Springer Undergraduate Texts in Mathematics and Technology Tomas Sauer Continued Fractions and Signal Processing Springer Undergraduate Texts in Mathematics and Technology Series Editors Helge Holden, Department of Mathematical Sciences, Norwegian University of Science and Technology, Trondheim, Norway KeriA.Kornelson,DepartmentofMathematics,UniversityofOklahoma,Norman, OK, USA Editorial Board Lisa Goldberg, Department of Statistics, University of California, Berkeley, Berkeley, CA, USA Armin Iske, Department of Mathematics, University of Hamburg, Hamburg, Germany Palle E.T. Jorgensen, Department of Mathematics, University of Iowa, Iowa City, IA, USA Springer Undergraduate Texts in Mathematics and Technology (SUMAT) publishes textbooks aimed primarily at the undergraduate. Each text is designed principally for students who are considering careers either in the mathematical sciences or in technology-based areas such as engineering, finance, information technology and computer science, bioscience and medicine, optimization or industry. Texts aim to be accessible introductions to a wide range of core mathematical disciplines and their practical, real-world applications; and are fashioned both for course use and for independent study. More information about this series at http://www.springer.com/series/7438 Tomas Sauer Continued Fractions and Signal Processing 123 TomasSauer UniversitätPassau Passau, Germany ISSN 1867-5506 ISSN 1867-5514 (electronic) SpringerUndergraduate Textsin MathematicsandTechnology ISBN978-3-030-84359-5 ISBN978-3-030-84360-1 (eBook) https://doi.org/10.1007/978-3-030-84360-1 MathematicsSubjectClassification: 11A55,40A15,65D32 ©TheEditor(s)(ifapplicable)andTheAuthor(s),underexclusivelicensetoSpringerNature SwitzerlandAG2021 Thisworkissubjecttocopyright.AllrightsaresolelyandexclusivelylicensedbythePublisher,whether thewholeorpartofthematerialisconcerned,specificallytherightsoftranslation,reprinting,reuseof illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmissionorinformationstorageandretrieval,electronicadaptation,computersoftware,orbysimilar ordissimilarmethodologynowknownorhereafterdeveloped. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publicationdoesnotimply,evenintheabsenceofaspecificstatement,thatsuchnamesareexemptfrom therelevantprotectivelawsandregulationsandthereforefreeforgeneraluse. The publisher, the authors and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, expressed or implied, with respect to the material contained hereinorforanyerrorsoromissionsthatmayhavebeenmade.Thepublisherremainsneutralwithregard tojurisdictionalclaimsinpublishedmapsandinstitutionalaffiliations. ThisSpringerimprintispublishedbytheregisteredcompanySpringerNatureSwitzerlandAG Theregisteredcompanyaddressis:Gewerbestrasse11,6330Cham,Switzerland Dedicated to the memory of my thesis advisor Hubert Berens who was a “Doktorvater” in the best meaning of the word. Preface Fractionsarestrangeanimals.Nowadays,everyoneisoratleastshouldbeawareof what 3 and 6 stand for, and that these two expressions actually mean the same 4 8 rationalnumber,althoughtheydonotrepresentthesamemeterindancemusic.For the ancient Egyptians on the other hand, these fractions were not really what they wanted;theypreferredtodealwithpurereciprocalsorunitfractionslike1or1.The 5 9 GermanwordforthatisStammbruch;Google’sautomatictranslationStembreakis ahintthatartificialintelligencestillhasthepotentialforimprovement.InEgyptian mathematics, fractions were actually written as sums of unit fractions and there even existed tables for such decompositions [53]. In this spirit, we (nowadays) consider a with a[b and write this fraction as b a c 1 ¼a þ ¼a þ ; b 0 b 0 b c where the fraction appearing in the denominator of the unit fraction on the right-hand side is again similar to the type a; the numerator is larger than the b denominator. This idea defines a recursive process and allows us to write each fraction formally as a 1 ¼a þ ¼:½a ;a ;a ;...(cid:2): b 0 ½a ;a ;...(cid:2) 0 1 2 1 2 This is the definition of a continued fraction . It turns out that any rational number, i.e., any fraction, has exactly one finite representation of the form ½a ;a ;...;a (cid:2)forsomen,atleastiftrivialambiguitiesareexcluded.Therefore,the 0 1 n embarrassmentofmultiplerepresentationssuchas3¼6isavoidedhere.Moreover, 4 8 anyrealnumbercanbeexpressed,againuniquely,byaninfinitecontinuedfraction expansion ½a ;a ;a ;...