16548 notes 1

Upload: akshit-sharma

Post on 05-Jul-2018

215 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/15/2019 16548 Notes 1

    1/61

    1

    16.548

    Coding, Information Theory (and

    Advanced Modulation

    !rof. "ay #eit$en

    %all 411

    "ay&'eit$enuml.edu

  • 8/15/2019 16548 Notes 1

    2/61

    )

    Cla** Coverage

    + undamental* of Information Theory (4'ee-*

    + %loc- Coding ( 'ee-*+ Advanced Coding and modulation a* a 'ay

    of achieving the /hannon Ca0acity ound2

    Convolutional coding, trelli* modulation,and turo modulation, *0ace time coding (3'ee-*

  • 8/15/2019 16548 Notes 1

    3/61

    Cour*e #e /ite

    + htt02faculty.uml.edu'eit$en16.548

      Cla** note*, a**ignment*, other material* on

    'e *ite

      !lea*e chec- at lea*t t'ice 0er 'ee- 

      7ecture* 'ill e *treamed, *ee cour*e 'e*ite

    http://faculty.uml.edu/jweitzen/16.548http://faculty.uml.edu/jweitzen/16.548

  • 8/15/2019 16548 Notes 1

    4/61

    4

    !rereui*ite* (#hat you need to

    -no' to thrive in thi* cla**+ 16.6 or 16.584 (A !roaility cla**

    + /ome !rogramming (C, 9%, Matla

    + :igital Communication Theory

  • 8/15/2019 16548 Notes 1

    5/61

    5

    ;rading !olicy

    + 4 Mini

    +7em0el $iv com0re**or 

    + Cyclic >edundancy Chec- 

    + Convolutional Coder:ecoder *oft

    deci*ion

    + Trelli* Modulator:emodulator 

  • 8/15/2019 16548 Notes 1

    6/61

    6

    Cour*e Information and Te?t

    %oo-*+ Coding and Information Theory y #ell*,

     0lu* hi* note* from @niver*ity of Idaho

    + :igital Communication y /-lar, or !roa-i*

    %oo- 

    + /hannon* original !a0er (1B48

    + ther material on #e *ite

  • 8/15/2019 16548 Notes 1

    7/61

    3

    Claude /hannon ound* /cience

    of Information theory in 1B48In hi* 1B48 0a0er, DD

    A Mathematical Theory of Communication,EE

    Claude F. /hannon formulated the theory of datacom0re**ion. /hannon e*tali*hed that there i* a

    fundamental limit to lo**le** data com0re**ion.

    Thi* limit, called the entro0y rate, i* denoted y

     H . The e?act value of H  de0end* on the

    information *ource

  • 8/15/2019 16548 Notes 1

    8/61

    8

  • 8/15/2019 16548 Notes 1

    9/61

    B

    Thi* i*

    Im0ortant

  • 8/15/2019 16548 Notes 1

    10/61

    1G

    /ource Modeling

  • 8/15/2019 16548 Notes 1

    11/61

    11

    Hero order model*

    It ha* een *aid, that if you get enough mon-ey*, and *it them do'n at enough

    ty0e'riter*, eventually they 'ill com0lete the 'or-* of /ha-e*0eare

  • 8/15/2019 16548 Notes 1

    12/61

    1)

    ir*t rder Model

  • 8/15/2019 16548 Notes 1

    13/61

    1

    igher rder Model*

  • 8/15/2019 16548 Notes 1

    14/61

    14

  • 8/15/2019 16548 Notes 1

    15/61

    15

  • 8/15/2019 16548 Notes 1

    16/61

    16

  • 8/15/2019 16548 Notes 1

    17/61

    13

  • 8/15/2019 16548 Notes 1

    18/61

    18

  • 8/15/2019 16548 Notes 1

    19/61

    1B

  • 8/15/2019 16548 Notes 1

    20/61

    )G

    Heroth rder Model

  • 8/15/2019 16548 Notes 1

    21/61

    )1

  • 8/15/2019 16548 Notes 1

    22/61

    ))

    :efinition of Fntro0y

    /hannon u*ed the idea* of randomne** and entro0y from the

    *tudy of thermodynamic* to e*timate the randomne** (e.g.

