e

te
Professional Development Short Course On:

at
m ica

lic
l
ia

up
co pl

er
s. u

at

D
se t D

ot
Practical Statistical Signal Processing using MATLAB

ur No

N
m TI
co A

o
co o

D
s. e
TI D

l
ia
l
l

er
se at
Instructor:

ia
w. a

ur lic

at
er
w ri

M
co up
w ate
A

t
s. Ma
Dr. Steven Kay

I
D

AT
m
M

AT t

I
o
o

ur AT
I

m e
c
AT

w. N
I

co at
w o

co
e
D

e
s

te
t

es pli
AT ca
l

om a
.c lic
u
ria

w. pli

ou D
w

es up
I
e

Ic ot
at

w D

rs D
s
AT N
M

r
ATI Course Schedule: http://www.ATIcourses.com/schedule.htm
ot

ou ot
o
w
N
I

http://www.aticourses.com/practical_statistical_signal.htm

Ic N
w. D
AT

ATI's Practical Statistical Signal Processing:

o

AT Do
w
D

a l
l

w.
er

w al
ia

at

w ri
er

w
IM

e
at

at
IM

AT

IM
AT

AT

349 Berkshire Drive Riva, Maryland 21140

888-501-2100 410-956-8805
Website: www.ATIcourses.com Email: ATI@ATIcourses.com
 Practical Statistical Signal Processing Using MATLAB
with Radar, Sonar, Communications, Speech & Imaging Applications

June 22-25, 2009

Middletown, Rhode Island
$1895 (8:30am - 4:00pm)
"Register 3 or More & Receive $10000 each
Off The Course Tuition."

Summary
This 4-day course covers signal processing
systems for radar, sonar, communications, speech,
imaging and other applications based on state-of-
the-art computer algorithms. These algorithms
Course Outline
1. MATLAB Basics. M-files, logical flow, graphing,
debugging, special characters, array manipulation,
vectorizing computations, useful toolboxes.
2. Computer Data Generation. Signals, Gaussian
noise, nonGaussian noise, colored and white noise,
AR/ARMA time series, real vs. complex data, linear
models, complex envelopes and demodulation.

e
include important tasks such as data simulation,
e

at
3. Parameter Estimation. Maximum likelihood,
at

parameter estimation, filtering, interpolation,

lic
l
ia
om lic

best linear unbiased, linear and nonlinear least

up
detection, spectral analysis, beamforming,
er
.c up

squares, recursive and sequential least squares,

at

classification, and tracking. Until now these

D
minimum mean square error, maximum a posteriori,
es D

IM

algorithms could only be learned by reading the

ot
general linear model, performance evaluation via Taylor
rs ot

N
latest technical journals. This course will take the
om AT
ou N

series and computer simulation methods.

o

mystery out of these designs by introducing the

Ic o

4. Filtering/Interpolation/Extrapolation. Wiener,
.c
AT l D

l
ia
algorithms with a minimum of mathematics and
es te

linear Kalman approaches, time series methods.

er
rs ca

illustrating the key ideas via numerous examples

ia
w. a

at
5. Detection. Matched filters, generalized matched
om er
i
w ri

using MATLAB.
ou pl

IM
w ate

filters, estimator-correlators, energy detectors,

t
Ic u

s. Ma

Designed for engineers, scientists, and other

D

detection of abrupt changes, min probability of error

AT
M

professionals who wish to study the practice of

AT ot

receivers, communication receivers, nonGaussian

rs TI
I

m e
c

statistical signal processing without the headaches,

AT

w. N

ou A

approaches, likelihood and generalized likelihood

co t
w o

this course will make extensive use of hands-on detectors, receiver operating characteristics, CFAR
e

s. ica
D

te
MATLAB implementations and demonstrations. receivers, performance evaluation by computer
t
AT ca

l
l

om a
rs up

Attendees will receive a suite of software source simulation.

.c lic
ria

w. li
Ic

ou D
w

w up

code and are encouraged to bring their own laptops

es up

6. Spectral Analysis. Periodogram, Blackman-

e

Ic ot
e
t

w D

to follow along with the demonstrations.

rs D
a

Tukey, autoregressive and other high resolution

AT N
M

ot

ou ot

Each participant will receive two books methods, eigenanalysis methods for sinusoids in noise.
o
N
I

Ic N
w. D
AT

Fundamentals of Statistical Signal Processing: Vol. I 7. Array Processing. Beamforming, narrowband

o

AT Do
w

and Vol. 2 by instructor Dr. Kay. A complete set of

D

vs. wideband considerations, space-time processing,

w ial
l

notes and a suite of MATLAB m-files will be interference suppression.

w.
er

w al
ia

distributed in source format for direct use or

at

8. Signal Processing Systems. Image processing,

w eri
er

modification by the user.

IM

active sonar receiver, passive sonar receiver, adaptive

at

at
IM

noise canceler, time difference of arrival localization,

AT

IM

channel identification and tracking, adaptive

AT

Instructor
AT

beamforming, data analysis.

Dr. Steven Kay is a Professor of Electrical
Engineering at the University of
Rhode Island and the President of
Signal Processing Systems, a
consulting firm to industry and the
government. He has over 25 years
of research and development
experience in designing optimal
statistical signal processing algorithms for radar,
sonar, speech, image, communications, vibration,
and financial data analysis. Much of his work has
been published in over 100 technical papers and
the three textbooks, Modern Spectral Estimation:
Theory and Application, Fundamentals of
Statistical Signal Processing: Estimation Theory,
and Fundamentals of Statistical Signal
Processing: Detection Theory. Dr. Kay is a Fellow
of the IEEE.
9. Case Studies. Fault detection in bearings,
acoustic imaging, active sonar detection, passive sonar
detection, infrared surveillance, radar Doppler
estimation, speaker separation, stock market data
analysis.
What You Will Learn
To translate system requirements into algorithms
that work.
To simulate and assess performance of key
algorithms.
To tradeoff algorithm performance for
computational complexity.
The limitations to signal processing performance.
To recognize and avoid common pitfalls and traps
in algorithmic development.
To generalize and solve practical problems using
the provided suite of MATLAB code.

