BOOK REVIEWS

R. Bruce Lindsay
Brown University, Providence, Rhode Island 02912
The opinions expressed are those of the individual reviewers and are not
necessarily endorsed by the Editorial Board of this Journal.

PsychologicalAcoustics: Benchmark Papers in arguewith a conclusionthat a selectedauthor reached,one cannot

argueeffectively with the inclusionof that author's work.
Acoustics, Volume 13
Rather than namethe 36 papersthat make up the volume,let
Earl D. Schubert, Ed. me simplyrecordthe majorheadingsunderwhich thosepapersfall:
Dowden, Hutchinson,and Ross,Inc., $troudsburg, Sensitivityof the Ear, ClassicalPsychoacoustics,
Pitch Mechanisms
Pennsylvania, 19 79. andPitchPerception,Separationof Simultaneous Signals,Time Re-
xii + 791 pp. Price: $?4.50. solution,and Advantagesof the BinauralSystem. The articlesthat
you expect to seeunder thoseheadingsare probably there. Even if
they weren't,though,the book is worth owningfor the essays that
Editing a collection of classicpapersin the field of psychological Schubertwrote to cementthe components together.
acousticscould be one of the easiesttasksin the world if one simply
JERRY V. TOBIAS
glancedback through one'snotecards,picked out the most impor-
tant, older, still-quoted articles, organizedand ordered them within Industrial A udiology
a reasonablylogicaloutline, and sentthe packageto the printer. Post Office Box 358
That sort of collagehasappearedfrom time to time and hasoften Norman, Oklahoma 73070
been useful to scholars as well as to •tudents.
But the thirteenth volume in the BenchmarkPapersin Acoustics
series,calledPsychologicalAcoustics,is far more. Earl Schubert's
commentary,like those in other volumesin this series,was planned
as a summary of the field and an explanation of the place and im- EngineeringApplications of Correlation and
portanceof the papersselected. The commentary servesthat pur-
posewell, but in addition, it shouldbe regardedas a valuablepart Spectral Analysis
of the literature in its own right. Julius S. Bendat and Allan G. Piersol
With simplicity and clarity, Schuberthaswritten a seriesof essays Wiley-Interscience, New York, 1980.
on half a dozen major areasin psychologicalacoustics,and those 50 xiv + 302 pp. Price $29.95.
or so pagesform a model, historically basedtextbook that covers
not only the papersthat are reprinted, but a coupleof hundred The centralsubjectof this book is the dynamicresponseof
others that developedfrom, expandedon, modified the interpre- complex mechanicalsystemsexcited by multiple random forces.
tations of, or challengedthe findingsreported in theseBenchmark The authorsdescribehow they haveestimatedsystemfrequency
Papers. Becausethis seriesof essaysis publishedtogether with the responsefunctions ("system identification") and excitation source
selectedarticlesthemselves,the volume is amongthe most valuable spectralcharacteristics
("energysourceidentification")after making
backgroundbooks available. somesimplifyingassumptions.All examplesgivenare mechanical
Spacepreventsmy reporting on every detail of Schubert'scontri- and/or acoustical.
bution. Instead, let me include an example of the issueshe writes The book is not a text, reference manual, or "cookbook". It
about; this one is from his general introduction. In it, he stresses comesnearestto being a monographlaced with technical anecdotes
the amazingability of the auditory systemto tell the differencebe- fromthe authors'professionalpractice.The authorsstatethat the
tween component soundsin a complex signalthat is heard under the "book attemptsto bridgethe differencesbetweenmeasuredresults
worst possibleconditions. In fact, whereasa few yearsago, writers from engineering
practiceasopposedto expectedtheoreticalresults
on "auditory theory" were generallydiscussing problemsof pitch fromanalyticalmodelsby discussing the detailsfrommanyengi-
perception, today the central problem is signalselection. Certainly neeringexamples."It supplements (anddoesnot replace)the
the mysteries of how the auditory system breaks down the single, authors'earlier book, Random Data: Analysisand Measurement
complex wavethat reachesthe eardrumand managesto differenti- Procedures(Wfley-Interscience,1971).
ate the fiddle from the oboe from the coloratura from the air condi- The elevenchaptersare entitled, 1. Introduction and Back-
tioner remain lessclear than the mysteriesof determining the pitch ground;2. ProbabilityFunctionsand Amplitude Measures;3. Cor-
of the tone that the first three are striking in unison. relationand SpectralDensityFunctions;4. SingleInput/SingleOut-
List making is a common and enjoyable human enterprise,and put Relationships;5. SystemIdentificationand Response; 6. Propa-
wheneverone seessomeoneelse'slist, a temptation arisesto try to gationPath Identification;7. SingleInput/Multiple Output Prob-
substitutea list of one's own. Selectingitems to include in a Bench- lems;8. MultipleInput/OutputRelationships; 9. ErtergySource
mark Papersvolume is the sort of high-levellist makingthat invites Identification;10. Procedures
for SolvingMultiple Input/Output
othersto invent revisionsof the Table of Contents. I was surprised Problems;and 11. StatisticalErrorsin Measurements.Each chapter
to find that my list differedfrom Schubert'sin no significantway. hasa shortlist of references,and a summarylist is givenseparately.
He includedhalf a dozen papersthat I wouldn't havethoughtof in The book has a reasonableIndex and a useful Glossaryof Symbols,
the context in which he included them (but he was right), and I althoughneither is complete.
would have added one early, German-languagepaper by yon Btk•sy The book is aimed at engineersin practice, particularly those
(mostof whosework belongsin a physiological
acoustics
rather with mechanical,civil, ocean,and aerospaceacademicbackgrounds.
than in a psychologicalacousticscollection). I haveto conclude To read the book with reasonablecomprehension requiressolid
that the articlesare quite properly chosenfor inclusionand that understandingof probability theory, modal analysismethods, the
they form a nearlyidealresource.Evenwhereonemight want to transfer function approach,Fourier series,and transform calculus.

