Random Processes and Type

3/31/2013
Random Processes
Dr. Ali Muqaibel
Dr. Ali Hussein Muqaibel 1
Introduction
• Real life: time (space) waveform (desired +undesired)
• Our progress and development relays on our ability to deal with such wave
forms.
• The set of all the functions that are available (or the menu) is call the
ensemble of the random process.
The graph of the function , ,
versus t for fixed, is called a
realization, Sample path, or sample
function of the random process.
For each fixed from the indexed set
, , is a random variable
1
3/31/2013
Formal Definition
• Consider a random experiment specified by the outcomes from some
sample space , and by the probabilities on these events.
• Suppose that to every outcome ∈ ,we assign a function of time
according to some rule: , , ∈ .
• We have created an indexed family of random variables, , , ∈ .
• This family is called a random process (stochastic processes).
• We usually suppress the and use to denote a random process.
• A stochastic process is said to be discrete‐time if the index set is a
countable set (i.e., the set of integers or the set of nonnegative integers).

• A continuous‐time stochastic process is one which is continuous
(thermal noise)
Deterministic and non‐deterministic
Processes
• Non‐deterministic: future values cannot be
predicted from current ones.
– most of the random processes are non‐
deterministic. Sinusoid with random amplitude [‐1,1]
• Deterministic:
• like :
X (t , s)  s cos(2 t )  t  
Y (t , s )  cos(2 t  s )
Sinusoid with random phase ,
2
3/31/2013
Distribution and Density Functions
• A r.v. is fully characterized by a pdf or CDF. How do we characterize
random processes?
• To fully define a random processes, we need dimensional joint
density function.
• Distribution and Density Functions
• First order:
 ;
• Second‐order joint distribution function
 , ; , ,
• Nth order joint distribution function
• ,…, ; ,…, ,….,
• ,…, ; ,…, ,…, ; ,…,
…
Stationary and Independence
• Statistical Independence
• ,…, , ,…, ; ,…, , ́ ,…, ́ =
,…, ,; ,…, ,…, ; ́ , … , ́
• Stationary
– If all statistical properties do not change with time
• First order Stationary Process
• ; ; ∆ , stationary to order one
• =>
• Proof
 ,
 ;
 ;
 Let ∆
 ∆
3
3/31/2013
Cyclostationary
A discrete‐time or continuous‐time random process is said to be
cyclostationary if the joint cumulative distribution function of any set of
samples is invariant with respect to shifts of the origin by integer multiples of
some period
FX ( t1 ), X ( t 2 ),.., X ( t k ) ( x1 , x 2 ,..., x k ) For all , and all

choices of sampling
 FX ( t1  mT ), X ( t 2  mT ),..., X ( t k  mT ) ( x1 , x 2 ,..., x k ). times , , …
We say that is wide‐sense cyclostationary if the mean and autocovariance

functions are invariant with respect to shifts in the time origin by integer
multiples of , that is , for every integer .
m X (t  mT )  m X (t )
C X (t1  mT , t 2  mT )  C X (t1 , t 2 ).
Note that if is cyclostationary, then if follows that is also wide‐sense
cyclostationary.
4
3/31/2013
Cross‐Correlation Function and its properties
• ,
• If X and Y are jointly w.s.s. we may write
.
• Orthogonal processes , 0
• If X and Y are statistically independent
 =
• If in addition to being independent they are at
least w.s.s.

Some properties for
•
• 0 0
• 0 0
• The geometric mean is tighter than the
arithmetic mean
• 0 0 0 0
1
3/31/2013
Measurement of Correlation Function
• In real life, we can never measure the true correlation .
• We assume ergodicity and use portion of the available time.
• Assume ergodicity, no need to prove mathematically “physical sense”
• Assume jointly ergodic => stationary
• Let 0, 2
• Similarly, we may find &
1
Delay 2
2
1 2
Product .
2
Delay
Example
• Use the above system to measure the for .
• 2 cos cos
• cos cos 2 2
• where cos
• cos 2
• If we require the to be at least 20 times less than the largest value of
the true autocorrelation 0.05 0
• 0.05 ⇒
• Wait enough time! Depending on the frequency
2
3/31/2013
Matlab: Measuring the correlation True Variance