Register online at www.ATIcourses.com or call ATI at 888.501.2100 or 410.956.8805 Vol. 97 41

 e
te

at
m ica

lic
l
ia

up
co pl

er
s. u

at

D
se t D

ot
ur No

N
m TI
co A

o
co o

D
s. e
TI D

l
ia
www.ATIcourses.com

l
l

er
se at

ia
w. a

ur lic

at
er
w ri

M
co up
w ate
A

t
s. Ma

I
D

AT
m
M

Boost Your Skills

AT t

I
349 Berkshire Drive

o
o

ur AT
I

m e
c
AT

w. N

Riva, Maryland 21140

I
with On-Site Courses

co at
w o

co
e
D

Telephone 1-888-501-2100 / (410) 965-8805

e
s

te
t

es pli
AT ca
l

om a
Tailored to Your Needs
Fax (410) 956-5785

.c lic
u
ria

w. pli

ou D
w

Email: ATI@ATIcourses.com

es up
I
e

Ic ot
at

w D

rs D
s
AT N
M

The Applied Technology Institute specializes in training programs for technical professionals. Our courses keep you

r
ot

ou ot
o
current in the state-of-the-art technology that is essential to keep your company on the cutting edge in todays highly
w
N
I

Ic N
w. D
AT

competitive marketplace. Since 1984, ATI has earned the trust of training departments nationwide, and has presented
o

AT Do
w
D

on-site training at the major Navy, Air Force and NASA centers, and for a large number of contractors. Our training
a l
l

increases effectiveness and productivity. Learn from the proven best.

i

w.
er

w al
ia

at

w ri
er

For a Free On-Site Quote Visit Us At: http://www.ATIcourses.com/free_onsite_quote.asp

IM

e
at

at
IM

AT

IM
For Our Current Public Course Schedule Go To: http://www.ATIcourses.com/schedule.htm
AT

AT
 References

1. S. Kay, Fundamentals of Statistical Signal

Processing: Estimation Theory, Prentice-Hall, 1993
2. S. Kay, Fundamentals of Statistical Signal

e
te

at
m ica

lic
Processing: Detection Theory, Prentice-Hall, 1998

l
ia

up
co pl

er
3. L. Scharf, Statistical Signal Processing, Addison-
s. u

at

D
se t D

IM

ot
Wesley,
ur No

N
om AT

o
Reading, MA, 1991 (more advanced treatment)
co o

D
.c
TI D

l
ia
es te

l
4. R.N. McDonough, A.D. Whalen, Detection of
l

er
rs ica

m eria
w. a

at
w ri

ou pl

Signals in Noise, Academic Press, New York, 1995

M
w ate
A

t
Ic Du

I
es I M

AT
5. H.L. Van Trees, Detection, Estimation, and
M

AT t

o
o

ur AT
I

m
c
AT

w. N

Modulation Theory, Vol. I, J. Wiley, New York, 1968

e
.

co t
w o

Ic e

a
D

te
(fairly involved but a classic)
t

es li
AT ca
l

om a
rs up
o

.c lic
ria

w. li

6. G.M. Jenkins, D.G. Watts, Spectral Analysis and its

ou D
w

w up

es up
e

Ic ot
at

w D

Applications, Holden-Day, 1968

rs D
AT N
M

ot

ou ot
o
N

7. S. Kay, Modern Spectral Estimation: Theory and

I

Ic N
w. D
AT

AT Do
w
D

Application, Prentice-Hall, 1988

w ial
l

w.
er

w al
ia

8. M.B. Priestley, Spectral Analysis and Time Series,

at

w ri
er

IM

e
at

Academic Press, 1981

at
IM

AT

IM

9. R.A. Monzingo, T.W. Miller, Adaptive Arrays, J.

AT

AT

Wiley, 1980
10. D.H. Johnson, D.E. Dudgeon, Array Signal
Processing, Prentice-Hall, 1993

Provided as part of course materials
 AT
M I
w ate AT
M w ri I
AT at w. a
IM e A TI D
l
ria
at l co o
er w D ur No
ia w o se t D

4. Detection
l w. N s. u
D AT t o co pl
o

7. Case Studies
Ic Du m ica
N ou pl te 1. Matlab Basics
ot

6. Array Processing
w D

5. Spectral Analysis
rs ica
AT w up es te
w. li .c

3. Parameter Estimation
IM AT ca om AT
at Ic e t IM
er our AT at
2. Computer Data Generation

w ial
w s er
es I M ia
w. D . c a t l
AT N o o m eria
AT Ic ot
Summary of Slides

8. Description of MATLAB Programs

IM ou D
D
at rs up o
e es li
. c N
w ri co t a ot
w al
w. m e D
AT Do AT up
Ic N I lic
Slide number

4 -16

M
17 - 49

at
ou ot e
108 - 182
50 - 107

183 - 208

248 - 253
223 - 247 at
209 - 222

rs D er
es up ia
.c lic
om a l
te
 AT
M I
AT w ate
MI w ri
AT at w. a
IM e A TI D
l
ria

