7.

10 MLE Examples
Well now apply the MLE theory to several examples of
practical signal processing problems.
These are the same examples for which we derived the CRLB
in Ch. 3
1.

Range Estimation

2.

Sinusoidal Parameter Estimation (Amp., Frequency, Phase)

3.
We
Will
Cover

sonar, radar, robotics, emitter location

sonar, radar, communication receivers (recall DSB Example), etc.

Bearing Estimation

4.

sonar, radar, emitter location

Autoregressive Parameter Estimation

speech processing, econometrics

See Book
1

Ex. 1 Range Estimation Problem

Transmit Pulse:

s(t)

nonzero over t[0,Ts]

Receive Reflection:

s(t o)

Measure Time Delay: o

C-T Signal Model

x ( t ) = s ( t o ) + w( t )
$
!#!
"

0 t T = Ts + o,max

s ( t ; o )

s(t)

Bandlimited
White Gaussian

BPF
& Amp

x(t)

PSD of w(t)
No/2

Ts

s(t o)
f

Range Estimation D-T Signal Model

ACF of w(t)

PSD of w(t)
No/2

2 = BNo

f
1/2B

x[n ] = s[n no ] + w[n ]

Sample Every = 1/2B sec

w[n] = w(n)

1/B 3/2B

n = 0,1, , N 1

DT White
Gaussian Noise
Var 2 = BNo

s[n;no] has M non-zero samples starting at no

w[n ]

x[n ] = s[n no ] + w[n ]

w[n ]

no o /

0 n no 1
no n no + M 1
no + M n N 1
3

Range Estimation Likelihood Function

White and Gaussian Independent Product of PDFs
3 different PDFs one for each subinterval
# 3 !!!!&
# 2 !!!!!!!
#1 !!!& (!!!!!!!'
& (!!!!'
(!!!'
no 1
( x[ n ] s[ n no ]) 2 no + M 1
x 2 [n ]
x 2 [ n ] no + M 1
p ( x; no ) = C exp
C exp
C exp

2
2
2
2
n = 0
2
2 n = no

n = no + M

C=

1
2 2

Expand to get an x2[n]

term group it with
the other x2[n] term

N 1 2

x [n ]

no + M 1
1

p ( x; no ) = C N exp n =0 2 exp
(
2
x
[
n
]
s
[
n
n
]
s
[
n
n
])

0
o

2
2
n
=
n

$!o!!!!!
!#!!!!!!!
"

$!!!
#!!!"

Does not depend on no

must minimize this or maximize

its negative over values of no

Range Estimation ML Condition

So maximize this: 2

no + M 1

no + M 1

x[n]s[n n0 ] +

n = no
$!
!!#!!!"

s 2 [n no ]

= no
$n!
!
!#!!!
"

Doesnt depend on no!

Summand moves with
the limits as no changes.

Because s[n no] = 0

outside summation
range so can extend it!
N 1

So maximize this:

x[n]s[n n0 ]

n =0

So. MLE Implementation is based on Cross-correlation:

Correlate Received signal x[n] with transmitted signal s[n]
no = arg max {C xs [m]}
0 m N M

C xs [m] =

N 1

x[n]s[n m],

n =0

Range Estimation MLE Viewpoint

Doesnt depend on no!
Summand moves with
the limits as no changes.

Cxs[m]

m
no
C xs [m] =

Warning: When
signals are complex
(e.g., ELPS) take find
peak of |Cxs[m] |

N 1

x[n]s[n m],

n =0

Think of this as an inner product for each m

Compare data x[n] to all possible delays of signal s[n]
! pick no to make them most alike

Ex. 2 Sinusoid Parameter Estimation Problem

Given DT signal samples of a sinusoid in noise.
Estimate its amplitude, frequency, and phase
x[n ] = A cos( o n + ) + w[n ]
o is DT frequency in
cycles/sample: 0 < o <

n = 0, 1, , N 1
DT White Gaussian Noise
Zero Mean & Variance of 2

Multiple parameters so parameter vector: = [ A o

]T

The likelihood function is:

p ( x; ) = C

1
exp
2
2

N 1

n =0

( x[ n ] A cos( o n + ))

= J ( A, o , )
For MLE: Minimize This

Sinusoid Parameter Estimation ML Condition

To make things easier
Define an equivalent parameter set:
1 = Acos() 2 = Asin()

[1 2 o ]T
Then J'(1 ,2,o) = J(A,o,)

= [1 2]T

Define:
c(o) = [1 cos(o) cos(o2)

cos(o(N-1))]T

s(o) = [0 sin(o) sin(o2)

sin(o(N-1))]T

and
H(o) = [c(o) s(o)]

an Nx2 matrix
8

Then:

J'(1 ,2,o) = [x H (o) ]T [x H (o) ]

Looks like the linear model case except for o dependence of H (o)

Thus, for any fixed o value, the optimal estimate is

= H ( o )H( o )
T

HT ( o )x

Then plug that into J'(1 ,2,o):

J (1 , 2 , o ) = [x H( o ) ]T [x H( o ) ]

= x T T HT ( o ) [x H( o ) ]

1 T

T
= x I H( o ) H ( o ) H( o ) H ( o ) x
!!!!!!!

$
!#!!!!!!!!
"
T

= I H ( o ) HT ( o ) H ( o )

HT ( o )

= x T x x T H( o ) H T ( o ) H( o ) H T ( o ) x
$!!!!!!!#!!!!!!!"
minimize w.r.t. o

Sinusoid Parms. Exact MLE Procedure

Step 1: Minimize this term over o to find o

1 T

T
T

o = arg min x H( o ) H ( o )H( o ) H ( o )x

0o

Step 2: Use result of Step 1 to get

Done Numerically

1 T

)x
= H ( o )H( o ) H (
o
T

Step 3: Convert Step 2 result by solving

1 = A cos()
2 = A sin()

for

A &

10

Sinusoid Parms. Approx. MLE Procedure

First we look at a specific structure:
c T ( o ) x
1

x T H( o ) H T ( o )H( o ) H T ( o )x =

sT ( ) x
o

c T ( o ) c ( o ) c T ( o ) s( o ) c T ( o ) x

sT ( )c ( )
sT ( o )s( o ) sT ( o )x
o
$!!o!!!
!#!!!!!!"

N
2

Then if o is not near 0 or , then approximately

and Step 1 becomes

= arg min 2

o
0o N

N 1

x
[
n
]
exp(
j
n
)
=
arg
min
X
(

o
0
n =0

and Steps 2 & 3 become

DTFT of Data x[n]

2
)
A =
X (
o
N
)
= X (
o
11

The processing is implemented as follows:

Given the data: x[n], n = 0, 1, 2, , N-1
1. Compute the DFT X[m], m = 0, 1, 2, , M-1 of the data

Zero-pad to length M = 4N to ensure dense grid of frequency points

Use the FFT algorithm for computational efficiency

2. Find location of peak

|X()|

Use quadratic interpolation of |X[m]|

3. Find height at peak

Use quadratic interpolation of |X[m]|

X()

4. Find angle at peak

Use linear interpolation of X[m]

12

Ex. 3 Bearing Estimation MLE

Figure 3.8
from textbook:

Emits or reflects
signal s(t)

s(t ) = At cos(2f o t + )
Simple model

Grab one snapshot of all M sensors at a single instant ts:

x[n ] = sn (t s ) + w[n ] = A cos s n + + w[n ]

Same as Sinusoidal Estimation!!
So Compute DFT and Find Location of Peak!!
If emitted signal is not a sinusoid then you
get a different MLE!!

13