1
• % Dr. Ali Muqaibel -1
-40 -30 -20 -10 0 10 20 30 40
• % Measurement of Correlation function Measured
•
•
clear all
close all
T=20, OMEGA=0.2 1
• clc 0
• T=100;
-1
• A=1; -40 -30 -20 -10 0 10 20 30 40
• omeg=0.2; Error
0.1
• t=‐T:T;
• thet=2*pi*rand(1,1);
0
• X=A*cos(omeg*t+thet);
• [R,tau]=xcorr(X,'unbiased'); -0.1
• %R=R/(2*T); -40 -30 -20 -10 0 10 20 30 40
True Variance
• True_R=A^2/2*cos(omeg*tau); 0.5
• Err=A^2/2*cos(omeg*tau+2*thet)*sin(2*omeg*T)/(2*omeg*T);
• subplot(3,1,1) 0
• plot(tau,True_R,tau,R+Err,':')
• title ('True Variance') -0.5
-100 -80 -60 -40 -20 0 20 40 60 80 100
• subplot(3,1,2) Measured
• 0.5
• plot (tau,R,':')
0
• title ('Measured')
• % error
-0.5
• -100 -80 -60 -40 -20 0 20 40 60 80 100
• subplot (3,1,3) Note the error is less than 5% Error
0.05
• plot (tau,Err)
• title ('Error')
0
• % error T=50, OMEGA=0.2
-0.05
-100 -80 -60 -40 -20 0 20 40 60 80 100
Gaussian Random Processes
A random process is a Gaussian random process if the samples

, ,…
are jointly Gaussian random variables for all ,and all choices of , , … .
This definition applies for discrete‐time and continuous‐time processes.
The joint pdf of jointly Gaussian random variables is determined by the vector of
means and by the covariance matrix:
 1 ( X  m )T C 1 ( X  m )
2
e
f X 1 , X 2 ,....., X k ( x1 ,...., xk )  1
where
(2 ) k / 2 C 2
 m X (t1 )   C X (t1 , t1 ) C X (t1 , t 2 ) . . . C X (t1,t k ) 

 .  C (t , t ) C (t , t ) . . . C X (t 2 , t k ) 
 X 2 1 X 2 2
   . . . 
m .  C 
  .
 . . 
 .   . 
m X (t k )  C X (t k ,t1 ) ... C X (t k ,t k ) 
3
3/31/2013
Example iid Gaussian Sequence

Let the discrete‐time random process be a sequence of independent
Gaussian random variables with mean and variance
The covariance matrix for the times , … , is
{C X (t1 , t j )}  { 2 ij }   2 I ,
where 1 when and 0 otherwise, and is the identity matrix. Thus the

corresponding joint pdf is
1  k 
f X 1 ,..... X k ( x1 , x2 ,....., xk )  k
exp  ( xi  m) 2 / 2 2 
(2 2 ) 2  i 1 
 f X ( x1 ) f X ( x2 )... f X ( xk )
Example of a Gaussian Random
Process
• A Gaussian Random Process which is W.S.S. 4 and
25 | |
16
• Specify the joint density function for three r.v. , 1,2,3 … ,
,
• , 1,2,3, …
• 25 16
• 25 16 4
1
• 25 1
1
4
3/31/2013
Complex Random Processes
• A complex random process is given by
•
∗
• ,
• ,
∗
• Note the conjugate
• There could be a factor of in some books
• See example in Peebles
Example Signal Plus Noise
Suppose we observe a process , which consists of a desired signal
plus noise .
Find the cross‐correlation between the observed signal and the desired
signal assuming that and are independent random processes.
RX ,Y (t1 , t2 )  E[ X (t1 )Y (t2 )]

 E[ X (t1 ){ X (t2 )  N (t2 )}]
 E[ X (t1 ) X (t2 )]  E[ X (t1 ) N (t2 )]
 RXX (t1 , t2 )  E[ X (t1 )]E[ N (t2 )]
 RXX (t1 , t2 )  mX (t1 ) mN (t2 )
where the third equality followed
from the fact that and
are independent.
5
3/31/2013
EXAMPLES OF DISCRETE_TIME &

Continuous-Time RANDOM
PROCESSES
See Leon Garcia
Probability, Statistics, and Random Processes
for Electrical Engineers, 3rd Edition
9.5 GAUSSIAN RANDOM PROCESSES,WIENER
PROCESS,AND BROWNIAN MOTION
EXAMPLES OF DISCRETE_TIME RANDOM PROCESSES

iid Random Processes
Let be a discrete‐time random process consisting of a sequence of in‐
dependent, identically distributed (iid) random variables with common cdf
mean and variance . The sequence is called the iid random process.
The joint cdf for any time instants , … . , is given by
FX1 ,... X K ( x1 , x2 ,...., xk )  P[ X 1  x1 , X 2  x2 ,..., X k  xk )