    information content or entro0y of a 0roce**

  • 8/15/2019 16548 Notes 1

    23/61

    )

    Juic- >evie'2 #or-ing 'ith

    7ogarithm*

  • 8/15/2019 16548 Notes 1

    24/61

  • 8/15/2019 16548 Notes 1

    25/61

    )5

  • 8/15/2019 16548 Notes 1

    26/61

    )6

  • 8/15/2019 16548 Notes 1

    27/61

    )3

  • 8/15/2019 16548 Notes 1

    28/61

    )8

    Fntro0y of Fngli*h Al0haet

  • 8/15/2019 16548 Notes 1

    29/61

    )B

  • 8/15/2019 16548 Notes 1

    30/61

    G

  • 8/15/2019 16548 Notes 1

    31/61

    1

  • 8/15/2019 16548 Notes 1

    32/61

    )

  • 8/15/2019 16548 Notes 1

    33/61

  • 8/15/2019 16548 Notes 1

    34/61

    4

    Kind of

    Intuitive,

     ut hard to

     0rove

  • 8/15/2019 16548 Notes 1

    35/61

    5

  • 8/15/2019 16548 Notes 1

    36/61

    3

    %ound* on Fntro0y

  • 8/15/2019 16548 Notes 1

    37/61

  • 8/15/2019 16548 Notes 1

    38/61

    B

    In thi* 0re*entation, 'ell di*cu** the

     joint density of t'o random variale*.Thi* i* a mathematical tool for

    re0re*enting the interde0endence of

    t'o event*.

    ir*t, 'e need *ome random

    variale*.

    7ot* of tho*e in %edroc-.

  • 8/15/2019 16548 Notes 1

    39/61

    4G

    7et N e the numer of day* red

    lint*tone i* late to 'or- in a given'ee-. Then N i* a random varialeO

    here i* it* den*ity function2

    Ama$ingly, another re*ident of %edroc- i* late 'ith

    e?actly the *ame di*triution. It*...

    red* o**, Mr. /lateP

     L 1 )

    (L .5 . .)

    L 1 )

  • 8/15/2019 16548 Notes 1

    40/61

    41

     L 1 )

    (L .5 . .) >ememer thi* mean* that!(NQ Q .).

    7et e the numer of day* 'hen /late i* late. /u00o*e 'e

    'ant to record %T N and for a given 'ee-. o' li-ely

    are different 0air*R

    #ere tal-ing aout the joint density of N and , and 'e recordthi* information a* a function of t'o variale*, li-e thi*2

    1 ) 1 .5 .1 .G5

    ) .15 .1 .G5

    G .1 .1

    Thi* mean* that

    !(NQ and Q) Q .G5.

    #e lael it f(,).

    L 1 )

  • 8/15/2019 16548 Notes 1

    41/61

    4)

     L 1 )

    (L .5 . .)

    1 )

    1 .5 .1 .G5) .15 .1 .G5

    G .1 .1

    The fir*t o*ervation to ma-e i* that

    thi* oint 0roaility function

    contain* all the information from

    the den*ity function* for N and

    ('hich are the *ame here.

    or e?am0le, to recover !(NQ, 'e

    can add f(,1Sf(,)Sf(,.

    .) The individual 0roaility function*recovered in thi* 'ay are calledmarginal .

    Another o*ervation here i* that /late i* never late three day* in

    a 'ee- 'hen red i* only late once.

    L 1 )

  • 8/15/2019 16548 Notes 1

    42/61

    4

     L 1 )

    (L .5 . .)

    /ince he ride* to 'or- 'ith red (at lea*t until the directing career'or-* out, %arney >ule i* late to 'or- 'ith the *ame 0roaility

    function too. #hat do you thin- the oint 0roaility function for

    red and %arney loo-* li-eR

    1 )

    1 .5 G G

    ) G . G

    G G .)

    It* diagonalP

    Thi* *hould ma-e *en*e, *ince in any

    'ee- red and %arney are late the

    *ame numer of day*.

    Thi* i*, in *ome *en*e, a ma?imum

    amount of interaction2 if you -no'

    one, you -no' the other.

