Statistical Signal Processing: ECE 5615 Lecture Notes Spring 201 9

Statistical
Signal
Processing
ECE 5615 Lecture Notes
Spring 2019
Distorted Equalized Estimated

Input Input +1 Data
Equalizer
x[n] y[n] -1 d[n]
-
e[n] +
© 2007–2019
Mark A. Wickert
Chapter 1
Course Introduction/Overview
Contents
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . 1-3
1.2 Where are we in the Comm/DSP Curriculum? . . . . 1-4
1.3 Instructor Policies . . . . . . . . . . . . . . . . . . . . 1-5
1.4 The Role of Computer Analysis/Simulation Tools . . . 1-6
1.5 Required Background . . . . . . . . . . . . . . . . . . 1-7
1.6 Statistical Signal Processing? . . . . . . . . . . . . . . 1-8
1.7 Course Syllabus-1 . . . . . . . . . . . . . . . . . . . . 1-9
1.8 Course Syllabus-2 . . . . . . . . . . . . . . . . . . . . 1-10
1.9 Mathematical Models . . . . . . . . . . . . . . . . . . 1-11
1.10 Engineering Applications . . . . . . . . . . . . . . . . 1-13
1.11 Random Signals and Statistical Signal Processing in
Practice . . . . . . . . . . . . . . . . . . . . . . . . . . 1-14
1-1
CHAPTER 1. COURSE INTRODUCTION/OVERVIEW
1-2 ECE 5615/4615 Statistical Signal Processing

1.1. INTRODUCTION
1.1 Introduction
Course perspective
Course syllabus
Instructor policies
Software tools for this course
Required background
Statistical signal processing overview
ECE 5615/4615 Statistical Signal Processing 1-3

1.2 Where are we in the Comm/DSP

Curriculum?
ECE 2610 ECE 2205 ECE 3205 ECE 3610
Signals & Signals & Signals & Eng. Prob.
Systems Systems I Systems II & Stats.
ECE 5650 ECE 4680 ECE 5625 ECE 4670

Modern DSP Comm. Comm.
DSP Lab Systems I Lab
You are Here!
ECE 5655 ECE 5615 ECE 5630 ECE 5645

Real-Time Statistical Comm. Info Thy &
DSP Signal Proc Systems II Coding
ECE 5610 ECE 5675 ECE 6640 ECE 6620

Random PLL & Spread Detect. &
Signals Applic. Spectrum Estim. Thy.
ECE 5720 ECE 5635 ECE 6650 Wireless Net,

Image Proc,
Optical Wireless Estim. & Sat. Comm,
Comm. Comm. Adapt. Fil. Radar Sys
Courses Offered According to Demand

1.3. INSTRUCTOR POLICIES
1.3 Instructor Policies

Working homework problems will be a very important aspect
of this course
Each student is to his/her own work and be diligent in keeping

up with problems assignments
Homework papers are due at the start of class
If work travel keeps you from attending class on some evening,

please inform me ahead of time so I can plan accordingly, and
you can make arrangements for turning in papers
The course web site
http://www.eas.uccs.edu/wickert/ece5615/
will serve as an information source between weekly class meet-

ings
Please check the web site updated course notes, assignments,

hints pages, and other important course news; particularly on
days when weather may result in a late afternoon closing of the
campus
Grading is done on a straight 90, 80, 70, ... scale with curving
below these thresholds if needed
Homework solutions will be placed on the course Web site in

PDF format with security password required; hints pages may
also be provided

1.4 The Role of Computer Analysis/

Simulation Tools
In working homework problems pencil and paper type solu-
tions are mostly all that is needed
– It may be that problems will be worked at the board by

students
– In any case pencil and paper solutions are still required to
be turned in later
Occasionally an analytical expression may need to be plotted,

here a visualization tool such as Python (via the Jupyter qtcon-
sole or notebook), Mathematica or MATLAB/Octave1 will be
very helpful
Simple simulations can be useful in enhancing your under-

standing of mathematical concepts
The use of Python for computer work is encouraged since it is

fast and efficient at evaluating mathematical models and run-
ning Monte-Carlo system simulations
There are multiple Python based simulation projects embed-

ded throughout the course, as the Hayes text supports this with
MATLAB based exercises at the end of each chapter
– Text MATLAB examples are easily converted to Python
1
https://www.gnu.org/software/octave/

