961, and M.I.T. Press, Cambridge,
the sum of a sine wave and Gaussian random process. Matched Mass., 1961. Second Edition. 212 pages + XVI.)
filters and minimum square error filters are covered under the More t,han thirteen years have passed since the first edition of
heading of optimum systems. The author avoids variational methods, “Cybernetics” was published. In this new edition Prof. Wiener
and consequently the reader is asked to take certain funda>mental has added a special preface, a chapter on Learning and Self-
results on faith. However, in spite of the omission of most proofs, Reproducing Machines, one on Brain Waves and Self-Organizing
the minimization of the integrals arising in this section gives a
Systems (together 34 pages) and a g-page index.
false impression as to the difficulty of deriving the results. For It is only in Chapter III, where Wiener gives a short survey of
example, the author states that a certain integral is clearly a mini- his own work on time series, that the reader gets into deep mathe-
mum when the integrand is zero, whereas the variational approach matical problems. This subject is difficult and the text concise.
show-s that a lower minimum can be obtained for a negative inte- Even so the new reader is luckier than earlier ones were, as numerous
grand. Fortunately, this error does not affect subsequent results. misprints have been removed. (Dr. G. Raizbeck is mentioned on
On the other hand, the non-negative integrals for which a minimum the cover as Technical Editor.) It is, however, not so much the
could be arrived at by inspection are not placed in the form where
great mathematician Wiener we meet in this book, as the man of
the result becomes evident. The latter difficulty is largely due to universal knowledge for whom the unity of science is still a reality.
the fact that, having employed s-plane terminology in the derivation, “Cybernetics” was perhaps the first book to draw attention to the
the author retains the s notation rather than converting to &I in analogy between information and control problems in statistical
the integrals representing the mean-square error. In addition to theory, computers, the nervous system and the society. It has had
the nonrealizable optimum filters, spectrum factorization for a great and well-deserved success in indeed bringing people of
realizable filter condit,ions is treated. different disciplines together to look over their problems, and it
Chapter V is concerned with information theory and covers the has led to new alleys in research, not so much by direct suggest,ions as
basic ideas of entropy, rate and channel capacity. Coding for the by suggesting a new frame of mind. Cybernetics covers so wide a field
discrete noiseless channel is treated in great detail, and coding for a that few symposia nowadays try to cover the whole of it. Those that
noisy channel is discussed heuristically in conjunction with the do have to divide into special sections very soon. For this reason
capacity of a Gaussian channel. By equating the information rate the name cybernetics may have been used less often in later years.
in the RF and audio sections of the receiver a comparison is made But in many symposia, especially in the bio-sciences, one can
between the ideal bandwidth-SNR t,radeofi and the results of Chap- easily see the lasting influence of Wiener and the group around
ter II on various modulation systems. A possible source of confusion him. even when cybernet,ics is not mentioned specifically. The
is the failure to distinguish between the quantization noise of new ideas by themselves did not lead to quick and easy results,
pulse-code modulation and the additive channel noise. but they triggered an effort in many laboratories all over the
Chapter VI is devoted to noise, its physical sources, spectral world that otherwise might not have developed so soon.
characterist,ics and attendant circuit analysis. Noise figure cal- The two new chapters and the new introduction fit well together
culations are covered thoroughly for both vacuum tube and tran- with the rest of the book. Special attention is paid in the first new
sistor circuits, and the entire chapter has a decided practical tone. chapter to Gabor’s work on nonlinear adaptive transducers and
This practical tendency may explain why the author treats the Wiener’s own representation of nonlinear problems in noise. In the
asymptotrs of functions as if they were the actual functions. chapter on Brain Waves the work on autocorrelation functions of
Without an explanatory stat,ement, 1 + f’ is considered “constant electroencephalograms by Barlow, Brazier, and Rosenblith in
with frequency up to [f = l] at which point it increases at a rate collaboration with Wiener is sketched. The effect of finite time-
of 6 db per octave.” sampling approximations in the calculation of the autocorrelation
The final chapter is entitled “Communication Systems” and function is comput’ed.
begins with a section on SNR in envelope detectors, which presents It would not be Wiener’s book if some of the analogies did not
the standard limiting results for high and low SNR. A short section seem rather far-fetched, like that between the construction of a
is devot,ed to further comparisons of modulation systems. This nonlinear network from ‘<white” boxes and the formation of a
material could well benefit from the inclusion of more detail and virus, or the importance of infrared radiation in the construction
clarification of the ground rules of the comparison. Elementary of viruses, genes and cancer. But who can see into the future?
radar systems are covered last in several sections on basic radar The index gives amusing examples of the wide scope of the book,
principles, the range equation, probabilities of detection and false with, under Bo, references to Bogey, Colonel; Bohr, Niels; Boltz-
alarm, and Doppler radar. mann L; Bonapart,e N. The cover gives a recent photograph of
Each chapter is followed by a set of problems and selected Prof. Wiener explaining his theories before the blackboard.
references to textbooks and key journal articles. The author is to Aft,er the first edition, Shannon in reviewing this book con-
be commended for including numerous illustrative examples through- gratulated Prof. Wiener on writing an excellent introduction to a
out the text. new and challenging branch of science. Now we can see this book
In at,tempting to cover such a vast amount, of material in a as one of the classics, which everyone interested in the state of
rather short book (253 pages) the author has set himself a difficult science should read. Wiener’s book will fascinate the reader from
task. Virtually any criticism one might make would inevitably the first introduction to the last chapter.
lead to an enlargement of the text. Additional emphasis in some
sections on methods of analysis as contrasted with results and F. L. H. M. STIJMPERS
applications would have made the text a more valuable educational Philips Research Labs.,
experience. Certainly an opportunity for unifying several topics Eindhoven, The Netherlands