Math:
at l co o

Fortran:
er w D ur No
ia w o se t D
l w. N s. u
D AT t o co pl
o Ic Du m ica
N

Signal generation
ou pl te
m files: script files
ot

N=25
w D rs ica
es te
AT w up
w. li .c

f0=0.25
IM AT ca om AT
t
Version: 5.2 for Windows

pi=3.14159
at Ic e IM
o

s[n ] = cos(2 f 0n )
er ur AT at

do 10 I=1,N
w ial
w s er
ia
Fortran vs. MATLAB example:
es I M

MATLAB: f0=0.25;N=25;
w. D . c a t l
AT N o o m eria
AT
MATLAB Basics

Ic ot l
IM ou D
D
at rs up o
e es li
. c N
w ri co t a ot
w al
w. m e D
AT Do AT up

10 s(I)=cos(2*pi*f0*(I-1))
n = 0,1, K, N 1

Ic N I lic
M at

s=cos(2*pi*f0*[0:N-1]);
ou ot
rs D at e
es up er
ia
l
Useful toolboxes: signal processing, statistics, symbolic

.c lic
om a
te
 AT
I M
AT w ate
I
M w ri
AT at w. a
IM e A TI D
l
ria
at l co o
er w D ur No
ia w o se t D
l w. N s. u
D AT t o co pl
o Ic Du m ica
N ou pl te
ot
w D rs ica
es te
AT w up
w. li .c
IM AT ca om AT
at Ic e t IM
er our AT at
w ial
w s er
es I M ia
w. D . c a t l
AT N o o m eria
AT Ic ot
IM ou D l
D
at rs up o
e es li
. c N
w ri co t a ot
w al
w. m e D
AT Do AT up
Ic N I lic
M
MATLAB treats vectors and matrices as elements

at
ou ot
rs D at e
cosine of vector of samples produces a vector output,

er
Notes: pi already defined, [0:N-1] is a column vector,

es up ia
.c lic
om a l
te
 Noise Generation

Simplest model for observation noise is white Gaussian

noise (WGN)

Definition: zero mean, all samples are uncorrelated,

e
te

at
m ica
power

lic
l
ia

up
co pl

er
spectral density (PSD) is flat, and first order
s. u

at

D
se t D

IM

ot
probability density function (PDF) is
ur No

N
om AT
Gaussian

o
co o

D
.c
TI D

l
ia
es te

l
l

er
rs ica

m eria
w. a

at
exp 2 x 2
1 1
w ri

ou pl

M
p( x ) =
w ate
A

PDF: t
Ic Du

I
2
p(x )

es I M

2 2

AT
M

AT t

o
o

ur AT
I

m
c
AT

w. N

e
.

co t
w o

Ic e

a
D

where 2 = variance
s

te
t

es li
AT ca
l

om a
rs up
o

.c lic
ria

w. li

ou D
w

w up

es up
e

Ic ot
at

w D

rs D
AT N
M

ot

ou ot
o

2 = 1
N
I

Ic N
w. D
AT

MATLAB Example:
o

AT Do
w
D

w ial
l

w.
er

4
w al
ia

at

w ri
er

IM

e
at

2
at
IM

AT

IM

] 0
AT

AT

n[
x
-2

-4
0 20 40 60 80 100
n
wgn.m
0 25
0
1
f
o 20
ut
o
s 15
e
m
o
ct 10
u
o
f 5
o
r
e
b 0
m
u -3 -2 -1 0 1 2 3
n x
 AT
M I
w ate AT
I

%
M w ri
w. a
AT at A l
IM e ria TI D
l co o

,
at

x(
)
p
F
P
D
er w D ur No

% wgn.m
ia w o se t D

0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1

-3
l w. N s. u
D AT t o co pl
o m ica

MATLAB code:
N Ic Du te
ot ou pl

-2
w D rs ica
es te
AT w up
w. li .c
IM AT ca om AT
t

-1
at Ic e IM
er our AT at
w ial
w s er
es I M ia

x
0
w. D . c a l
wgn.m

AT N o o t
m eria
AT Ic ot
ou D l

1
IM D
at rs up o
e es li
. c N
w ri co t a ot
w al e D
2

w. m

the time series, histogram, and

AT Do AT up
Ic N I lic
ou ot M at
e
3

at

% This program generates and plots

rs D
default seed and thus generates the same set of
random numbers each time the program is run.
es up er
ia
Note: randn(state,0) sets random number generator to
.c lic
om a l
te
 % estimated PDF for real white
Gaussian noise.
randn('state',0)
x=randn(100,1);
subplot(2,1,1)
plot(x)

e
te

at
xlabel('n')
m ica

lic
l
ia

up
co pl

er
ylabel('x[n]')
s. u

at

D
se t D

IM
grid

ot
ur No

N
om AT
subplot(2,1,2)

o
co o

D
.c
TI D

l
ia
hist(x)
es te

l
l

er
rs ica

m eria
w. a

at
xlabel('x')
w ri

ou pl

M
w ate
A

t
Ic Du

a
ylabel('number of outcomes out of

I
es I M

AT
M

AT t

o
o

ur AT

100')
I

m
c
AT

w. N

e
.

co t
w o

Ic e

a
title('wgn.m')
D

te
t

es li
AT ca
l

om a
rs up
figure
o

.c lic
ria

w. li

ou D
w

w up

es up
e

pdf(x,100,10,-3,3,1)
Ic ot
at

w D

rs D
AT N
M

ot

xlabel('x')
ou ot
o
N
I

Ic N
w. D
AT

ylabel('PDF, p(x)')
o

AT Do
w
D

w ial
l

title('wgn.m')
w.
er

w al
ia

at

w ri
er

IM

e
at

at
IM

AT

IM
AT

AT

% pdf.m
%
function
pdf(x,N,nbins,xmin,xmax,ymax)
%
 AT
I M
AT w ate
I

%
%
%
%
%
%
%
M w ri
w. a
AT at A l
IM e ria TI D
at l co o

grid
N
x
er w D ur No
ia w o se t D
l w. N s. u
D AT t o co pl
o Ic Du m ica
N ou pl te
ot
and plots the

w D rs ica
es te
AT w up
w. li .c
IM AT ca om AT
at Ic e t IM
o

delx=xx(2)-xx(1);
er ur AT at

bar(xx,y/(N*delx))
% Input parameters:

w ial
w s er
es I M ia
w. D . c a t l
AT N o o
% PDF of a set of data.