1.5. REQUIRED BACKGROUND
1.5 Required Background

Undergraduate probability and random variables (ECE 3610 or
similar)
Linear systems theory
– A course in discrete-time linear systems such as, ECE

5650 Modern DSP, is highly recommend (a review is pro-
vided in Chapter 2)
Familiarity with linear algebra and matrix theory, as matrix no-

tation will be used throughout the course (a review is provided
in Chapter 2)
A desire to use the Jupyter/IPython Notebook, but using Math-

ematica or back to MATLAB/Octave as needed
Homework problems will require the use of some vector/matrix

computational tool (prefer Python at this point in time), and a
reasonable degree of proficiency will allow your work to pro-
ceed quickly
A desire to dig in!

1.6 Statistical Signal Processing?

The author points out that the text title is not unique, in fact
A Second Course in Discrete-Time Signal Processing is also
appropriate
The Hayes text covers:
– Review of discrete-time signal processing and matrix the-

ory for statistical signal processing
– Discrete-time random processes
– Signal modeling
– The Levinson and Related Recursions
– Wiener and Kalman filtering
– Spectrum estimation
– Adaptive filters
The intent of this course is not entirely aligned with the text
topics, as this course is also attempting to fill the void left
by Random Signals, Detection and Extraction of Signals from
Noise, and Estimation Theory and Adaptive Filtering
A second edition to the classic detection and estimation theory

text by Van Trees is another optional text for 2015; This is the
text I first learned from
Three Steven Kay books on detection and estimation are now

optional texts, and may take the place of the Hayes book in the
future

1.7. COURSE SYLLABUS-1
1.7 Course Syllabus-1

ECE 5615/4615
Statistical Signal Processing
Spring Semester 2019
Instructor: Dr. Mark Wickert Office: EN292 Phone: 255-3500
mwickert@uccs.edu Fax: 255-3589
http://www.eas.uccs.edu/~mwickert/ece5615/
Office Hrs: Wednesday 10:40–11:45 AM or alternative?, other times by appointment.
Required Monson H. Hayes, Statistical Digital Signal Processing and Modeling, John
Texts: Wiley, 1996. Also Python Machine Learning from Packt for $5.
Optional H.L. Van Trees, K.L. Bell, and Z. Tian, Detection, Estimation, and Modulation
Texts: Theory, Part I, 2nd. ed., Wiley, 2013. Steven Kay, Fundamentals of Statistical
Signal Processing, Vol I: Estimation Theory, Vol II: Detection Theory, Vol III:
Practical Algorithm Development, Prentice Hall, 1993/1998. ISBN-978-0-13-
345711-7/978-0-13-504135-2/978-0-13-280803-3.
Optional Scientific Python via the Jupyter Notebook/Lab (installed with https://ana-
Software: conda.org/anaconda/python). See the second page of the syllabus. Mathemat-
ica, available to UCCS students (see course Web Site), is also very useful.
Grading: 1.) Graded homework assignments and computer projects worth 55%.
2.) Midterm exam (takehome) worth 20%.
3.) Final exam with computer sim problems (takehome) worth 25%.
Topics Text Sections

1. Introduction and course overview Notes ch 1
2. Background: discrete-time signal processing, linear algebra 2.1–2.4 (N ch2)
3. Random variables & discrete-time random processes 3.1–3.7 (N ch3)
4. Classical detection and estimation theory N ch4
5. Overview of Rashka & Mirjalili, Python Machine Learning Packt Pub.
6. The Levinson recursion 5.1–5.5 (N ch6)
7. Optimal filters including the Kalman filter; consider Phil Kim, 7.1–7.5 (N ch8)
Kalman Filtering for Beginners with MATLAB Examples.
8. Spectrum estimation 8.1–8.8 (N ch9)
9. Adaptive filtering 9.1–9.4
(N ch10)
Drop Date The last day to drop is April 5, the Friday after Spring break week.