262 J. Acoust.Soc. Am. 70(1 ), July 1981 0001-4966/81/070262-02500.80 ¸ 1981 AcousticalSocietyof America 262

Redistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 128.240.225.44 On: Sat, 20 Dec 2014 12:13:08
 cordsare usedin computation, they easeover the interpretive de-
tailswith little explanation. Until very late in the book, they use
the term "f'mite Fourier transform" rather than "discrete Fourier

x,(t) v(t) transform" or "FFT" to describe a transformation where digital

computationis implied. Modernengineersusingdigitalcomputers
preferthe discretearithmeticof actualcomputation.How elsedoes
one know what the computeris really doing?
The third problemconcernssubstantiation of the equationsfor
conditionedspectralanalysis.The authorsrefer throughoutthe
bookto an "optimumfrequencyresponse function." The concept
is discussedwith respectto the "old" systemon pp. 98-99 and 191-
192 [seeFig.R1(a)of thisreview]. It ismentionedwith respectto
(c) the "new" systemon pp. 198-199, and a derivationis givenon pp.
224-245 [Fig.Rl(b)]. Forbothsystems, theauthorstakeFourier
x(t) + transformsand form expressions N(f) .... andX2.• (f) .... . They
squarethe expressions, take expectations,expressas power spectra,
and set i)N/i)H = 0 and •)X2.•/i)L•2 = 0. The resultshavethe familiar
forms
H =Sxy[S
• andL•2=S•2[S•2. Theauthors
statethatbothof
theseexpressions
are optimumleast-squares
estimatesfor their re-
spectivetransfer functions.
This claim is not correct for the "old" system. In this system,
FIG. R1. (a) "old" system(from authors' Fig. 5.1); (b) "new" system H(f) isassumed
to be independentof n(t). SettingaN[i)H = 0 simply
(from authors' Fig. 8.4); (c)Wiener filter representationfor "new" expressesthis independence.
system. The claim is correct for the "new" system. In fact, it is the
heart of the substantiationfor conditioned spectralanalysis. The
authors'explanation,however,is opaque,evenmisleading.In Fig.
Rl(b),x• (t) andx2(t) areassumed
to bepartiallycorrelated.
The
An understandingof matrix methodsis helpful. Prior exposureto authorsdefinex2.•(t) to be "x2(t) with the lineareffectsof x• (t)
the theory of digital signalprocessing
aidsin interpretingmany removed"(p. 198). x2q(t) is thusan output, not an independent
statementsmade without adequateexplanation. inputasimplied[Fig.R1(b)]. Thepertinentquestion is,"whatmust
For anyonewith an adequatebackground,an effectiveway to L•2 be suchthatx2q(t) fulfillstherequiredcondition."Theauthors
get the mostfrom thisbook is to startwith Chap.11 (the last chap- answerscan be more clearly understoodby usingFig. Rl(c). When
ter, on measurementerrors),and then moveto and finishwith Chap. Lt2 iSselected
to minimizex2.t(t) in theleast-mean-squares
sense,
it
10 (computationprocedures). Some of the nomenclatureis unique, is an optimumfilter well knownto electricalengineers
asa Wiener
but it canbe decipheredby usingthe glossaryandindex. Readers ffiter. Readers will find an extensive literature available concerning
withoutan adequatebackground mightfind otherbooks,including optimum filters.
the authors' 1971 book, more satisfyingthan the first nine chapters The subjectmatter of this book is becomingincreasingly
im-
for obtaining that background. portantto mechanicallyorientedcollegecurricula.Instructors