AT m eria
Ic ot
- Nx1 data array

IM ou D l
D
o
axis([xmin xmax 0 ymax]);
at rs up
e es li N
[y,xx]=hist(x(1:N),nbins);

w ri . co t c a ot
w al
w. m e D
AT Do AT up
- number of data points

Ic N I lic
M at
xmin,xmax,ymax - axis scaling

ou ot
at e
nbins - number of bins (<N/10)

rs D er
% This function subprogram computes

es up ia
.c lic
om a l
te
 Complex White Gaussian Noise

Definition: x [n ] = w1[n ] + jw2 [n ]

where w1[n ] and w2 [n ] are independent of each other

e
and te

at
m ica

lic
l
each one is real WGN with variance of 2 / 2

ia

up
co pl

er
s. u

at

D
se t D

IM

ot
ur No

Mean: E (x [n ]) = 0

N
om AT

o
co o

D
.c
TI D

l
ia
es te

l
l

er
rs ica

Variance: var(x [n ]) = var(w1[n ]) + var(w 2 [n ]) = 2

m eria
w. a

at
w ri

ou pl

M
w ate
A

t
Ic Du

I
es I M

AT
M

AT t

o
o

ur AT
I

m
c
AT

w. N

MATLAB code:

e
.

co t
w o

Ic e

a
D

te
t

es li
AT ca
l

om a
rs up
o

.c lic
% cwgn.m
ria

w. li

ou D
w

w up

es up
e

Ic ot

%
at

w D

rs D
AT N
M

ot

ou ot
% This program generates complex
o
N
I

Ic N
w. D
AT

AT Do
w

white Gaussian noise and

D

w ial
l

w.
er

% then estimates its mean and

w al
ia

at

w ri
er

IM

variance.
e
at

at
IM

AT

IM

%
AT

AT

N=100;
varw=1;
x=sqrt(varw/2)*randn(N,1)+j*sqrt(varw
/2)*randn(N,1);
muest=mean(x)
varest=cov(x)
 AT
M I
AT w ate
M I w ri
AT at w. a
A l

[0,1]
e

PDF.
IM ria TI D
at l co o
D

Example:
w ur No

Laplacian
er
ia w o se t D memoryless
l w. N s. u
D AT t o co pl
o Ic Du m ica
N ou pl te
ot

p( x ) =
w D rs ica
es te
w up

Use the transformation

AT w. li .c
IM AT ca om AT 1
2 2
t

1
at Ic e IM
o
transformation

er

F (x )
Laplacian noise

ur AT at
w ial
w s er
ia

es I M
. c a l

x = F 1 (w )
w. D
AT N o o t
m eria

AT
2
NonGaussian Noise

Ic ot l
IM ou D
D
at rs up o
e es li
c N

exp 2 x

w ri . co t a ot
w al
w. m e D

x
up
Generation: transform WGN using a nonlinear

AT Do AT
Ic N I lic
ou ot M at
e
and F is the cumulative distribution function of the
where w is uniform random variable on the interval

rs D at
es up er
ia
.c lic
om a l
te
 AT
IM
AT w ate
I
M w ri
AT at w. a
IM e A TI D
l
ria
l co o

,
)
at

P
D
n[
]
x
er w D ur No
ia w o se t D

0
0.2
F 0.4
-5

x( 0.6
0

0.8
1
5

-5
0
l w. N s. u
D AT t o co pl
o Ic Du m ica
N te
Example: 2 = 1

ot ou pl
w D rs ica

200
AT w up es te
w. li .c
IM AT ca om AT
at Ic e t IM
er o

400
ur AT at
w ial
w s er
es I M ia

x
n

0
w. D . c a t l
AT N o o
laplaciannoise.m
AT m eria 600
Ic ot l
IM ou D
D
at rs up o
e es li
. c N
w ri co t a ot
800

w al
w. m e D
AT Do AT up
Ic N I lic
ou ot M at
e

5
rs D at
1000

es up er
ia
.c lic
om a l
te
 MATLAB Code:

% laplaciannoise.m
%
% This program uses a memoryless
transformation of a uniform

e
te

at
% random variable to generate a set
m ica

lic
l
ia

up
co pl

er
of independent Laplacian
s. u

at

D
se t D

IM
% noise samples.

ot
ur No

N
om AT
%

o
co o

D
.c
TI D

l
ia
es te

rand('state',0)

l
l

er
rs ica

m eria
w. a

at
varx=1;N=1000;
w ri

ou pl

M
w ate
A

t
Ic Du

a
u=rand(N,1);

I
es I M

AT
M

AT t

o
o

ur AT

for i=1:N
I

m
c
AT

w. N

e
.

co t
w o

Ic e

a
if u(i)>0.5
D

te
t

es li
AT ca
l

om a
rs up
o

.c lic
ria

w. li

ou D
w

w up

es up
e

x(i,1)=sqrt(varx)*(1/sqrt(2))*log(1/(
Ic ot
at

w D

rs D
AT N
M

ot

ou ot
2*(1-u(i))));
o
N
I

Ic N
w. D
AT

else
AT Do
w
D

w ial
l

w.
er

w al
ia

at

w ri
er

x(i,1)=sqrt(varx)*(1/sqrt(2))*log(2*u
IM

e
at

at
IM

(i));
AT

IM
AT

AT

end
end
subplot(2,1,1)
plot(x)
xlabel('n')
ylabel('x[n]')
 AT
I M
AT w ate
I
M w ri
AT at w. a
IM e A TI D
l
ria
at l co o
er w D ur No
ia w o se t D
l w. N s. u
D AT t o co pl
o Ic Du m ica
N ou pl te
ot
w D rs ica
es te
AT w up
w. li .c
subplot(2,1,2)