1.8 Course Syllabus-2

Installing Download and Install the Anaconda Python 3.6 Distribution: https://ana-
Python & conda.org/anaconda/python. Do not choose the defaults! I recom-
scikit-dsp- mend you have Anaconda configure paths to the tools. This is not the default.
comm
Jupyter Notebook and QT Console will be installed with the Anaconda distri-
bution, but more formation can be found at: http://ipython.org/.
Download and Install the distributed version control application git: https://
git-scm.com/ on Windows systems (included on macOS and Linux).
Optionally Download and Install a code development environment that also

integrates git version control such as Visual Studio Code (VS Code): https://
code.visualstudio.com/ or PyCharm Community Edition IDE for Python:
https://www.jetbrains.com/pycharm/.
Clone and Install the package scikit-dsp-comm by following the GitHub

README page at: https://github.com/mwickert/scikit-dsp-comm.
Maintain scikit-dsp-comm using git pull origin master (see the informa-
tion in the GitHub link).
Optional The Jupyter notebook is the perfect place to write code, document code, write
Jupyter text using markdown, import figures, and typeset math equations using
Notebook LaTeX syntax. To render a Jupyter notebook as a PDF document a few more
to PDF open source software components are needed:
Conversion • Install Pandoc for file conversion to LaTeX and other formats: https://
pandoc.org/index.html
• Install MikTeX for converting LaTeX documents to PDF on Windows:
https://miktex.org/
• When installing MikTeX be sure to choose the option to automatically
download needed LaTeX packages on-the-fly
• Install TeXLive for converting LaTeX documents to PDF on macOS and
Linux: https://www.tug.org/texlive/
• Install Inkscape for converting embedded SVG graphics in Jupyter note-
books via Pandoc to LaTeX and then PDF: https://inkscape.org/en/
release/0.92.2/. This gives you the ability to have nice looking graphics
in the notebook and easily convert to a PDF, using just the File: Down-
load Notebook menu item. On macOS you just install Inkscape. On
Windows you may have to manually tweak the registry to get Inkscape to
launch by the build script.
Install Typora: As an alternative to installing LaTeX (MikTeX or TeXLive),
install the markdown editor Typora: https://typora.io/. Now you can export
as *.md and then open the file in Typora and save to PDF directly. You can
also do some nice file editing if need be.

1.9. MATHEMATICAL MODELS
1.9 Mathematical Models

Mathematical models serve as tools in the analysis and design
of complex systems
A mathematical model is used to represent, in an approximate

way, a physical process or system where measurable quantities
are involved
Typically a computer program is written to evaluate the math-

ematical model of the system and plot performance curves
– The model can more rapidly answer questions about sys-

tem performance than building expensive hardware pro-
totypes
Mathematical models may be developed with differing degrees

of fidelity
A system prototype is ultimately needed, but a computer sim-

ulation model may be the first step in this process
A computer simulation model tries to accurately represent all

relevant aspects of the system under study
Digital signal processing (DSP) often plays an important role

in the implementation of the simulation model
If the system being simulated is to be DSP based itself, the sim-

ulation model may share code with the actual hardware proto-
type

The mathematical model may employ both deterministic and

random signal models
Formulate
Hypothesis
Define Experiment to
Test Hypothesis
Physical Modify
or Simulation of Model
Process/System
Observations Predictions
Sufficient No
Agreement?
All Aspects
No
of Interest
Investigated?
Stop
The Mathematical Modeling Process2
2
Alberto Leon-Garcia, Probability and Random Processes for Electrical Engineering,

1.10. ENGINEERING APPLICATIONS
1.10 Engineering Applications

Communications, Computer networks, Decision theory and decision
making, Estimation and filtering, Information processing, Power en-
gineering, Quality control, Reliability, Signal detection, Signal and
data processing, Stochastic systems, and others.
Relation to Other Subjects3
Estimation
and Filtering Information
Signal Theory
Processing
;ommunication
!elia2ility = >ireless
!andom Signals
and Systems
:inear
Systems
Bthers
Pro2a2ility
!andom
Aaria2les
Statistics
Decision
Theory
9ame Mathematics
Theory
Addison-Wesley, Reading, MA, 1989

