dc02 detection theory

Upload: trannguyen-nhat-linh

Post on 01-Jun-2018

222 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/9/2019 DC02 Detection Theory

    1/90

    DETECTION THEORY

  • 8/9/2019 DC02 Detection Theory

    2/90

    Gram-Schmidt Procedure

    Permits the representation of any set ofMenergy signals, {si(t)}, as linear

    combinations ofNorthonormal basisfunctions { }, whereN M)(tj

  • 8/9/2019 DC02 Detection Theory

    3/90

    Gram-Schmidt Procedure

    Represent the given set of real-valued energysignals1(t), s2(t), ,sM(t), each of duration Tseconds, in the form

    N

    j

    jiji tsts1

    )()( Mi

    Tt

    ,....,2,1

    0

    T

    jiij dtttss0

    )()( Nj

    Mi

    ,....,2,1

    ,....,2,1

  • 8/9/2019 DC02 Detection Theory

    4/90

    Gram-Schmidt Procedure

    The real-value basis functions , .are orthonormal

    Each basis function is normalized to have unit energy

    The basis functions are orthogonal with respect toeach other over the interval 0 t T

    T

    ji dttt0

    01)()(

    jiji

    )(1 t )(2 t

  • 8/9/2019 DC02 Detection Theory

    5/90

    Example

  • 8/9/2019 DC02 Detection Theory

    6/90

    Gram-Schmidt Procedure

    Stage 1: We have to establish whether or not the given

    set of signals1(t), s2(t), , sM(t) is linearly

    independent Lets1(t), s2(t), , sN(t) denote subset of linearly

    independent signals, whereN M eachmember of the original set of signals s1(t), s2(t),

    , sM(t) may be expressed as a linearcombination of this subset of Nsignals

  • 8/9/2019 DC02 Detection Theory

    7/90

    Gram-Schmidt Procedure

    Stage 2:

    Construct a set of Northonormal basisfunctions

    1(t),

    2(t), ,

    N(t) from the linearly

    independent signalss1(t), s2(t), , sN(t)

  • 8/9/2019 DC02 Detection Theory

    8/90

    Geometric Interpretations ofSignals

    With orthonormal basis functions {i(t)},j = 1,2, , N, each real-valued energy signals1(t),

    s2(t), ,sM(t), can be expanded as

    N

    j

    jiji tsts1

    )()( Mi

    Tt

    ,....,2,1

    0

    T

    jiij dtttss0

    )()( Nj

    Mi

    ,....,2,1

    ,....,2,1

  • 8/9/2019 DC02 Detection Theory

    9/90

    Geometric Interpretations ofSignals

  • 8/9/2019 DC02 Detection Theory

    10/90

    Geometric Interpretations ofSignals

  • 8/9/2019 DC02 Detection Theory

    11/90

    Example

  • 8/9/2019 DC02 Detection Theory

    12/90

    Solution

  • 8/9/2019 DC02 Detection Theory

    13/90

    Geometric Interpretations ofSignals

    Each signal in the set {si(t)} is completely determined

    by the vector of its coefficients

    Si is called the signal vector

    iN

    i

    i

    i

    s

    ss

    2

    1

    s Mi ,....,2,1

  • 8/9/2019 DC02 Detection Theory

    14/90

    Geometric Interpretations ofSignals

    Set of signal vector {si}, i = 1, 2, , Masdefining a corresponding set of Mpoints inaN-dimensional Euclidean space

    N-dimensional Euclidean space is called

    signal space

  • 8/9/2019 DC02 Detection Theory

    15/90

    Geometric Interpretations ofSignals

    In anN-dimensional Euclidean space,

    define lengths of vectors and anglesbetween vectors

  • 8/9/2019 DC02 Detection Theory

    16/90

    Geometric Interpretations ofSignals

  • 8/9/2019 DC02 Detection Theory

    17/90

    Geometric Interpretations ofSignals

    Square-length of any signal vector si is define tobe the inner product or dot product of si