Chapter10 is the technicalheartof the book. In it, the shouldbe wary, however. Thisbook is neitherdesignednor suited
authorsdiscuss"conditionedspectralanalysis,"a new analysis for useas a coursetext at any level. There are no worked-out
procedure
for multipleinput/outputproblems.Oneof the authors examples
andno problems.The authors'language
is oftendiffuse.
(J. S. Bendat) stated at the Fall 1980 meetingof the Society in Considerablematurity is often necessaryto identify hidden assump-
LosAngelesthat thisnewmethodhasmadethe "older"method tionsor to interpretwhat the authorsmeanfrom what they have
obsolete. The new procedureis discussedalso in a recent paper said.
by J. S. Bendat,"Modernanalysis procedures
for multipleinput• In summary,the book providessomevaluablepracticalinsights
outputproblems"[J. Acoust.Soc.Am. 158,498-503(1980)]. The which the authors share with readers. The discussionof conditioned
principalclaimmadeis that the newmethodis computationally spectralanalysisis a welcomeadditionto the literature. Large
"more efficient than older techniquesbut also providesgreater technicallibrariesshouldhavea copy. Smalllibrariesand individ-
engineering
insightinto the physicalmeaningof variousrelation- ualsmay wish to review a copy before acquiringone.
ships."
This book has its shareof typographicalerrorsand hazy defini-
tions. Few of thesewill trouble an adequatelypreparedreader. JOHN C. BURGESS
Three problemsare more serious. Universityof Hawaii
The first concernsskillsthat a reader can expect to acquire by Honolulu, Hawaii 96822
reading
thebook.Thereviewer's
opinion
isthattheauthors
were
overlyoptimisticin statingan objective"to bridgethe differences
betweenmeasuredresults... and expectedtheoreticalresults."
Nowhere in the book is there a worked-out example comparing
theory and experimentalresultswith sufficientdetail that a reader Voice Identification: Theory and Legal
can easilycheckhis understandingof the authors'methods. The Applications
authorsdiscuss
qualitativelywhat onemightdo ratherthanquant- Oscar Tosi
itatively the relativeadvantages
and disadvantagesof different
approaches to specificproblems.Readerswill acquireideasabout UniversityPark Press,Baltimore,1979.
procedures.They will haveto developtheir own specificapplica- 182 pp. Price $19.50.
tions examples.
The secondproblemis that, in practice,one usesdigital proce- The reader should be advised at the outset of this review that,
duxeswith finite4ength recordsto calculatesignalspectraand basedon personalresearchand study,thisrevieweris seriously
systemresponsefunctions. Almost all theoreticalexpressions in the biasedagainst
the "voiceprint"or "voicegram"approachto speaker
text are written in continuous-timeor continuous-frequencyform identification.
In hisopinion,
themethod
lacks
meritantishould
with the euphemism"limit T-• oo"implied or statedexplicitly. not be employed.In all fairness,therefore,the followingcomments
While the authors clearly state that discrete-time finite-length re- should be read with this bias in mind.

263 J. Acoust. Soc. Am. 70(1 ), July 1981; 0001-4966/81/070263-03500.80; ¸ 1981 Acoust. Soc. Am.; Book Reviews 263

Redistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 128.240.225.44 On: Sat, 20 Dec 2014 12:13:08