3
X. Rong Li, Probability, Random Signals, and Statistics, CRC Press, Boca Raton, FL, 1999

1.11 Random Signals and Statistical Sig-

nal Processing in Practice
A typical application of random signals concepts involves one
or more of the following:
– Probability
– Random variables
– Random (stochastic) processes
Example 1.1: Modeling with Probability
Consider a digital communication system with a binary sym-

metric channel and a coder and decoder
Input (utput
"–!
! !
Binary !
Channel !
Model
" "
"–!
! = Error!Pro-a-ility
Binary Received
Information Binary Information
Coder Channel Decoder
Communication!System!with!Error!Control
A data link with error correction

1.11. RANDOM SIGNALS AND STATISTICAL SIGNAL PROCESSING IN PRACTICE
The channel introduces bit errors with probability Pe .bit/ D
A simple code scheme to combat channel errors is to repetition

code the input bits by say sending each bit three times
The decoder then decides which bit was sent by using a major-
ity vote decision rule
The system can tolerate one channel bit error without the de-
coder making an error
The probability of a symbol error is given by
Pe .symbol/ D P .2 bit errors/ C P .3 bit errors/
Assuming bit errors are statistically independent we can write
P .2 bit errors/ D .1 / C .1 /
C .1 / D 3 2.1 /
P .3 errors/ D D 3
The symbol error probability is thus
Pe .symbol/ D 3 2 2 3
Suppose Pe .bit/ D D 10 3, then Pe .symbol/ D 2:998

10 6
The error probability is reduced by three orders of magnitude,

but the coding reduces the throughput by a factor of three

Example 1.2: Separate Queues vs A Common Queue

A well known queuing theory result4 is that multiple servers, with
a common queue for all servers, gives better performance than mul-
tiple servers each having their own queue. It is interesting to see
probability theory in action modeling a scenario we all deal with in
our lives.
7ustomers Servers
(ne
*ueue !
:andom ,-aiting line4
Arrivals at ! " Departing
per unit time
7ustomers
,exponentially ...
distriAuted4 :ate
! m
T s = Average Service
Time
Separate
7ustomers Servers
*ueues
!
:andom
Arrivals at ! " Departing
per unit time 7ustomers
,exponentially
...
distriAuted4
m
:ate
Assume customers T s = Average Service
!!m Time
randomly picE Fueues
(ne long serviceBtime customer
forces those Aehind into a long -ait
Common queue versus separate queues

4
Mike Tanner, Practical Queuing Analysis, The IBM-McGraw-Hill Series, New York, 1995.

Common Queue Analysis
The number of servers is defined to be m, the mean customer

arrival rate is per unit of time, and the mean customer service
time is Ts units of time
In the single queue case we let u D Ts D traffic intensity
Let D u=m D server utilization
For stability we must have u < m and < 1
As a customer we are usually interested in the average time in

the queue, which is defined as the waiting time plus the service
time (Tanner)
Ec .m; u/Ts
TQCQ D Tw C Ts D C Ts
m.1 /
Ec .m; u/ C m.1 / Ts
D
m.1 /
where
um=mŠ
Ec .m; u/ D
um=mŠ C .1 / m
P 1 k
kD0 u =kŠ
is known as the Erlang-C formula
To keep this problem in terms of normalized time units, we

will plot TQ =Ts versus the traffic intensity u D Ts
The normalized queuing time is
TQCQ Ec .m; u/ C .m u/ Ec .m; u/

D D C1
Ts m u m u
Separate Queue Analysis
Since the customers randomly choose a queue, arrival rate into

each queue is just =m
The server utilization is D .=m/ Ts which is the same as

the single queue case
The average queuing time is (Tanner)