    The cosine of the angle between the vectors siand sj equals the ratio

    N

    j

    ijiii s1

    22 ),( sss

    ji

    ji

    ss

    ss ),(

  • 8/9/2019 DC02 Detection Theory

    18/90

    Geometric Interpretations ofSignals

    Energy of a signalsi(t) of duration T seconds is

    equal to

    Euclidean distance between the pointsrepresented by the signal vector si and sk

    T N

    jijii sdttsE 0 1

    22

    )(

    N

    j

    T

    kikjijki dttstsss1 0

    22 )]()([ss

  • 8/9/2019 DC02 Detection Theory

    19/90

    Model of a DigitalCommunication System

  • 8/9/2019 DC02 Detection Theory

    20/90

    Model of a DigitalCommunication System

    At the transmitter input, we have amessage source that emits one symbolevery T seconds, denote by m

    1

    , m2

    ,, mM

    Assume that all symbols are equally likely

    ( )i iP P m 1

    M

  • 8/9/2019 DC02 Detection Theory

    21/90

    Model of a DigitalCommunication System

    The output of message source is presentedto a vector transmitter, producing a vectorof real numbers

    1

    2

    .

    .

    .

    i

    i

    i

    i N

    s

    s

    s

    s

  • 8/9/2019 DC02 Detection Theory

    22/90

    Model of a DigitalCommunication System

    The channel is assumed to have twocharacteristics:

    The channel is linear, with a bandwidth that islarge enough to accommodate the transmissionof thesi(t) without distortion

    The transmitted signalsi(t) is perturbed by an

    additive, zero-mean, stationary, white Gaussiannoise process, denote by W(t)

  • 8/9/2019 DC02 Detection Theory

    23/90

    Model of a DigitalCommunication System

    We expressed the received randomprocess, X(t)

    x(t) is referred as the received signal

    ( ) ( ) ( )iX t S t W t 01,2,...,t T

    i M

  • 8/9/2019 DC02 Detection Theory

    24/90

    Model of a DigitalCommunication System

    The requirements is to design the vectorreceiver so as to minimize the averageprobability of symbol error defined as

    mi is the transmitted symbol and is estimateproduced by the vector receiver

    ( )e iP P m m

    m

  • 8/9/2019 DC02 Detection Theory

    25/90

    Model of a DigitalCommunication System

    If the receiver is phase locked to thetransmitter, it is referred to coherentreceiver

    On the other hand, the receiver is referredto non-coherent receiver

  • 8/9/2019 DC02 Detection Theory

    26/90

    Response of Bank of Correlatorsto Noisy Input

    Suppose that the input of the bank of N

    product-integrator or correlators israndom processX(t)

    W(t) is a white Gaussian noise process of zero

    mean and power spectral densityN0/2

    ( ) ( ) ( )iX t S t W t 0

    1, 2,...,

    t T

    i M

    f k f C l

  • 8/9/2019 DC02 Detection Theory

    27/90

    Response of Bank of Correlatorsto Noisy Input

    R f B k f C l

  • 8/9/2019 DC02 Detection Theory

    28/90

    Response of Bank of Correlatorsto Noisy Input

    The output of each correlator is

    0

    ( ) ( )T

    i jX X t t dt ij js W 1, 2,...,j N

    R f B k f C l t

  • 8/9/2019 DC02 Detection Theory

    29/90

    Response of Bank of Correlatorsto Noisy Input

    The first component, sij, is a deterministicquantity contributed by the transmitted signalsi(t)

    The second component, Wj, is a random variablethat arises because of the presence of noise atthe input

    0( ) ( )Tij i js s t t dt

    0( ) ( )

    T

    j jW W t t dt

    R f B k f C l t

  • 8/9/2019 DC02 Detection Theory

    30/90

    Response of Bank of Correlatorsto Noisy Input

    Consider next a new random process,X(t)

    1'( ) ( ) ( )

    N

    j jJX t X t X t

    1

    '( ) ( ) ( ) ( ) ( )N

    i ij j j

    j

    X t s t W t s w t

    1

    ( ) ( )

    '( )