IM AT ca om AT
at Ic e t IM
er our AT at
pdf(x,N,50,-5,5,1)

w ial
w s er
es I M ia
. a l
axis([0 1000 -5 5]);

w. D
AT N o c o t
AT m eria
Ic ot l
IM ou D
D
o
title('laplaciannoise.m')

at rs up
e es li
. c N
w ri co t a ot
w al
w. m e D
AT Do AT up
Ic N I lic
ou ot M at
e
rs D at
es up er
ia
.c lic
om a l
te
 AT
M I
AT w ate
I
M w ri
AT at w. a
IM e A TI D
l
ria
co o

problem
at l
er w D ur No
w o
Approach:
ia se t D
l w. N s. u
D AT t o co pl
o Ic Du m ica
N ou pl te
ot
w D rs ica
es te
AT w up
w. li .c
IM AT ca om AT
at Ic e t IM
o

b. By computer simulation
er ur AT at
w ial s er

4. Evaluate actual performance

w es I M ia
w. D . c a t l
AT N o o m eria
AT Ic ot
IM ou D l
D
a. Analytically exact or approximate
at rs up o
e es li
. c N
w ri a
3. Choose optimal or suboptimal procedure

co t ot
w al
w. m e D
up
2. Evaluate best possible performance (bounds)

AT Do AT
Solving Parameter Estimation Problems

lic
1. Translate problem into manageable estimation

Ic N I M at
ou ot
rs D at e
es up er
ia
.c lic
om a l
te
 AT
I M
AT w ate
I
M w ri
AT at w. a
IM e A TI D
l
ria
at l co o
er w D ur No
ia w o se t D
l w. N s. u
D AT t o co pl
o Ic Du m ica
N

receive
ot ou pl te
w D

approaching
rs ica
AT w up es te
w. li .c
IM AT ca om AT
at Ic e t IM
er our AT at
s
Physical basis: Doppler effect

w ial er
w es I M ia
(Step 1)

w. D . c a t l
o o
determine speed to within 0.5 mph

AT N m eria
AT Ic ot
IM ou D l
D
rs up o
Radar Doppler Estimation

at
e es li
c N
transmit

w ri . a ot
receive-

co t
w al
w. m e D
up
Problem: Given radar returns from automobile,

AT Do AT
moving away

Ic N I lic
ou ot M at
e
rs D at
es up er
ia
.c lic
om a l
te
 AT
M I
AT w ate
I
M w ri
AT at w. a
IM e A TI D
l
ria
co o

transmit
at l
w D ur No

frequency
er
ia w o se t D
l w. N s. u
D AT t o co pl
o Ic Du m ica
N ou pl te
ot
w D rs ica
es te
w up
Received frequency is

AT w. li .c

v =
v =
IM AT ca om AT
at Ic e t IM
er our AT at
To measure the velocity use
F = F0 +

w ial
w s er
es I M ia
c

2 F0
2 F0
l
2v

w. D . c a
FD

o o t

c F F0
c F F0

and estimate the frequency to yield

AT N
F0
{

AT m eria
Ic ot l
IM ou D
D
at rs up o
e es li
. c N
w ri co t a ot
w al
w. m e D
AT Do AT up
Ic N I lic
ou ot M at
e
where v = velocity, c= speed of light, F0 = sinusoidal

rs D at
es up er
ia
.c lic
om a l
te
 Modeling and Best Possible Performance
(Step 2)

Preprocessing: first demodulate to baseband to produce

the
sampled complex envelope or

e
te

at
m ica

lic
l
ia
s%[n ] = (A / 2) exp( j 2 FD n + )

up
co pl

er
s. u

at

D
se t D

IM

ot
F = 2v F
ur No

N
D
om AT

o
c 0
co o

D
.c
TI D

l
ia
es te

l
l

er
rs ica

m eria
w. a

at
w ri

ou pl

M
Fs = 1/ > 2FD = 2 max F0
2v
w ate
A

t
Ic Du

I
Must sample at
es I M

AT
M

c
AT t

o
o

ur AT
I

m
c
AT

w. N

e
.

co t
w o

Ic e

a
D

te
t

es li
AT ca
l

Example: v max =300 mph, F0 =10.5 Ghz (X-band),

om a
rs up
o

.c lic
ria

w. li

ou D
w

w up

es up
c = 3x108 m/s
e

Ic ot
at

w D

rs D
AT N
M

ot

ou ot
o
N
I

Ic N
w. D
AT

2v max
o

AT Do
w

FD max = F0 9388 Hz
D

w ial
l

w.
er

c
w al
ia

at

w ri
er

IM

e
at

at

Fs > 18, 776 complex samples/sec

IM

AT

IM
AT

AT

How many samples do we need?