    L 1 )

  • 8/15/2019 16548 Notes 1

    43/61

    44

     L 1 )

    (L .5 . .)

    A little

  • 8/15/2019 16548 Notes 1

    44/61

    45

     L 1 )

    (L .5 . .)

    1 )

    1 .)5 .15 .1) .15 .GB .G6

    .1 .G6 .G4

    /ince 'e -no' the variale* N

    and H (for /0oc- are

    inde0endent, 'e can calculateeach of the oint 0roailitie*

     y multi0lying.

    or e?am0le, f(), Q !(NQ) and HQ

    Q !(NQ)!(HQ Q (.(.) Q .G6.

    Thi* re0re*ent* a minimal amount of

    interaction.

  • 8/15/2019 16548 Notes 1

    45/61

    46

    :e0endence of t'o event* mean* that -no'ledge of one give*

    information aout the other.

     Lo' 'eve *een that the oint den*ity of t'o variale* i* ale to

    reveal that t'o event* are inde0endent ( and , com0letely

    de0endent ( and , or *ome'here in the middle ( and .

    7ater in the cour*e 'e 'ill learn 'ay* to uantify de0endence.

    /tay tuned.

    A%%A :A%%A :P

  • 8/15/2019 16548 Notes 1

    46/61

    4B

  • 8/15/2019 16548 Notes 1

    47/61

    5G

  • 8/15/2019 16548 Notes 1

    48/61

  • 8/15/2019 16548 Notes 1

    49/61

    5)

    Conditional !roaility

    another event

    (

    ,(U(

     B P 

     B A P  B A P    =

  • 8/15/2019 16548 Notes 1

    50/61

    5

    Conditional !roaility (contd

    !(%UA

  • 8/15/2019 16548 Notes 1

    51/61

    54

    :efinition of conditional 0roaility

    + If !(% i* not eual to $ero, then the conditional 0roaility of A

    relative to %, namely, the 0roaility of A given %, i*

    !(AU% Q !(A %!(%

    !(A % Q !(% !(AU%

    or 

    !(A % Q !(A !(%UA

    ∩ •

    ∩ •

  • 8/15/2019 16548 Notes 1

    52/61

    55

    Conditional !roaility

    G.45G.)5 G.)5

    A %

    !(A Q G.)5 S G.)5 Q G.5G

    !(% Q G.45 S G.)5 Q G.3G

    !(A Q 1 < G.5G QG.5G

    !(%Q 1

  • 8/15/2019 16548 Notes 1

    53/61

    56

    7a' of Total !roaility

    If % % ,......, and % are mutually e?clu*ive event* of 'hich

    one mu*t occur, then for any event A

    !(A Q !(% !(AU% S !(%

    1 ) - 

    1 1 )

    ,

    ( U ...... ( ( U ⋅ ⋅ + + ⋅

     P A B P B P A Bk k )

     P A P B P A B P B P A B( ( ( U ( ( U E E= ⋅ + ⋅

    Special case of rule of Total Probability

  • 8/15/2019 16548 Notes 1

    54/61

    53

    %aye* Theorem

  • 8/15/2019 16548 Notes 1

    55/61

    58

  • 8/15/2019 16548 Notes 1

    56/61

    5B

    ;enerali$ed %aye* theorem

     If    % % and % are mutually e?clu*ive event* of 'hich

    one mu*t occur, then

    1 ) - , ,....

    4((.......4((4((

    4((4(

    ))11   k k 

    ii

    i

     B A P  B P  B A P  B P  B A P  B P 

     B A P  B P  A B P 

    ⋅++⋅+⋅

    ⋅=

    -.),......,1,Qi for 

  • 8/15/2019 16548 Notes 1

    57/61

    6G

    @rn !rolem*

    + A00lication* of %aye* Theorem

    + %egin to thin- aout conce0t* of Ma?imum

    li-elihood and MA! detection*, 'hich 'e'ill u*e throughout codind theory

  • 8/15/2019 16548 Notes 1

    58/61

    61

  • 8/15/2019 16548 Notes 1

    59/61

    6)

  • 8/15/2019 16548 Notes 1

    60/61

    6

  • 8/15/2019 16548 Notes 1

    61/61

    Fnd of Lote* 1