    N

    j j

    j

    W t W t

    W t

    R f B k f C l t

  • 8/9/2019 DC02 Detection Theory

    31/90

    Response of Bank of Correlatorsto Noisy Input

    We may express the received randomprocess as

    1

    ( ) ( ) '( )

    N

    j j

    j

    X t X t X t

    1

    ( ) '( )N

    j j

    j

    X t W t

    R f B k f C l t

  • 8/9/2019 DC02 Detection Theory

    32/90

    Response of Bank of Correlatorsto Noisy Input

    The noise process W(t) has zero mean,hence the random varialbe Wj extractedfrom W(t) has zero mean

    The mean value of jth correlator output Xjdepends only onsij

    jx j

    m E X

    ij jE s w ij j ijs E W s

    Response of Bank of Co elato s

  • 8/9/2019 DC02 Detection Theory

    33/90

    Response of Bank of Correlatorsto Noisy Input

    The variance ofXj

    2

    j jx Var X

    2

    j ijE X s

    2

    jE W

    Response of Bank of Correlators

  • 8/9/2019 DC02 Detection Theory

    34/90

    Response of Bank of Correlatorsto Noisy Input

    2

    0 0( ) ( ) ( ) ( )

    j

    T T

    X j jE W t t dt W u u du

    0 0 ( ) ( ) ( ) ( )

    T T

    j jE t u W t W u dtdu

    0 0 ( ) ( ) ( , )T T

    j j Wt u R t u dtdu

    0 0

    ( ) ( ) ( ) ( )T T

    j t j u E W t W u dtdu

    Response of Bank of Correlators

  • 8/9/2019 DC02 Detection Theory

    35/90

    Response of Bank of Correlatorsto Noisy Input

    Since

    Then

    0( , ) ( )2

    W

    NR t u t u

    2 0

    0 0( ) ( ) ( )

    2jT T

    X j j

    Nt u t u dtdu

    20

    0

    ( )2

    T

    j

    Nt dt

    0

    2

    N

    Response of Bank of Correlators

  • 8/9/2019 DC02 Detection Theory

    36/90

    Response of Bank of Correlatorsto Noisy Input

    Similarly, we find thatXj are mutually uncorrelated

    j kj k j X k X

    Cov X X E X m X m

    0 0

    0 0

    0

    0 0

    0

    0

    ( ) ( ) ( ) ( )

    ( ) ( ) ( , )

    ( ) ( ) ( )2

    ( ) ( ) 02

    j ij k ik

    j k

    T T

    j k

    T T

    j k W

    T T

    j k

    T

    j k

    E X s X s

    E W W

    E W t t dt W u u du

    t u R t u dtdu

    Nt u t u dtdu

    Nt t dt

    Response of Bank of Correlators

  • 8/9/2019 DC02 Detection Theory

    37/90

    Response of Bank of Correlatorsto Noisy Input

    Define the vector of N random variables atthe correlator outputs as

    1

    2

    .

    .

    .

    N

    X

    X

    X

    X

    Response of Bank of Correlators

  • 8/9/2019 DC02 Detection Theory

    38/90

    Response of Bank of Correlatorsto Noisy Input

    Since the elements of the vector X arestatistically independent, we may express theconditional probability density function of thevector X, given that the signal si(t) or