Spec: error must be less than 0.5 mph for

 (A / 2)2
SNR = 10 log10 > 10 dB
2

Cramer-Rao Lower Bound for Frequency

e
te

at
m ica

lic
l
ia

up
co pl

er
s. u

at

D
se t D

IM

ot
tells us the minimum possible variance for estimator
ur No

N
om AT

o
co o

very useful for feasibility studies

D
.c
TI D

l
ia
es te

l
l

er
rs ica

m eria
w. a

at
w ri

ou pl

M
w ate

var( fD )
A

t
Ic Du

a(*) (see [Kay 1988])

I
es I M

(2 )2N (N 2 1)

AT
M

AT t

o
o

ur AT
I

m
c
AT

w. N

e
.

co t
w o

Ic e

a
D

te
where fD = FD / Fs , N = number of complex samples,
t

es li
AT ca
l

om a
rs up
o

.c lic
ria

w. li

ou D
w

w up

=linear SNR
.

es up
e

Ic ot
at

w D

rs D
AT N
M

ot

ou ot
o
N
I

Ic N
w. D
AT

cFs
AT Do
w

Since FD = (2v / c )F0 v =

D

w ial

fD
l

w.
er

2F0
w al
ia

at

w ri
er

IM

e
at

at
IM

AT

IM

and we can show that

AT

AT

2
cFs
var(v ) = var( fD )
2F0

For an error of 0.5 mph (0.22 m/s) set

 AT
M I
AT w ate
M I w ri
AT at w. a
IM e A TI D
l
ria
at l co o
er w D ur No
ia w o se t D
l w. N s. u
o

N >

D AT t co pl
o Ic Du m ica
N ou pl te
ot

2
w D rs ica
es te
w up

6
AT w. li .c
IM AT ca om AT
t IM

at Ic e
o
er ur AT at
s

(2 ) var( fD )


w ial er
w es I M ia
. l

1/ 3
a

and finally we have from (*) that

w. D
AT N o c o t
AT m eria
Ic ot l
IM ou D
D v 0.5 v v + 0.5
at rs up o
3 var(v ) = 0.22 var( fD ) = 7.47x10 8

e es li
. c N
w ri co t a ot
w al
w. m e D
99.8%

AT Do AT up

272 samples
Ic N I lic
ou ot M at
e
rs D at
er
v

es up ia
.c lic
om a l
te
 Descriptions of MATLAB Programs

1. analogsim simulates the action of an RC filter on a

pulse

e
te

at
m ica
2. arcov - estimates the AR power spectral density

lic
l
ia

up
co pl

er
using he covariance method for AR parameter
s. u

at

D
se t D

IM

ot
estimation for real data.
ur No

N
om AT

o
co o

D
.c
TI D

l
ia
es te

l
3. arexamples - gives examples of the time series and
l

er
rs ica

m eria
w. a

at
w ri

corresponding power spectral density for various AR

ou pl

M
w ate
A

t
Ic Du

I
es I M
models. It requires the function subprograms:

AT
M

AT t

o
o

ur AT
I

m
c
AT

w. N

gendata.m and armapsd.m.

e
.

co t
w o

Ic e

a
D

te
t

es li
AT ca
l

om a
rs up
o

.c lic
ria

4. armapsd - computes a set of PSD values, given the

w. li

ou D
w

w up

es up
e

Ic ot
at

parameters of a complex or real AR or MA or ARMA

w D

rs D
AT N
M

ot

ou ot
o

model.
N
I

Ic N
w. D
AT

AT Do
w
D

w ial
l

w.
er

5. arpsd - plots the AR power spectral density for some

w al
ia

at

w ri
er

IM

simple cases. The external subprogram armapsd.m is

at

at
IM

AT

IM

required.
AT

AT

6. arpsdexample - estimates the power spectral density

of two real sinusoids in white Gaussian noise using the
periodogram and AR spectral estimators.
It requires the functions subprograms: per.m and
arcov.m.
 7. arrivaltimeest - simulates the performance of an
arrival time estimator for a DC pulse. The estimator is
a running correlator which is the MLE for white
Gaussian noise.

e
te

at
m ica
8. avper - illustrates the effect of block averaging on

lic
l
ia

up
co pl

er
the periodogram for white Gaussian noise.
s. u

at

D
se t D

IM

ot
ur No

N
om AT
9. classicalbayesian - demonstrates the difference

o
co o

D
.c
TI D

l
ia
es te

l
between the classical approach and the Bayesian
l

er
rs ica

m eria
w. a

at
w ri

approaches to parameter modeling.

ou pl

M
w ate
A

t
Ic Du

I
es I M

AT
M

AT t

o
o

ur AT
I

m
c
AT

w. N

10. cwgn - generates complex white Gaussian noise and

e
.

co t
w o

Ic e

a
D

then estimates its mean and variance.

s

te
t

es li
AT ca
l

om a
rs up
o

.c lic
ria

w. li

ou D
w

w up

es up
e

Ic ot
at

11. DClevelhist - generates Figures 1.4, 1.5 in

w D

rs D
AT N
M

ot

ou ot
o

"Fundamentals of Statistical Signal Processing:

N
I

Ic N
w. D
AT

AT Do
w

Detection Theory", S. Kay

D

w ial
l

w.
er

w al
ia

at

w ri
er

IM

12. DCleveltime - generates a data set of white

at

at
IM

AT

IM

Gaussian noise only and also a DC level A in white

AT

AT

Gaussian noise

13. discretesinc plots the graph in linear and dB

quantities of a discrete sinc pulse in frequency
 14. estperform - compares the frequency estimation
performance for a single complex sinusoid in complex
white Gaussian using the peak location of the
periodogram and an AR(1) estimator.

