Technique to understand geometric representation and represent the
basis functions
Generalized Equations for Orthonormal basis functions
• We write generalized equation for orthonormal basis
functions φi (t) as:
x(t)
Conversion of the Continuous AWGN channel into a vector channel
Consider that the noisy received signal is expressed as a random process and
Consider the input to the correlators is the received signal x(t) defined in
accordance to AWGN channel
w( t) is sample of white Gaussian process having zero mean and PSD No/2.
Output of each correlator is the sample value of a random variable Xj, whose
sample values are
Substituting;
The output is random due to channel noise at the receiver input.
Received signal can be represented by in a vector form .The correlator
output can be characterized by mean and variance
Optimum Receivers using Coherant detection
(Detection of signals & Maximum Likelihood Decoding)
Consider a source producing M messages m1,m2,m3…mM that
are equiprobable, Let corresponding transmitted signals be
s1(t),s2(t),s3(t)…sM(t).For AWGN channel a sample function of
received Random process X(t) is given by
•Vector x and w – sample values of the random vectors X and W.
•Also si – vector representing si(t) & tip of this is a point in the N dimensional
Euclidean space
Co-ordinates of this vector si – transmitted signal point (mssg pt)
•Received signal x(t) –represented by a vector x & tip of this is a point in the
N dimensional Euclidean space – Received signal point
•Vector x – observation vector
•Detection problem – Given the N dimensional observation vector x we have
to determine the most likely transmitted signal si(t) or symbol mi such that the
avg probab of error is minimized.
The decision rule is obtained based on the foll.. analysis
a)Effect of noise perturbation b) location of received signal point
This decision rule is called Maximum a posteriori probability (MAP)
rule and can be expressed in terms of likelihood function
Since log function is a monotonically increasing function it is
convenient to use natural log(log likelihood function)
LIKELIHOOD FUNCTIONS
The set of N correlator outputs can be expressed as a vector X of N
random variables
Substituting- The likelihood functions of an AWGN channel are defined by:
Gaussian pdf :
Using the above equation we can make a decision regarding which
symbol mi is transmitted ; i= 1,2,3….M
•Log likelihood function is maximum when right hand side of equation is
maximum .
•..Since 2nd term negative,max value when magnitude is minimum. This leads to
decision rule as
Square of the Euclidean distance between received signal pt x and signal
pt sk
Observation vector x lies in region Zi if Euclidean distance is minimum for k=i
i.e Maximum likelihood decision rule is to choose the mssg point closest to
the received signal point
Hence expanding the summation;