    corresponding the symbol mi was transmitted

    This is called likelihood functions

    1

    | |j

    N

    i X j i

    j

    f m f x m

    X x 1, 2,...,i M

    Response of Bank of Correlators

  • 8/9/2019 DC02 Detection Theory

    39/90

    Response of Bank of Correlatorsto Noisy Input

    Since eachXj is a Gaussian random variablewith meansij and varianceN0/2, we have

    2

    00

    1 1| exp

    jX j i j ijf x m x s

    NN

    Response of Bank of Correlators

  • 8/9/2019 DC02 Detection Theory

    40/90

    Response of Bank of Correlatorsto Noisy Input

    Likelihood functions of AWGN channel aredefined by

    2

    2

    0

    10

    1| exp

    N N

    i j ij

    jf m N x sN

    x x

    1, 2,...,i M

  • 8/9/2019 DC02 Detection Theory

    41/90

    Maximum Likelihood Detector

    Suppose that, when the observation vectorhas the value x, we make decision

    The average probability of symbol error inthis decision, denoted byPe(mi,x), is

    mm

    ( , ) not sent|e i iP m x P m x

    1 sent|iP m x

  • 8/9/2019 DC02 Detection Theory

    42/90

    Maximum Likelihood Detector

    The optimum decision rule is

    set if

    for all k i

    Where k = 1, 2, , M

    This decision rule is referred to as maximum aposteriori probability (MAP)

    im m

    sent| sent|i kp m p mx x

  • 8/9/2019 DC02 Detection Theory

    43/90

    Maximum Likelihood Detector

    Applying Bayes rule, we may restate thisdecision rule as

    set if

    is maximum for k =i

    i

    m m

    |k kp f mf

    x

    x

    x

  • 8/9/2019 DC02 Detection Theory

    44/90

    Maximum Likelihood Detector

    SincefX(x) is independent of the transmittedsignal, and the a priori probability pk= pi when allthe signals are transmitted with equal probability,we may simplify the decision rule as

    set if

    is maximum for k =i

    The decision rule is referred to as maximum likelihood, and thedevice for its implementation is referred as maximum-likelihooddetector

    im m

    | kf mx x

  • 8/9/2019 DC02 Detection Theory

    45/90

    Maximum Likelihood Detector

    Since likelihood function is alwaysnonnegative, we may restate the decisionrule

    set if

    is maximum for k = i

    im m

    ln | kf m x x

  • 8/9/2019 DC02 Detection Theory

    46/90

    Maximum Likelihood Detector

    Let Z denote the N-dimensional space of allpossibly observed vector x. we refer to this space

    as observation space.

    Because we have assumed that the decision rulemust say , where i = 1, 2, , M, the totalobservation space Z is correspondinglypartitioned intoMdecision regions, denoted by

    Z1, Z2, , ZM

    imm

  • 8/9/2019 DC02 Detection Theory

    47/90

    Maximum Likelihood Detector

    We may restate the decision rule as follow:

    vector x lies inside region Zi if

    is maximum for k = i ln | kf m x x

  • 8/9/2019 DC02 Detection Theory

    48/90

    Maximum Likelihood Detector

  • 8/9/2019 DC02 Detection Theory

    49/90

    Maximum Likelihood Detector

    We may formulate the maximum likelihooddecision rule for AWGN channel as

    vector x lies inside region Zi if

    is maximum for k = i 2

    10

    1 N

    j kjj

    x sN

  • 8/9/2019 DC02 Detection Theory

    50/90

    Maximum Likelihood Detector

    Equivalently, we may restate

    vector x lies inside region Zi if

    is minimum for k = i2

    1

    ( )N

    j kj

    j

    x s

  • 8/9/2019 DC02 Detection Theory

    51/90

    Maximum Likelihood Detector

    Since

    We may rewrite the decision rule

    vector x lies inside region Zi if

    is minimum for k = i

    22

    1

    ( )N

    j kj kj

    x s

    x s

    2

    kx s

  • 8/9/2019 DC02 Detection Theory

    52/90

    Maximum Likelihood Detector

    This show that maximum-likelihooddecision rule is simply to choose themessage point closest to the received

    signal point

  • 8/9/2019 DC02 Detection Theory

    53/90

    Probability of Error

    The average probability of symbol errorequals

    i1 ( sent) ( does not lie inside Z | sent)

    M

    e i iiP P m P m

    x

    i

    1

    11 ( lies inside Z | sent)