Ts Ts mTs
TQSQ D D Ts
D
1 1 m
m Ts
The normalized queuing time is
TQSQ m m
D D
Ts m Ts m u
Create the Erlang-C function in MATLAB:
function Ec = erlang_c(m,u);
% Ec = erlangc(m,u)
%
% The Erlang-C formula
%
% Mark Wickert 2001
s = zeros(size(u));
for k=0,
s = s + u.^k/factorial(k);
end
Ec = (u.^m)/factorial(m)./(u.^m/factorial(m) + (1 - u/m).*s);
We will plot TQ =Ts versus u D Ts for m D 2 and 4

>> % m = 2 case
>> u = 0:.05:1.9;
>> m = 2;
>> Tqsq = m./(m - u);
>> Tqcq = erlang_c(m,u)./(m-u) + 1;
>> % m = 4 case
>> u = 0:.05:3.9;
>> m = 4;
>> Tqsq = m./(m - u);
>> Tqcq = erlang_c(m,u)./(m-u) + 1;
The plots are shown below:
&
*
m,# Separate
%
)
TQ
!!!!!!! $
Ts
(
Common
#
'
!
! !"# !"$ !"% !"& ' '"# '"$ '"% '"& #
Traffic Intensity = !T s
Queuing time of common queue and separate queues for m D 2 servers

)
m,'
( Separate
#
TQ
!!!!!!! '
Ts
&
% Common
$
!
! !"# $ $"# % %"# & &"# '
Traffic Intensity = !T s
Queuing time of common queue and separate queues for m D 4 servers
Example 1.3: Power Spectrum Estimation
The second generation wireless system Global System for Mo-

bile Communications (GSM), uses the Gaussian minimum shift-
keying (GMSK) modulation scheme
p 1
X
xc .t / D 2Pc cos 2f0t C 2fd ang.t nTb /
nD 1
where
r
1 2 t 1
g.t / D erf BTb
2 ln 2 Tb 2
r
2 t 1
C erf BTb C
ln 2 Tb 2
The GMSK shaping factor is BTb D 0:3 and the bit rate is
Rb D 1=Tb D 270:833 kbps
We can model the baseband GSM signal as a complex random

process
Suppose we would like to obtain the fraction of GSM signal

power contained in an RF bandwidth of B Hz centered about
the carrier frequency
There is no closed form expression for the power spectrum of a

GMSK signal, but a simulation, constructed in MATLAB, can
be used to produce a complex baseband version of the GSM
signal
>> [x,data] = gmsk(0.3, 10000, 6, 16, 1);

>> [Px,F] = psd(x,1024,16);
>> [Pbb,Fbb] = bb_spec(Px,F,16);
>> plot(Fbb,10*log10(Pbb)); axis([-400 400 -60 20]);
GSM Baseband Power Spectrum

=;
>;
Spectral Density 5d78
9>;
P fraction
9=;
9<;
9:; B
9@;
9?;
9:;; 9<;; 9=;; 9>;; ; >;; =;; <;; :;;
Fre$uency in +H-

Using averaged periodogram spectral estimation we can esti-

mate SGMSK.f / and then find the fractional power in any RF
bandwidth, B, centered on the carrier
R B=2
B=2 SGMSK .f / df
Pfraction D 1
R
1 SGMSK .f / df
– The integrals become finite sums in the MATLAB calcu-

lation
GSM Power Containment vs. RF Bandwidth

#
!&)
Fractional Po1er
!&(
B = 200 kHz ! 95.6%
!&' B = 100 kHz ! 67.8%
B = 50 kHz ! 38.0%
!&$
!
! "! #!! #"! $!! $"! %!!
*F -and1idth in 5H7
Fractional GSM signal power in a centered B Hz RF bandwidth
An expected result is that most of the signal power (95%) is

contained in a 200 kHz bandwidth, since the GSM channel
spacing is 200 kHz

Example 1.4: A Simple Binary Detection Problem

Signal x is measured as noise alone, or noise plus signal
(
n; only noise for hypothesis H0
xD
V C n; noise + signal for hypothesis H1
We model x as a random variable with a probability density

function dependent upon which hypothesis is present
A!Two!Hypothesis px ( X H0 ) px ( X H1 )
Scenario
VT
X
–1 0 V–1 1 V V+1
We decide that the hypothesis H1 is present if x > VT , where