(cid:2); most prominently known is the famous golden ratio 0 1 2 vii viii Preface pffiffiffi 1þ 5 ¼½1;1;1;...(cid:2); 2 whichwewillidentifyasthemostirrationalnumberpreciselyduetothefactthatit has the simplest possible continued fraction expansion. According to [61] who in turnrefersto[3],thiswasthewaytheancientGreeksintroducedrealnumbersonce pffiffiffi pffiffiffi theyrealizedthat 2¼½1;2;2;...(cid:2)isnotrational,thatis, 2cannotbewrittenas a fraction p=q with p2Z and q2N. Even after these very simple observations it should not come as a big surprise any more that continued fractions play a fundamental role in Number Theory and even in music. This will be the content of the first major chapter of this book. But onceonedealswithrationalnumbersandstartstothinkbeyondnumbers,onemay be tempted to also consider more general rational objects, in particular rational functions in the sense of quotients of polynomials. Continued fractions with polynomials in them were an important topic in the eighteenth and nineteenth centuries with major contributions by Bernoulli, Euler, Gauss and many more. Their contribution will occupy the major part of this book and will relate to con- cepts like quadrature, moment problems and sparse recovery which all appear, not quite accidentally in connection with Signal Processing, which will make these classicalmathematicaltoolsrelevantformodernapplicationsaswell.Notthatthey have ever been irrelevant, just to record that. The last topic is to describe polynomials which have zeros that are restricted to certainlocations,eitheroutsidetheunitcircleorinthelefthalf-plane.Thelatter,the so-called Hurwitz polynomials have a really nice characterization in terms of continued fractions. Prerequisites and Recommended Reading Thisbookemergedfromanintroductorylectureoncontinuedfractions,andstillis supposed to act as elaborated lecture notes that view continued fractions from a SignalProcessingperspectiveandfocusontheirroleinnumericalapplications.To thatend,itstartswithanintroductorychapterthatsummarizesthemainconceptsto beconsideredandoutlineswhatwillhappeninthechapterstofollowwhereresults are worked out together with all the necessary details of sometimes a technical nature.Thepresentationaimsatbeingmostlyself-containedandshouldnotrelyon toomuchpreviousknowledge,exceptforthestandardbasicknowledgeinAnalysis and Linear Algebra. Some more complex detailed aspects or material that is not centraltothemainthemeandthereforedidnotmakeitintothepresentation,butare providedintheexercisessothereadercanelaborateonthem.Theproblemsshould be structured enough to provide sufficient information on the intermediate steps, thereby helping to work out these additional facts and complete the theory. Preface ix Some more specific remarks on the requirements needed for the individual chapters might be helpful for using this book for lectures or seminars. Of course Chap. 1 only surveys things to come and requires only curiosity from the readers’ side. Chapter2isalmostcompletelyelementaryandrequiresnotmuchmorethanthe skill to manipulate fractions, to follow a “proof by induction” and to compute a derivativeofpolynomialsorexponentialfunctionsonceinawhile.Thepricetobe paidforthatisSect.2.6whichdealswithalgebraicnumbersandonlyremainsvery superficial. Here, it may be useful to have a look at [45] where, for example, also the transcendence of p and e are proved in part even by using continued fractions. For a deeper connection between math and music that could make a seminarof its own, the first choice is still [5], but also [82] and [69] are highly recommended. Chapter 3 makes use of some basic concepts of algebra, especially Euclidean rings. Any standard literature on algebraic structures is more than sufficient here and some acquaintance with these concepts is not really necessary, only helpful. Due to the algorithmic nature of the chapter, [31] is a good choice for further reading in case some things are not clear. Chapter 4 is mostly driven by explaining “Gauss’ quadrature”, i.e., the original approach by Gauss for developing what is nowadays known as Gaussian quadra- ture. The quadrature problem is treated in all classics in Numerical Analysis, for