15. Fig35new - computes Figure 3.5 (same as Figure

e
te

at
m ica
4.5) in "Fundamentals of Statistical Signal Processing:

lic
l
ia

up
co pl

er
Detection Theory", S. Kay. The function subprograms
s. u

at

D
se t D

IM

ot
Q.m and Qinv.m are required.
ur No

N
om AT

o
co o

D
.c
TI D

l
ia
es te

l
16. Fig39new - computes Figure 3.9 in "Fundamentals
l

er
rs ica

m eria
w. a

at
w ri

of Statistical Signal Processing: Detection Theory", S.

ou pl

M
w ate
A

t
Ic Du

I
es I M
Kay. The function subprograms Q.m and Qinv.m are

AT
M

AT t

o
o

ur AT
I

m
c
AT

w. N

required.

e
.

co t
w o

Ic e

a
D

te
t

es li
AT ca
l

om a
rs up
o

.c lic
ria

17. Fig77new - computes Figure 7.7 in "Fundamentals

w. li

ou D
w

w up

es up
e

Ic ot
at

of Statistical Signal Processing: Detection Theory", S.

w D

rs D
AT N
M

ot

ou ot
o

Kay.
N
I

Ic N
w. D
AT

AT Do
w
D

w ial
l

w.
er

18. gendata - generates a complex or real AR, MA, or

w al
ia

at

w ri
er

IM

ARMA time series given the filter parameters and

at

at
IM

AT

IM

excitation noise variance.

AT

AT

19. kalman - implementation of the vector state-scalar

observation linear Kalman filter. See (13.50)-(13.54) of
"Fundamentals of Statistical Signal Processing:
Estimation Theory" by S. Kay for more details.
 20. kalmanexample - uses the linear Kalman filter to
estimate the tap weights for a random TDL channel. It
generates Figures 13.16-13.18 in "Fundamentals of
Statistical Signal Processing: Estimation Theory", S.
Kay. It requires the function subprogram kalman.m.

e
te

at
m ica
21. laplaciannoise - uses a memoryless transformation

lic
l
ia

up
co pl

er
of a uniform random variable to generate a set of
s. u

at

D
se t D

IM

ot
independent Laplacian noise samples.
ur No

N
om AT

o
co o

D
.c
TI D

l
ia
es te

l
22. linearmodel - computes the optimal estimator of
l

er
rs ica

m eria
w. a

at
w ri

the parameters of a real or complex linear model.

ou pl

M
w ate
A

t
Ic Du

I
es I M
Alternatively, it is just the least squares estimator.

AT
M

AT t

o
o

ur AT
I

m
c
AT

w. N

e
.

co t
w o

Ic e

a
D

23. linearmodelexample - implements a line fit to a

s

te
t

es li
AT ca
l

om a
rs up
o

.c lic
ria

noise corrupted line. The linear model or least squares

w. li

ou D
w

w up

es up
e

Ic ot
at

estimator is used. The function subprogram

w D

rs D
AT N
M

ot

ou ot
o

linearmodel.m is required.
N
I

Ic N
w. D
AT

AT Do
w
D

w ial
l

w.
er

24. MAexample plots out the PDF of an MA process

w al
ia

at

w ri
er

IM

e
at

at
IM

AT

IM

25. mlevar - computes the mean, variance, PDF of the

AT

AT

MLE for the power of a WGN process and compares it

to the CRLB.

26. montecarloroc - uses a Monte Carlo approach to

determine the detection performance of a Neyman-
Pearson detector for a DC level in WGN. The true
 performance is shown in "Fundamentals of Statistical
Signal Processing: Detection Theory", S. Kay, in Figure
3.9 for d^2=1. The function subprogram roccurve.m is
required.

27. pcar - estimates the frequencies of real sinusoids

e
te

at
m ica
by using the principal component AR approach. Futher

lic
l
ia

up
co pl

er
details can be found in "Modern Spectral Estimation:
s. u

at

D
se t D

IM

ot
Theory and Application", by S. Kay.
ur No

N
om AT

o
co o

D
.c
TI D

l
ia
es te

l
28. pdf - computes and plots the PDF of a set of data.
l

er
rs ica

m eria
w. a

at
w ri

ou pl

M
w ate
A

t
Ic Du

I
es I M
29. per - computes the periodogram spectral estimator.

AT
M

AT t

o
o

ur AT
I

m
c
AT

w. N

Futher details can be found in "Modern Spectral

e
.

co t
w o

Ic e

a
D

Estimation: Theory and Application", by S. Kay.

s

te
t

es li
AT ca
l

om a
rs up
o

.c lic
ria

w. li

ou D
w

w up

es up
e

Ic ot
at

30. perdetectexample - illustrates the detection

w D

rs D
AT N
M

ot

ou ot
o

performance of a periodogram, which is an incoherent

N
I

Ic N
w. D
AT

AT Do
w

matched filter.
D

w ial
l

w.
er

w al
ia

at

w ri
er

IM

31. perexamples - illustrates the capability of the

at

at
IM

AT

IM

periodogram for resolving spectral lines.

AT

AT

32. plot1 plots a sinusoid

33. psk - implements a matched filter receiver for the

detection of a PSK signal. The data are assumed real.
 34. pskexample - illustrates the optimal
detection/decoding of a PSK encoded digital sequence.
The bits are decoded and the probability of error is
computed and compared to the number of actual errors.
The external function subprogram psk.m is required.

e
te

at
m ica
35. Q - computes the right-tail probability

lic
l
ia

up
co pl

er
(complementary cumulative distribution function) for a
s. u

at

D
se t D

IM

ot
N(0,1) random variable.
ur No

N
om AT

o
co o

D
.c
TI D

l
ia
es te

l
36. Qinv - computes the inverse Q function or the
l

er
rs ica

m eria
w. a

at
w ri

value which is exceeded by a N(0,1) random variable

ou pl

M
w ate
A

t
Ic Du

I
es I M
with a probability of x.

AT
M

AT t

o
o

ur AT
I

m
c
AT

w. N

e
.

co t
w o

Ic e

a
D

37. repcorr - implements a replica correlator for either

s

te
t

es li
AT ca
l

om a
rs up
o

.c lic
ria

real or complex data.

w. li

ou D
w

w up

es up
e

Ic ot
at

w D

rs D
AT N
M

ot

ou ot
o

38. repcorrexample - illustrates the replica-correlator.