VT is the so-called decision threshold
The probability of detection is given by
Z 1
PD D Pr.x > VT jH1/ D px .X jH1/ dX
VT
px ! X H1 "
VT P D # Pr ! x # v T H 1 "
X
!1 0 V!1 1 V V"1
Area corresponding to PD

We may choose VT such that the probability of false alarm,

defined as
Z 1
PF D Pr.x < VT jH0/ D px .XjH0/ dX
VT
is some desired value (typically small)
px ! X H0 "
P F # Pr ! x # v T H 0 "
VT
X
!1 0 V!1 1 V V"1
Area corresponding to PF
Example 1.5: A Waveform Estimation Theory Problem
Consider a received phase modulated signal of the form
r.t / D s.t / C n.t / D Ac cos.2fc t C .t // C n.t /
where .t / D kp m.t /, kp is the modulator phase deviation

constant, m.t / the modulation, and n.t / is additive white Gaus-
sian noise
The signal we wish to estimate is m.t /, information phase

modulated onto the carrier

We may choose to develop an estimation procedure that ob-

O /, the estimate of m.t / such that the mean square error
tains m.t
is minimized, i.e.,
MSE D Efjm.t / O /j2g

m.t
Another approach is maximum likelihood estimation
A practical implementation of a maximum likelihood based

scheme is a phase-locked loop (PLL)
As a specific example here we consider the MATLAB Simulink

block diagram shown below:
PM PM
Pulse Scope
PM PM
Generator
Modulator Demodulator
Baseband Baseband
Random Re
Number
Im
Real-Imag to
Scope1 Complex
Random
Number1
Check Signal
1 U( : ) num(z) ts(1)(z)
Attributes Im(u) Kv 1
den(z) 1-z -1
Discrete Integrator
Discrete
K u Modulo
Integrator
Ph
Complex baseband Simulink block diagram
Results from the simulation using a squarewave input for two

different signal-to-noise power ratios (20 dB and 0 dB) are
shown below

Input m(t) 0.5
0
0 1 2 3 4 5 6 7 8 9 10
Estimate at 20 dB
0.5
0
0 1 2 3 4 5 6 7 8 9 10
Estimate at 0 dB
0.5
0
0 1 2 3 4 5 6 7 8 9 10
Time
Input/output waveforms
Example 1.6: Nonrandom Parameter in White Gaussian Noise

We are given observations
ri D A C ni ; i D 1; : : : ; N
with the ni iid of the form N.0; n2/

The joint pdf is

N
.Ri A/2

Y 1
prja .RjA/ D exp
2n2
p
iD1 2n2
The log likelihood function is

N
N 1 1 X
ln prja .RjA/ D ln .Ri A/2
2 2n2 2n2 iD1
Solve the likelihood equation

ˇ N ˇ
@ lnŒprja .RjA/ ˇˇ 1 X ˇ
D 2 .Ri A/ˇˇ D0
@A ˇ
ADaO ml n i D1 ADaO ml
N
1 X
H) aO ml D Ri
N i D1
Check the bias:

N
1 X
EfaO ml.r/ Ag D E ri A
N i D1
N
1 X
D Efri g A D 0
N iD1 „ƒ‚…
A
No Bias!

Example 1.7: Sat-Com Adaptive Equalizer
Wideband satellite communication channels are subject to both

linear and non-linear distortion
Transmitter Uplink
Channel
Mod.
Data PSK Modulation Bandpass HPA
Source Mod Impairments Filtering (TWTA)
Z BPSK Z IQ amplitude imbalance Z Spurious PM WGN

Z QPSK Z IQ phase imbalance Z Incidental AM Other
Noise
Z OQPSK Z Waveform asymmetry Z Clock jitter Signals
and rise/fall time (off)
Transponder Downlink
Channel
Bandpass HPA
Mod.
Filtering (TWTA)
Z Phase noise WGN

Z Spurious PM Other
Z Incidental AM Noise
Signals
Z Spurious outputs (on)
Receiver
PSK Demod
Bandpass Adaptive Recovered
Mod.
(bit true with