    M

    i

    i

    P mM

    x

  • 8/9/2019 DC02 Detection Theory

    54/90

    Probability of Error

    i( lies inside Z | sent) ( | )

    i

    i i

    Z

    P m f m d xx x x

    1

    11 ( | )

    i

    i

    M

    ei Z

    P f m dM

    x x x

  • 8/9/2019 DC02 Detection Theory

    55/90

    Correlation Receiver

    The optimum receiver consists of two subsystems The detector part of the receiver consists of bank of M

    product-integrators or correlators supplied with acorresponding set of coherent reference signals or

    orthonormal basis functions The second part of the receiver, namely the vector

    receiver, is implemented in the form of a maximum-likelihood detector that operates on the observationvector x to produce an estimate of the transmitted

    symbol mi, i = 1, 2, , M, in a way that would minimizethe average probability of symbol error

    m

  • 8/9/2019 DC02 Detection Theory

    56/90

    Correlator Detector

  • 8/9/2019 DC02 Detection Theory

    57/90

    Correlator Receiver

  • 8/9/2019 DC02 Detection Theory

    58/90

    Matched Filter Receiver

    Since each of the orthonormal basis functions1(t), 2(t), , N(t) is assumed to be zero outsidethe interval 0t Tmay be avoided because

    analog multipliers are usually hard to build Consider a linear filter with impulse response hj(t).

    With the receiver signalx(t) used as the filterinput, the resulting outputyj(t)

    dthxty jj )()()(

  • 8/9/2019 DC02 Detection Theory

    59/90

    Matched Filter Receiver

    Suppose

    The resulting filter output is

    )()( tTth jj

    dtTxty jj )()()(

  • 8/9/2019 DC02 Detection Theory

    60/90

    Matched Filter Receiver

    Sampling the output at time t = T, we get

    Since j(t) is zero outside the interval 0t T,finally we get

    Note thatyj(t) = xj, wherexj is thejth

    correlator outputproduced by the received signalx(t)

    dxTy jj )()()(

    T

    jj dxTy0

    )()()(

  • 8/9/2019 DC02 Detection Theory

    61/90

    Matched Filter Receiver

    The detector part of the optimumreceiver may also be implemented asbank of matched filter

    The optimum receiver based on thisdetector is referred as the matched filter

    receiver

  • 8/9/2019 DC02 Detection Theory

    62/90

    Matched Filter Receiver

  • 8/9/2019 DC02 Detection Theory

    63/90

    Matched Filter

    Consider a linear filter of impulse response h(t),with an input that consist of a known signal, (t),and an additive noise component, w(t)

    T is the observation instant

    w(t) is the sample function of a white Gaussian noise

    process of zero mean and power spectral density N0/2

    )()()( twttx Tt0

  • 8/9/2019 DC02 Detection Theory

    64/90

    Matched Filter

    The resulting output,y(t), may be expressed as

    0(t) and n(t) are prodeuced by the signal and noise

    components of the input x(t), respectively

    )()()( 0 tntty

    h d l

  • 8/9/2019 DC02 Detection Theory

    65/90

    Matched Filter

    The output signal component 0(t) be considerablygreater than the output noise component n(t) is to

    have the filter make the instantaneous power in theoutput signal 0(t), measured at time t = T, as large

    as possible compared with the average poser of theoutput noise n(t)

    This is equivalent to maximizing the output signal-to-noise ratio

    )()(

    )(2

    20

    tnE

    TSNR O

    M h d Fil

  • 8/9/2019 DC02 Detection Theory

    66/90

    Matched Filter

    (f) denote the Fourier transform fo the knownsignal (t), andH(f) denote the transfer function

    of the filter

    When the filter output is sampled at time t = T

    dfftjffHt

    2exp)()()(0

    22

    0 2exp)()()( dffTjffHT

    M t h d Filt

  • 8/9/2019 DC02 Detection Theory

    67/90

    Matched Filter

    The power spectral density SN(f) of the outputnoise n(t)

    The average power of the output noise n(t)