 FX ( x1 ) FX ( x2 )...FX ( xk ),
where for simplicity denotes . The equation above implies that if is

discrete‐values, the joint pmf factors into the product of individual pmf’s, and if is
continuous‐valued, the joint pdf factors into the product of the individual pdf’s.
The mean of an iid process is obtained
m X ( n )  E[ X n ]  m for all n
Thus, the mean is constant.
The autocovariance function is obtained from as follows. If , then
6
3/31/2013
C X ( n1 , n2 )  E[( X n1  m)( X n2  m)]

 E[( X n1  m)]E[( X n2  m)]  0
since and are independent random variables. If then
C X (n1 , n2 )  E[( X n  m) 2 ]   2
We can express the autocovariance of the iid process in compact form as follows:
C X (n1 , n2 )   2 n1n2 ,
 n1n2  1
where if and 0 otherwise
The autocorrelation function of the iid process is:
RX ( n1 , n2 )  C X (n1 , n2 )  m 2
Example : Bernoulli Random Process
Let be a sequence of independent Bernoulli random variables. is then
an iid random process taking on values from the set {0,1}. A realization of such
a process is shown in Figure.
For example, could be an indicator function for the event “ a light bulb fails
and is replaced on day n.”
Since is a Bernoulli random variable, it has mean and variance
1
The independence of the makes probabilities easy to compute. For example,
the probability that the first 4 bits in the sequence are 1001 is
1, 0, 0, 1 1 0 0 1
1
Similarly, the probability that the second bit is 0 and the seventh is 1 is
0, 1 0 1 1
Realization of a Bernoulli process. 1 indicates

that a light bulb fails and is replaced in day .
7
3/31/2013
Sum Processes: The Binomial Counting and Random Walk Processes
(b) Realization of a binomial
process. denotes the
number of light bulbs that
have failed up to time n.
Many interesting random processes are obtained as the sum of a sequence
of iid random variables, , , … .
⋯. , 1,2, …
The sum process ⋯
0 , can be generated in this way.
where 0. We call the sum process. The pdf or pmf of is found using

the convolution .
Note that depends on the “past,” , … , only through , that is,
is independent of the past when is known. This can be seen clearly from the previous
Figure, which shows a recursive procedure for computing . Thus is a Markov process.
Example Binomial Counting Process
Let the be the sequence of independent Bernoulli random variables in
a previous Example, and let be the corresponding sum process. is then the
counting process that gives the number of successes in the first Bernoulli
trials. The sample function for corresponding to a particular sequence of
is shown in the Figure up. If indicates that a light bulb fails and is replaced
on day n, then denotes the number of light bulbs that have failed up to day .
Since is the sum of independent Bernoulli random variables, is a
binomial random variable with parameters n and 1
n
P[ S n  j ]    p j (1  p) n  j for 0  j  n,
 j
and zero otherwise. Thus has mean and variance 1 .
Note that the mean and variance of this process grow linearly with time ( ).
8
3/31/2013
Example One‐Dimensional Random Walk
Let be the iid process of 1 random variable as in the previous

example, and let be the corresponding sum process. is then the
position of the particle at time n.
The random process is an example of a one‐dimensional random walk.
A sample function of is shown in the Figure
The pmf of is found as follows. If there are " 1" in the first trials,
then there are " 1" and 2 .
Conversely, if the number of “ 1" is .
If is not an integer, then cannot equal .
 n k nk
Thus P[ S n  2k  n]   k  p (1  p) for k  {0,1,..., n}
 
(n‐j)/2 (n‐j)/2
n‐j J
Example Sum of iid Gaussian Sequence

Let be a sequence of iid Gaussian random variables with zero mean
and variance . Find the joint pdf of the corresponding sum process at
times and .
The sum process is also a Gaussian random process with mean zero and
variance . The joint pdf of at times and is given by
f S n1 , S n 2 ( y1 , y2 )  f S n 2n1 ( y2  y1 ) f S n1 ( y1 )
1 2 /[ 2 ( n2 n1 )  2 ] 1
 e ( y2  y1 ) e  y1 / 2 n1
2 2
2 (n2  n1 ) 2 2n1 2
9
3/31/2013
EXAMPLES OF CONTINUOUS-TIME RANDOM PROCESSES
Poisson Process
Consider a situation in which events occur at random instants of time at
an average rate of a customer to a service station or the breakdown of a
component in some system. Let be the number of event occurrences
in the time interval [0,t]. is then a nondecreasing, integer‐valued,
continuous‐time random process as shown in Figure.
A sample path of the Poisson
counting process. The event
occurrence times are denoted by
, , … . .The j th interevent time
is denoted by
Poisson Process.. From Binomial
If the probability of an event occurrence in each subinterval is p, then the expected
number of event occurrences in the interval [0,t] is np. Since events occur at a rate of
events per second, the average number of events in the interval [0,t] is also
.Thus we must have that
If we now let → ∞ . . , → 0 and → 0 while remains fixed, then