Filtering Equalizer Data
full synch)
Z Phase noise
Other Z Spurious PM
Signals Z Incidental AM
Z Spurious outputs
Wideband Sat-Comm simulation model
An adaptive filter can be used to estimate the channel dis-

tortion, for example a technique known as decision feedback
equalization

M2 Tap
Real
Decision
FIR
Feedback
M1 Tap + -
Recovered
Complex Re 2 I Data
FIR
Adapt +
Tap CM Error/ Mode DD Error/
Weight LMS Update LMS Update -
Soft I/Q outputs Update
from demod at
sample rate = 2Rs
µCM, µDD CM Error/ DD Error/ -
µDF, γ LMS Update LMS Update
+
M1 Tap Recovered
Complex Im z-1 2 Q Data
FIR + -
Stagger for
OQPSK, omit Decision
Feedback M2 Tap
for QPSK
Real
FIR
An adaptive baseband equalizer5
Since the distortion is both linear (bandlimiting) and nonlin-

ear (amplifiers and other interference), the distortion cannot be
completely eliminated
The following two figures show first the modulation 4-phase

signal points with and with out the equalizer, and then the bit
error probability (BEP) versus received energy per bit to noise
power spectral density ratio (Eb =N0)
5
Mark Wickert, Shaheen Samad, and Bryan Butler. “An Adaptive Baseband Equalizer for
High Data Rate Bandlimited Channels,ï£¡ Proceedings 2006 International Telemetry Conference,
Session 5, paper 06–5-03.

Before Equalization: Rb = 300 Mbps After Equalization: Rb = 300 Mbps

1.5 1.5
1 1
0.5 0.5
Quadrature
Quadrature
0 0
−0.5 −0.5
−1 −1
−1.5 −1.5
−1.5 −1 −0.5 0 0.5 1 1.5 −1.5 −1 −0.5 0 0.5 1 1.5
In−phase In−phase
OQPSK scatter plots with and without the equalizer
300 MBPS BER Performance with a 40/0 Equalizer

−2
10
Semi-Analytic Simulation
−3
10
Probability of Bit Error
−4
10
−5
10
Theory EQ NO EQ
−6
10
4.0 dB 8.1 dB
−7
10
6 8 10 12 14 16 18 20 22 24
Eb/N0 (dB)
BEP versus Eb =N0 in dB

Statistical Signal Processing: ECE 5615 Lecture Notes Spring 201 9

Uploaded by

Copyright:

Available Formats

Statistical Signal Processing: ECE 5615 Lecture Notes Spring 201 9

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Statistical Signal Processing: ECE 5615 Lecture Notes Spring 201 9

Uploaded by

Copyright:

Available Formats

Statistical

Distorted Equalized Estimated

1-2 ECE 5615/4615 Statistical Signal Processing

 Software tools for this course

 Statistical signal processing overview

ECE 5615/4615 Statistical Signal Processing 1-3

1.2 Where are we in the Comm/DSP

ECE 5650 ECE 4680 ECE 5625 ECE 4670

You are Here!

ECE 5655 ECE 5615 ECE 5630 ECE 5645

ECE 5610 ECE 5675 ECE 6640 ECE 6620

ECE 5720 ECE 5635 ECE 6650 Wireless Net,

1-4 ECE 5615/4615 Statistical Signal Processing

1.3 Instructor Policies

 Each student is to his/her own work and be diligent in keeping

 Homework papers are due at the start of class

 If work travel keeps you from attending class on some evening,

 The course web site

will serve as an information source between weekly class meet-

 Please check the web site updated course notes, assignments,

 Homework solutions will be placed on the course Web site in

ECE 5615/4615 Statistical Signal Processing 1-5

1.4 The Role of Computer Analysis/

– It may be that problems will be worked at the board by

 Occasionally an analytical expression may need to be plotted,

 Simple simulations can be useful in enhancing your under-

 The use of Python for computer work is encouraged since it is

 There are multiple Python based simulation projects embed-

– Text MATLAB examples are easily converted to Python

1-6 ECE 5615/4615 Statistical Signal Processing

1.5 Required Background

 Linear systems theory

– A course in discrete-time linear systems such as, ECE

 Familiarity with linear algebra and matrix theory, as matrix no-

 A desire to use the Jupyter/IPython Notebook, but using Math-

 Homework problems will require the use of some vector/matrix

 A desire to dig in!

