 Chapter 2 concept of stochastic process

necessary

 From the stochastic processes, we can

find the model of some signals and
communication system.
 It is the key basis of the research
direction:
Electronic and communication Engineering.
 2.1 what is stochastic
process

 Random event
 Random variable
 Stochastic process
 What is the random event?
 The result that we can observe is not certain.
Example1:
A coin has two sides, the one is called
obverse side and the other is called reverso.
 Throwing the coin , the obverse side or
the reverso will be upturned.
 Before throwing it ,we can not make
sure which side will be upturned.
 In a word, the result we can observe is
not certain.
Example2:
How many guests will go into the supermarket ,
from 9 to 10 in the morning ?
The result maybe any positive integer (0,1,2,
…),which is not certain.
 Random variable
 Let X(A) represent the functional relationship
between a random event A and a real number.
X(A) is called a random variable.
 The random variable may be discrete or
continuous.
when X(A) is countable , it is discrete;
and X(A) is an interval of a real line , it is
continuous.
  if the obverse side
 1,
 is upturned
X (t )  
0, if the reverso
 is upturned
 The probability of X(t)=1 is 0.5 ;
 The probability of X(t)=0 is 0.5 ;
The random variable X(t) is as following :
 Let X (ti ) denote state of
throwing coin at time ti .
 we observe the state of throwing
coin at successive times
The collection of random variables
X   X (ti ), ti  T  can show process of
throwing coin in any time period.
The collection of random variables
X   X (ti ), ti  T  is called Stochastic
process .
Definition of Stochastic process .
Let X (t ) denote state of
throwing coin at time t .
The collection of random variables
X   X (t ), t  T  is called Stochastic
process .
 The set T is called index set.
 When the index set T is countable ,X is
called a discrete-time process.
 When the index set T is an interval of the
real line ,X is called a continuous-time
process.
 For example , we observe the state of throwing
coin every minute , the index set T is {0,1,2,…}
which is countable , thus the collection of
random variable X   X (t ), t  T  can be
called a discrete-time process.
A Stochastic process X(A,t) can be
viewed as a function of two
variables: random event A and time t.
 For a specific event Aj ,we have a single time
function X ( Aj , t ) which is called a sample
function.
 For a specific time tk , X ( A, tk ) is a random
variable whose value depends on the event A.
 For notational convenience we shall designate
the random process by X (t ) ,and let the
functional dependence upon A be implicit.
 1.2 number character

 Stochastic process X   X (t ), t  T 
Let X(t) denote state of X at time t;
X(t) is the random variable obtained by
observing the Stochastic process at tIme t.
 Definition1 : we define the mean of the Stochastic
process X as
  when X (t ) is a continous


 
xf X t dx
random var iable
m X (t )  E{ X (t )}   + 
 X P{ X (t )  X } when X (t ) is a discrete
 
i=0
i i
random var iable

Where E{} is called the expected value operator

f X ( x)
t is the probability density function of X(t);
P{ X (t )  X i } is the probability distribution of X(t).
 Definition 2 :we define the variance function of
the Stochastic process X as

DX (t )  D{ X (t )}  E{| X (t )  m X (t ) |2 }
 E{| X (t ) |2 } | m X (t ) |2

 It is a measure of the “randomness” of the

Stochastic process X at time t (designated by
X(t)).
 the smaller is DX (t ) , The more the samples of
X(t) concentrate .
 It is equal to the difference between the mean-
square value and the square of the mean.
 Definition 3: we define the autocorrelation function of
the Stochastic process X(t) as:

RX (t1 , t2 )  E{ X (t1 )  X (t2 )}

where X(t1) and X(t2) are random variables obtained
by observing X(t) at times t1 and t2 ,respectively.
 It is a function of two variables,t1 and t2;
 It is the measure of the degree to which two time
samples of the same Stochastic process are related.
 Definition 4: we define the Covariance function of the
Stochastic process X(t) as:
C X (t1 , t2 )  E{[ X (t1 )  m X (t1 )]  [ X (t2 )  m X (t2 )]}
 E{ X (t1 )  X (t2 )}  m X (t1 )  m X (t2 )
where X(t1) and X(t2) are random variables obtained
by observing X(t) at times t1 and t2 ,respectively.
 It is a function of two variables,t1 and t2;
 It is the measure of the degree to which two time
samples of the same Stochastic process are related.
Example:
Consider process X (t )  a  cos(t   ) ,
a,  are constant. RV  is uniform In the
interval (-π, π)
Find the mean function 、 variance function
and autocorrelation function.
 Solution:
The probability density function of RV 
 1
 ,     
f ( )     (  )
 0 other

E{ X (t )}  E{a  cos(t   )}

 a  cos(t   )  f ( )d

 1
  a  cos(t   )  d
 2
0
Continuing…
RX (t , t   )  E{ X (t )  X (t   )}
 E{a  cos(t   )  a  cos(t     )}
a2
 E{  [cos(2t  2   )  cos( )]}
2
2
 a
  [cos(2t  2   )  cos( )]  f ( )d
 2

a2
  cos( )
2
DX (t )  E{| X (t ) |2 } | E{ X (t )} |2
a2
 RX (t , t )  0 
2
Continuing…

C X (t , t   )  E{ X (t )  X (t   )}  m X (t )  m X (t   )
 RX (t , t   )  m X (t )  m X (t   )
a2
  cos( )  0
2
a2
  cos( )
2
Example:
Give a RV  with density f ( ) , and an RV
 is uniform In the interval (-π, π) and
independent of  , we form the process:
X (t )  a  cos(t   )
Find the mean function 、 variance function
and autocorrelation function.
 Solution1:

X (t )  a  cos(t   )
 a[cos(t ) cos   sin(t )sin  )]
E{ X (t )}  E{a[cos(t ) cos   sin(t )sin  )]}
 a  E{cos(t )  cos  }  a  E{sin(t )  sin  )}
 a  E{cos(t )}  E{cos  }  a  E{sin(t )}  E{sin  )}
0
Continue:
The uniform probability density function of RV ( ,  )
 1
 ,     
f ( )     (  )
 0 others

 1
  f ( ),     
f (,  )  f ( )  f ( )   2
 0 others
E{ X (t )}  E{a  cos(t   )}
  a  cos(t   )  f (,  )d d
  1
  [ a  cos(t   )   f ( )d ]d 
  2
 a
  f ( )  [sin(t   )  sin(t   )]d 
 2
0
 Continuing:
RX (t , t   )  E{ X (t )  X (t   )}
 E{a  cos(t   )  a  cos(t     )}
  a 2  cos(t   )  cos(t     )  f (,  )d d
1
  a  [cos(2t    2 )  cos( )]  f (,  )d d
2

2
a2
  E{cos( )}  RX ( )
2
DX (t )  D{ X (t )}  E{| X (t ) |2 } | E{ X (t )} |2
 RX (t ,  0)  0
a2

2
Example:
Consider random series { X (n ), n  0, 1, 2,} ,
For all i,j , X (i ) and X ( j ) are independent to
each other.n
And for all i , E{ X (i )}  0, D{ X (i )}  1
Let Yn = X (n ) , then {Yn ,n  1, 2,} is a
i =1
random series.
Find the mean function and autocorrelation
function.
Solution:

for all i, j
E{ X (i )}  0, D{ X (i )}  1
 E{ X (i )}  E{ X ( j )}  0 i  j
E{ X (i )  X ( j )}  
 E{| X (i ) | }  1
2
i j
n n
E{Yn }  E{ X (i )}   E{ X (i )  0  n  0
i 1 i 1
 Continuing…
RY (n, n  m)  E{Yn  Yn  m }
n nm n nm
 E{ X (i )   X ( j )}  E{ [ X (i ) X ( j )]}
i 1 j 1 i 1 j 1
n nm
  E [ X (i ) X ( j )]
i 1 j 1