the binomial distribution approaches a Poisson distribution with parameter . We
therefore conclude that the number of event occurrences in the interval
0, has a Poisson distribution with mean :
, 0,1, …
!
For this reason is called the Poisson process.
Replace p with /
For detailed derivation , please see n
http://www.vosesoftware.com/ModelRiskHelp/index.htm#P P[ S n  j ]    p j (1  p ) n  j for 0  j  n,
robability_theory_and_statistics/Stochastic_processes/Derivi  j
ng_the_Poisson_distribution_from_the_Binomial.htm
10
3/31/2013
Poisson Random Process
• Also known as Poisson Counting Process
• Arrival of customers, failure of parts, lightning,….internet t>0
• Two conditions:
 Events do not coincide.
 # of occurrence in any given time interval is independent of the number in any non overlapping
time interval. (independent increments)
• Average rate of occurrence= .
•
!
, 0,1,2, … 0,
• ∑
!
•
•
• 1
• The probability distribution of the waiting time until the next occurrence is an
exponential distribution.
• The occurrences are distributed uniformly on any interval of time.
http://en.wikipedia.org/wiki/Poisson_process
Joint probability density function for
Poisson Random Process
• The joint probability density function for the poison process at times 0
• , 0,1,2, …
!
• The probability of occurrence over 0, given that events occurred over
0, , is just the probability that events occurred over , , which is
•
!
• For , the joint probability is given by
• , ]
•
! !
• The joint density becomes
• , ∑ ∑ ,
• Example : demonstrate the higher‐dimensional pdf
11
3/31/2013
Example I
Inquiries arrive at a recorded message device according to a Poisson process of rate 15
inquiries per minute. Find the probability that in a 1‐minute period, 3 inquiries arrive
during the first 10 seconds and 2 inquiries arrive during the last 15 seconds.
The arrival rate in seconds is inquiries per second.
Writing time in seconds, the probability of interest is
10 3 60 45 2
By applying first the independent increments property, and then the stationary
increments property, we obtain
P[ N (10)  3 and N (60)  N (45)  2]

 P[ N (10)  3]P[ N (60)  N (45)  2]
 P[ N (10)  3]P[ N (60  45)  2]
(10 4) 3 e 10 4 (15 4) 2 e 15 4

3! 2!
Example II
Find the mean and variance of the time until the arrival of the tenth
inquiry in the previous Example. The arrival rate is 1/4 inquiries per
second, so the inter‐arrival times are exponential random variables with
parameter .
From Tables, the mean and variance of an inter‐arrival time are then
1/ and 1/ , respectively.
The time of the tenth arrival is the sum of ten such iid random variables,
thus 10
E[ S10 ]  10 E[T ]   40 sec

10
VAR[ S10 ]  10VAR[T]   160 sec 2
2
12
3/31/2013
Example Random Telegraph Signal
Consider a random process that assumes the values 1 . Suppose that
0 1 with probability and suppose that then changes polarity
with each occurrence of an event in a Poisson process of rate . The next figure
shows a sample function of .
The pmf of is given by
P[ X (t )  1]  P[ X (t )   1 X (0)  1]P[ X (0)  1]

 P[X (t )  1 X (0)  1]P[ X (0)  1].
Sample path of a random
telegraph signal. The times
between transitions are iid
exponential random variables.
The conditional pmf’s are found by noting that will have the same polarity

as 0 only when an even number of events occur in the interval 0, . Thus
P[ X (t )  1 X (0)  1]  P[ N (t )  even integer]


(t ) 2 j t
 e
j  0 ( 2 j )!
1
 e t {et  e t }
2
1
 {1  e  2t }
2
and X(0) will differ in sign if the number of events in t is odd:

(t ) 2 j 1 t
P[ X (t )  1 X (0)  1]   e
j 0 (2 j  1)!
1
 e t {et  e t }
2
1
 {1  e  2t }.
2
13
3/31/2013
Mean & Variance of the Random
Telegraph Signal
We obtain the pmf for by substituting into :
P[ X (t )  1]  P[ X (t )   1 X (0)  1]P[ X (0)  1]
 P[X (t )  1 X (0)  1]P[ X (0)  1].
11 11 1
P[ X (t )  1]  {1  e  2t }  {1  e  2t } 
22 22 2
1
P[ X (t )  1]  1  P[ X (t )  1] 
2
Thus the random telegraph signal is equally likely to be 1 at any time
The mean and variance of are
m X (t )  1P[ X (t )  1]  (1) P[ X (t )  1]  0 t 0