ECE 5615/4615 Statistical Signal Processing 1-7

1.6 Statistical Signal Processing?

 The Hayes text covers:

– Review of discrete-time signal processing and matrix the-

 A second edition to the classic detection and estimation theory

 Three Steven Kay books on detection and estimation are now

1-8 ECE 5615/4615 Statistical Signal Processing

1.7 Course Syllabus-1

Topics Text Sections

ECE 5615/4615 Statistical Signal Processing 1-9

1.8 Course Syllabus-2

Optionally Download and Install a code development environment that also

Clone and Install the package scikit-dsp-comm by following the GitHub

1-10 ECE 5615/4615 Statistical Signal Processing

1.9 Mathematical Models

 A mathematical model is used to represent, in an approximate

 Typically a computer program is written to evaluate the math-

– The model can more rapidly answer questions about sys-

 Mathematical models may be developed with differing degrees

 A system prototype is ultimately needed, but a computer sim-

 A computer simulation model tries to accurately represent all

 Digital signal processing (DSP) often plays an important role

Software tools for this course

Statistical signal processing overview

Each student is to his/her own work and be diligent in keeping

Homework papers are due at the start of class

If work travel keeps you from attending class on some evening,

The course web site

Please check the web site updated course notes, assignments,

Homework solutions will be placed on the course Web site in

Occasionally an analytical expression may need to be plotted,

Simple simulations can be useful in enhancing your under-

The use of Python for computer work is encouraged since it is

There are multiple Python based simulation projects embed-

Linear systems theory

Familiarity with linear algebra and matrix theory, as matrix no-

A desire to use the Jupyter/IPython Notebook, but using Math-

Homework problems will require the use of some vector/matrix

A desire to dig in!

The Hayes text covers:

A second edition to the classic detection and estimation theory

Three Steven Kay books on detection and estimation are now

A mathematical model is used to represent, in an approximate

Typically a computer program is written to evaluate the math-

Mathematical models may be developed with differing degrees

A system prototype is ultimately needed, but a computer sim-

A computer simulation model tries to accurately represent all

Digital signal processing (DSP) often plays an important role

If the system being simulated is to be DSP based itself, the sim-

The mathematical model may employ both deterministic and

Consider a digital communication system with a binary sym-

The channel introduces bit errors with probability Pe .bit/ D

A simple code scheme to combat channel errors is to repetition

The probability of a symbol error is given by

Assuming bit errors are statistically independent we can write

The symbol error probability is thus

Suppose Pe .bit/ D D 10 3, then Pe .symbol/ D 2:998

The error probability is reduced by three orders of magnitude,

The number of servers is defined to be m, the mean customer

In the single queue case we let u D Ts D traffic intensity

Let D u=m D server utilization

For stability we must have u < m and < 1

As a customer we are usually interested in the average time in

To keep this problem in terms of normalized time units, we

The normalized queuing time is

Since the customers randomly choose a queue, arrival rate into

The server utilization is D .=m/ Ts which is the same as

The average queuing time is (Tanner)

The normalized queuing time is

Create the Erlang-C function in MATLAB:

We will plot TQ =Ts versus u D Ts for m D 2 and 4

The plots are shown below:

The second generation wireless system Global System for Mo-

We can model the baseband GSM signal as a complex random

Suppose we would like to obtain the fraction of GSM signal

There is no closed form expression for the power spectrum of a

Using averaged periodogram spectral estimation we can esti-

An expected result is that most of the signal power (95%) is

We model x as a random variable with a probability density

We decide that the hypothesis H1 is present if x > VT , where

We may choose VT such that the probability of false alarm,

Consider a received phase modulated signal of the form

r.t / D s.t / C n.t / D Ac cos.2fc t C .t // C n.t /