N
I

Ic N
w. D
AT

AT Do
w

It requires the subprogram repcorr.m.

D

w ial
l

w.
er

w al
ia

at

w ri
er

IM

39. roccurve - determines the ROCs for a given set of

at

at
IM

AT

IM

detector outputs under H0 and H1.

AT

AT

40. sampling plots out an analog sinusoid and the

samples taken

41. seqls - implements a sequential least squares

estimator for a DC level
 in WGN of constant variance.

42. shift - shifts the given sequence by a specified

number of samples. Zeros are shifted in either from
the left or right.

e
te

at
m ica
43. signdetexample - implements a sign detector for a

lic
l
ia

up
co pl

er
DC level in Gaussian-mixture noise. A comparison is
s. u

at

D
se t D

IM

ot
made to a replica correlator, which is just the sample
ur No

N
om AT
mean.

o
co o

D
.c
TI D

l
ia
es te

l
l

er
rs ica

m eria
w. a

at
w ri

44. sinusoid - generates a sinusoid

ou pl

M
w ate
A

t
Ic Du

I
es I M

AT
M

AT t

o
o

ur AT
I

m
c
AT

w. N

45. stepdown - implements the step-down procedure to

e
.

co t
w o

Ic e

a
D

find the coefficients and prediction error powers for all

s

te
t

es li
AT ca
l

om a
rs up
o

.c lic
ria

the lower order predictors given the filter parameters

w. li

ou D
w

w up

es up
e

Ic ot
at

and white noise variance of a pth order AR model. See

w D

rs D
AT N
M

ot

ou ot
o

(6.51) and (6.52). This program has been translated

N
I

Ic N
w. D
AT

AT Do
w

from Fortran into Matlab. See "Modern Spectral

D

w ial
l

w.
er

Estimation" by S. Kay for further details.

w al
ia

at

w ri
er

IM

e
at

at
IM

AT

IM

46. timedelaybfr - implements a time delay

AT

AT

beamformer for a line array of 3 sensors. The emitted

signal is sinusoidal and is assumed to be at
broadside or at 90 degrees (perpendicular to line array).
 AT
I M
AT w ate
I
M w ri
AT at w. a
e A l
IM ria TI D
co o noise.
at l
er w D ur No
ia w o se t D
l w. N s. u
o

noise present.
D AT t co pl
o Ic Du m ica
N ou pl te
ot
w D rs ica
es te
AT w up
w. li .c
IM AT ca om AT
at Ic e t IM
er our AT at
w ial
w s er
es I M ia
w. D . c a t l
AT N o o m eria
AT Ic ot
IM ou D l
D
at rs up o
e es li
. c N
w ri co t a ot
w al e D
47. wgn - generates and plots the time series,

w. m
AT Do AT up
48. wiener - implements a Wiener smoother for

Ic N I lic
ou ot M at
e
rs D at
er
also for predicting an AR(1) signal for no observation

es up ia
histogram, and estimated PDF for real white Gaussian

l
extracting an AR(1) signal in white Gaussian noise and

.c lic
om a
te
 Boost Your Skills
with On-Site Courses
Tailored to Your Needs
The Applied Technology Institute specializes in training programs for technical
professionals. Our courses keep you current in the state-of-the-art technology that is
essential to keep your company on the cutting edge in todays highly competitive

e
te
marketplace. For 20 years, we have earned the trust of training departments nationwide,

at
m ica

lic
and have presented on-site training at the major Navy, Air Force and NASA centers, and for a

l
ia

up
co pl

er
large number of contractors. Our training increases effectiveness and productivity. Learn
s. u

at

D
from the proven best.
se t D

IM

ot
ur No

ATIs on-site courses offer these cost-effective advantages:

N
om AT

o
co o

You design, control, and schedule the course.

D
.c
TI D

l
ia
es te

l
l

er
Since the program involves only your personnel, confidentiality is maintained. You can
rs ica

m eria
w. a

at
freely discuss company issues and programs. Classified programs can also be arranged.
w ri

ou pl

M
w ate
A

t
Ic Du

a
Your employees may attend all or only the most relevant part of the course.

I
es I M

AT
M

AT t

o
o

Our instructors are the best in the business, averaging 25 to 35 years of practical, real-
ur AT
I

m
c
AT

w. N

e
.

world experience. Carefully selected for both technical expertise and teaching ability, they

co t
w o

Ic e

a
D

provide information that is practical and ready to use immediately.

s

te
t

es li
AT ca
l

om a
rs up
Our on-site programs can save your facility 30% to 50%, plus additional savings by
o

.c lic
ria

w. li

ou D
w

w up

.
eliminating employee travel time and expenses.

es up
e

Ic ot
at

w D

rs D
The ATI Satisfaction Guarantee: You must be completely satisfied with our program.
AT N
M

ot

ou ot
o
N
I

Ic N
w. D
AT

We suggest you look at ATI course descriptions in this catalog and on the ATI website.
o

AT Do
w

Visit and bookmark ATIs website at http://www.ATIcourses.com for descriptions of all

D

w ial
l

of our courses in these areas:

w.
er

w al
ia

at

w ri
er

Communications & Computer Programming

IM

e
at

at
IM

Radar/EW/Combat Systems
AT

IM
AT

AT

Signal Processing & Information Technology

Sonar & Acoustic Engineering
Spacecraft & Satellite Engineering
I suggest that you read through these course descriptions and then call me personally, Jim
Jenkins, at (410) 531-6034, and Ill explain what we can do for you, what it will cost, and what
you can expect in results and future capabilities.
Our training helps you and your organization
remain competitive in this changing world.

Register online at www.aticourses.com or call ATI at 888.501.2100 or 410.531.6034