    2)(

    2

    )( 0

    fHN

    fSN

    dffStnE N )()(

    2

    dffHN 20 )(2

    M t h d Filt

  • 8/9/2019 DC02 Detection Theory

    68/90

    Matched Filter

    The output signal-to-noise ratio

    2

    2)(

    2

    2exp)()(

    0

    dffHN

    dffTjffH

    SNR O

    M t h d Filt

  • 8/9/2019 DC02 Detection Theory

    69/90

    Matched Filter

    According to Schwarzs inequality

    The output signal-to-noise ratio

    dffdffHdffTjffH222

    )()(2exp)()(

    dff

    NSNR O

    2

    0

    )(2

    M t h d Filt

  • 8/9/2019 DC02 Detection Theory

    70/90

    Matched Filter

    The signal energy given

    The noise power spectral densityN0/2

    dtfdtt22

    )()(

    M t h d Filt

  • 8/9/2019 DC02 Detection Theory

    71/90

    Matched Filter

    Consequently, the output signal-to-noise ratio willbe a maximum whenH(f) is chosen so that the

    equality holds

    The optimum value of this transfer function isdefined by

    dffNSNRO

    2

    0

    max, )(2

    fTjffHop t 2exp)()( *

    Matched Filter

  • 8/9/2019 DC02 Detection Theory

    72/90

    Matched Filter

    The impulse response of matched filter

    Since for real-valued signal (t), we have*(f) = (-f)

    dftTfjfthop t )(2exp)()(

    *

    dftTfjfthopt )(2exp)()( *

    tT

    Matched Filter

  • 8/9/2019 DC02 Detection Theory

    73/90

    Matched Filter

    Properties of Matched filter

    The spectrum of the output signal of a matchedfilter with the matched signal as input is, except

    for a time delay factor, proportional to the energyspectral density of the input signal

    )()()(0 ffHf opt

    fTjf

    fTjff

    2exp

    2exp)()(

    2

    *

    Matched Filter

  • 8/9/2019 DC02 Detection Theory

    74/90

    Matched Filter

    Properties of Matched filter

    The output signal of a matched filter isproportional to a shifted version of the

    autocorrelation function of the input signal towhich the filter is matched

    TtRt )(0

    ERT )0()(0

    Matched Filter

  • 8/9/2019 DC02 Detection Theory

    75/90

    Matched Filter

    Properties of Matched filter

    The output signal-to-noise ratio of a matched filterdepends only on the ratio of the signal energy to

    the power spectral density of the white noise atfilter input

    ENdffNtnE2

    )(2

    )( 02

    02

    00

    2

    max,2

    2 NE

    ENESNR O

    Matched Filter

  • 8/9/2019 DC02 Detection Theory

    76/90

    Matched Filter

    Properties of Matched filter

    The matched filter operation may be separatedinto two matching conditions; namely, spectral

    phase matching that produces the desired outputpeak at time T, and spectral amplitude matchingthat gives this peak value its optimum signal-to-noise density ratio

    )()( ffH fTjfjfHfH 2)(exp)()(

    Detection of Signals WithUnknown Phase In Noise

  • 8/9/2019 DC02 Detection Theory

    77/90

    Unknown Phase In Noise

    Up to this point in our discussion, we have assumed thatthe information bearing signal is completely known at thereceiver.

    In practice, however, it is often found that in addition tothe uncertainty due to the additive noise of a receiver,there is an additional uncertainty due to the randomnessof certain signal parameters

    The usual cause of this uncertainty is distortion in thetransmission medium. The most common random signalparameter is the phase, which is especially true fornarrow-band signals.

    Detection of Signals WithUnknown Phase In Noise

  • 8/9/2019 DC02 Detection Theory

    78/90

    Unknown Phase In Noise

    Synchronization with the phase of thetransmitted carrier may be too costly, andthe designer may simply choose to

    disregard the phase information in thereceived signal at the expense of somedegradation in the noise performance ofthe system

    Detection of Signals WithUnknown Phase In Noise

  • 8/9/2019 DC02 Detection Theory

    79/90

    Unknown Phase In Noise

    Consider a digital communication system inwhich the transmitted signal equals

    Eis the signal energy

    T is the duration of the signaling interval

    fi is an integral multiple of 1/2T

    2

    ( ) cos 2i i

    E

    s t f tT

    0 t T

    Detection of Signals WithUnknown Phase In Noise

  • 8/9/2019 DC02 Detection Theory

    80/90

    Unknown Phase In Noise

    When no provision is made to phase synchronize thereceiver with the transmitter, the received signal will, foran AWGN channel, be of the form

    w(t) is the sample function of a white Gaussian noise process ofzero mean and power spectral density N0/2