VAR[ X (t )]  E[ X (t ) 2 ]  (1) 2 P[ X (t )  1]
 (1) 2 P[ X (t )  1]  1
Auto‐covariance of the Random
Telegraph Signal
The autocovariance of is found as follows:
C X (t1 , t 2 )  E [ X (t1 ) X ( t 2 )]
 1P[ X (t1 )  X (t 2 )]  (  1) P[ X ( t1 )  X (t 2 )]
1  2 t 2  t1 1  2 t 2  t1
 {1  e }  {1  e }
2 2
 2 t 2  t1
e
Thus time samples of become less and less correlated as the time between
them increases.
14
3/31/2013
Wiener Process and Brownian Motion
• A continuous‐time Gaussian random process as a limit of a
discrete time process.
• Suppose that the symmetric random walk process (i.e.,
0.5 ) takes steps of magnitude every seconds.
• We obtain a continuous‐time process by letting be
the accumulated sum of the random step process up to
time t.
• is a staircase function of time that takes jumps of
every seconds.
• At time t, the process will have taken jumps, so it is
equal to
• The mean and variance of
 E[ 0
 VAR[ VAR
 We used the fact that 4 1 1
• By shrinking the time between jumps and letting → 0 and →
0 with
• then has a mean and variance
 0

• X(t) is called the Wiener random process. It is used to model
Brownian motion, the motion of particles suspended in a fluid that
move under the rapid and random impact of neighboring particles.
15
3/31/2013
Wiener Process
• As approaches the sum of infinite number of random variables since
→∞
• lim → lim →
• By the central limit theorem the pdf therefor approaches that o a
Gaussian variable with mean zero and variance :
•
• inherits the property of independent and stationary increments from
the random walk process from which it is derived.
• The independent increments property and the same sequence of
steps can be used to show that the autocovariance of X(t) is given by
• , min ,
• Wiener and Poisson process have the same covariance despite the fact that
they are different.
Practice Problem :Poisson Process
• Suppose that a secretary receives calls that
arrive according to a Poisson process with a
rate of 10 calls per hour.
• What is the probability that no calls go
unanswered if the secretary is a way from the
office for the first and last 15 minutes of an
hour?
16
3/31/2013
In class practice: Wide‐Sense
Stationary Random Process
• Let be an iid sequence of Gaussian random variables with
zero mean and variance , and let be the average of two
consecutive values of ,
2
• Find the mean of .
• Find the covariance ,
• What is the distribution of the random variable . Is it
stationary?
17

Random Processes and Type

Uploaded by

Copyright:

Available Formats

Random Processes and Type

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

Random Processes and Type

Uploaded by

Copyright:

Available Formats

3/31/2013

FX ( t1 ), X ( t 2 ),.., X ( t k ) ( x1 , x 2 ,..., x k ) For all , and all

We say that is wide‐sense cyclostationary if the mean and autocovariance

Matlab: Measuring the correlation True Variance

Gaussian Random Processes

A random process is a Gaussian random process if the samples

 m X (t1 )   C X (t1 , t1 ) C X (t1 , t 2 ) . . . C X (t1,t k ) 

Example iid Gaussian Sequence

where 1 when and 0 otherwise, and is the identity matrix. Thus the

RX ,Y (t1 , t2 )  E[ X (t1 )Y (t2 )]

EXAMPLES OF DISCRETE_TIME &

EXAMPLES OF DISCRETE_TIME RANDOM PROCESSES

FX1 ,... X K ( x1 , x2 ,...., xk )  P[ X 1  x1 , X 2  x2 ,..., X k  xk )

where for simplicity denotes . The equation above implies that if is

C X ( n1 , n2 )  E[( X n1  m)( X n2  m)]

We can express the autocovariance of the iid process in compact form as follows:

Realization of a Bernoulli process. 1 indicates

where 0. We call the sum process. The pdf or pmf of is found using

Example Binomial Counting Process

Let be the iid process of 1 random variable as in the previous

Example Sum of iid Gaussian Sequence

EXAMPLES OF CONTINUOUS-TIME RANDOM PROCESSES

If we now let → ∞ . . , → 0 and → 0 while remains fixed, then

P[ N (10)  3 and N (60)  N (45)  2]

The pmf of is given by

P[ X (t )  1]  P[ X (t )   1 X (0)  1]P[ X (0)  1]

The conditional pmf’s are found by noting that will have the same polarity

P[ X (t )  1 X (0)  1]  P[ N (t )  even integer]

m X (t )  1P[ X (t )  1]  (1) P[ X (t )  1]  0 t 0

You might also like