example[54].Forfurtherbackground,Iwouldrecommend[33]whichwaswritten byanexpertinquadratureandorthogonalpolynomials.Althoughthingsbecomea bit more intricate and tricky in that chapter, everything is still mostly elementary and should be understandable from the presentation. Chapter 5 uses the Dirac distribution which is popular in the engineering liter- ature and on Wikipedia pages where it often is used like a function which fre- quently leads to incorrect proofs. Basic facts on distributions are pointed out in [114] in a very condensed way; a more substantial presentation to understand generalized functions isto view functions asfunctionalsas provided in [34], but it isreallynotnecessaryhere.Whatisimportantisanawarenessthatdistributionsare beautifulbuttreacherous.Again,therestofthechapterisfairlyelementary;infinite matrices and operators on sequence spaces work in a straightforward extension ofthefinitecase.Somebasicknowledge,however,inFunctionalAnalysis,see,for example [63], could be useful. Chapter 6 introduces basic concepts of mathematical signal processing, still basedontheclassicslike[44].Usefuladditionalreading,especiallyonallpassand wavelets would be [106]; and of course [71]. This additional seconding will be helpful to obtain a more complete image of the underlying theory than what is brieflysurveyedinChap.6;buttheknowledgewillbeespeciallyusefulforworking out seminar talks based on this chapter. Chapter7usesafundamentalconceptinFunctionTheory,namelytheso-called argument principle which will be recalled in Theorem 7.4. It is a consequence of the residue theorem, hence it is located at the end of a one-semester lecture on Function Theory, and can be found, e.g., in [25, 50, 100]. Its main application is identity (7.11) and if one is willing to “believe” this identity, one can restrict the x Preface needed Complex Analysis to elementary manipulations of complex numbers that should be known from calculus/analysis. The other concepts in this chapter are provided and nicely introduced in the first chapters of [27]. Literature on Continued Fractions The literature on continued fractions is not overly huge, but definitely existent. Besides the classic, Perron’s two volumes [78, 79], one has to mention Khinchin [60]astheotherextreme;whilePerronistrulyencyclopedic,coversalotofissues, he is not so easy to read (even besides the fact that the books are in German), Khinchin’sbookisthin,butaverywell-writtenintroductiontocontinuedfractions. Even if some material has been added meanwhile and some of the proofs were modified, it still may be recognized as the guideline for Chap. 2. Another classic bookwherebasicsoncontinuedfractionscanbefound,ofcourseinthecontextof Number Theory, is the one by Hardy and Wright [45]. The special application to musicfromSect.2.7iscoveredinalmostallbooksonMathematicalMusicTheory, from the highly recommended and mathematically substantial masterpiece by Benson [5] to [23, 82]. Their connection is also mentioned more from the per- spectiveofComputerAlgebrain[31]and[61].ApologiestoallIforgottomention here. Applications in Analysis similar to what we consider in later chapters were already treated in [107] which is even older than Perron’s work; and in [68]. Connectionstoorthogonalpolynomialscanbefoundin[16].Thestatisticalcontext is presented in greater detail in [10]. There are also quite recent books on the issue: Hensley [48] covers Number Theory, Ergodic Theory and complex numbers, and a very careful overview especially about number-theoretic aspects is also provided in [64]; geometric aspectsandevengeometry-basedextensionstothemultivariatecasearepointedout in[57].TheroleofcontinuedfractionsinNumberTheoryisonceagainconsidered in quite some depth in [9] which combines classical with more modern aspects. Finally, a more educational approach from a somewhat recreational perspective is provided in [101]. Some Personal Remarks Myfirstpersonalencounterwithcontinuedfractionshappenedbyaccidentandwas totallyunintendedasIlookback.WorkingasateachingassistantattheUniversity ofErlangen,I was responsiblefor theexercisesinthe Numerical Analysisclass of my boss and thesis supervisor Hubert Berens. When he arrived to teach Gaussian quadrature,hetoldmethathehadread,probablyinChihara’sbook[16],thatGauss himselfhadnotusedorthogonalpolynomials(whichwereunknownatthistimeas