    The phase is unknown, and is usually considered to be the

    sample value of a random variable uniformly distributed between0 and 2radians

    2

    ( ) cos 2 iE

    x t f t w tT

    0 t T

    Detection of Signals WithUnknown Phase In Noise

  • 8/9/2019 DC02 Detection Theory

    81/90

    Unknown Phase In Noise

    Using a well-known trigonometric identity, we may write

    Suppose that the received signal x(t) is applied to a pair ofcorrelators; we assume that one correlator is suppliedwith the reference signal and the other issupplied with the reference signal

    For Both correlators, the observation interval is 0 t T

    2 2

    ( ) cos cos 2 sin sin 2i iE E

    x t f t f t w tT T

    0 t T

    2 cos 2 iT f t 2 sin 2 iT f t

    Detection of Signals WithUnknown Phase In Noise

  • 8/9/2019 DC02 Detection Theory

    82/90

    Unknown Phase In Noise

    In the absence of noise, we find that the firstcorrelator output equals and the secondcorrelator output equals

    The dependence on unknown phase may be

    removed by summing the squares of the twocorrelator outputs, and then taking the squareroot of the sum

    When the noise w(t) is zero, the result of these

    operations is simply , which is independent ofthe unknown phase

    cosE

    sinE

    E

    Detection of Signals WithUnknown Phase In Noise

  • 8/9/2019 DC02 Detection Theory

    83/90

    Unknown Phase In Noise

    The suggests that for the detection of asinusodal signal of arbitrary phase, andwhich is corrupted by an additive white

    Gaussian noise, we may use the so-calledquadrature receiver

    This receiver is optimum in the sense that

    it realizes this detection with the minimumprobability of error

    Detection of Signals WithUnknown Phase In Noise

  • 8/9/2019 DC02 Detection Theory

    84/90

    Unknown Phase In Noise

    Detection of Signals WithUnknown Phase In Noise

  • 8/9/2019 DC02 Detection Theory

    85/90

    Unknown Phase In Noise

    To obtain the second equivalent form of thequadrature receiver, suppose we have a filter thatis matched to

    The envelope of the matched filter output isobviously unaffected by the value of the phase

    2 cos 2 is t f tT 0 t T

    Detection of Signals WithUnknown Phase In Noise

  • 8/9/2019 DC02 Detection Theory

    86/90

    Unknown Phase In Noise

    The output of such a filter in response tothe received signalx(t) is given

    0

    2cos 2

    T

    i

    y t x f T t dT

    0

    0

    2cos 2 cos 2

    2 sin 2 sin 2

    T

    i i

    T

    i i

    f T t x f dT

    f T x f dT

    Detection of Signals WithUnknown Phase In Noise

  • 8/9/2019 DC02 Detection Theory

    87/90

    Unknown Phase In Noise

    The envelop of matched filter output, evaluatedat time t = T, will be

    This is just the output of the quadrature receiver.Therefore, the output at time T) of a filter matched to the

    signal , of arbitrary phase ,followed by an envelope detector is the same as thecorresponding ouput of the quadrature receiver

    1/2 2

    0 0

    2 2cos 2 sin 2

    T T

    i i il x f d x f d

    T T

    2 cos 2 is t T f t

    Detection of Signals WithUnknown Phase In Noise

  • 8/9/2019 DC02 Detection Theory

    88/90

    Unknown Phase In Noise

    Detection of Signals WithUnknown Phase In Noise

  • 8/9/2019 DC02 Detection Theory

    89/90

    Unknown Phase In Noise

    The combination of matched filter andenvelope detector is called a noncoherentmatched filter

    Detection of Signals WithUnknown Phase In Noise

  • 8/9/2019 DC02 Detection Theory

    90/90