 n
E[ X (i ) X ( j )]  1  n  n, i j

 


 n ( n  m )n
E [ X (i ) X ( j )]  0  [n( n  m)  n ]  0, i  j

n
 Distribution of random process
The first-order distribution
for each t , X(t) is a RV with distribution
F ( x, t )  P{ X (t )  x}
This function depends on t, and it equals the
probability of the event { X (t )  x} .
the function F ( x, t ) will be called the first-order
Distribution of the random process X(t) . Its derivative
With respect to x : F ( x, t )
f ( x, t ) 
x
Is the first-order density of random process X(t).
 The second-order distribution of the random
process X(t) is the joint distribution of the RVs
X(t1) and X(t2) .
F ( x1 , x2 ; t1 , t2 )  P{ X (t1 )  x1, X (t 2 )  x 2}
The corresponding density equals
F ( x1 , x2 ; t1 , t2 )
f ( x1 , x2 ; t1 , t2 ) 
x1x2
 The nth-order distribution of the random
process X(t) is the joint distribution of the RVs
X(t1) ,X(t2),…,X(tn) .
F ( x1 , x2 ,, xn ; t1 , t2 ,, tn ) 
P{ X (t1 )  x1 , X (t2 )  x2 ,, X (tn )  xn }

The corresponding density equals

f ( x1 , x2 ,, xn ; t1 , t2 , , tn ) 
F ( x1 , x2 ,, xn ; t1 , t2 ,, tn )
x1x2  xn
 Example:
Throwing the coin every 0.5 second from time t=0,
let :
cos( t ), obverse side is upturned at time t
X (t )  
 2t , reverso is upturned at time t
(1) Find the first-order distribution F ( x ;0.5) , F ( x ;1)
(2) Find the second-order distribution F ( x1 , x2 ;0.5 ,1)
(3) Find the mean function mX (t ) , and the
variance function DX (t )
 Solution:
P{obverse is upturned}  0.5
P{reverso is upturned}  0.5
at time t , the distribution of RV X ( t ) is as :
the event : X (t )=cos( t)  obverse is upturned
and X ( t )=2t  reverso is upturned
as result :
P{ X (t )  cos( t )}=P{obverse is upturned}  0.5 ;
P{ X (t )  2t}=P{reverso is upturned}  0.5

X(t) cos(πt) 2t
P{X(t)} 0.5 0.5
 Continuing…
(1) At time t=0.5
The distribution of RV X(t=0.5)
cos(  0.5)  0, obverse is upturned at time 0.5
X (0.5)  
 2  0.5=1, reverso is upturned at time 0.5
P{ X (0.5)  0}  0.5 , P{ X (0.5)  1}  0.5
so the distribution function of RV X(0.5) is :
 0, if x  0

F ( x,0.5)  0.5, if 0  x  1
 1 if x  1


X(0.5) 0 1
P{X(0.5)} 0.5 0.5
 Continuing…
(1) At time t=1
The distribution of RV X(t=1)
cos(  1)  1, obverse is upturned at time 0.5
X (1)  
 2  1=2, reverso is upturned at time 0.5
P{ X (1)  1}  0.5 , P{ X (1)  2}  0.5
so the distribution function of RV X(1) is :
 0, if x  1

F ( x;1)  0.5, if  1  x  2
 1 if x  2


X(1) -1 2
P{X(1)} 0.5 0.5
(2) RV X(0.5) and X(1) are mutually independent. So ,
the joint probability of X(0.5) and X(1) :

P{ X (0.5)  i, X (1)  j}  P{ X (0.5)  i}  P{ X (1)  j}

as result:
1 1 1
P{ X (0.5)  0, X (1)  1}  P{ X (0.5)  0}  P{ X (1)  1}   
2 2 4
1 1 1
P{ X (0.5)  0, X (1)  2}  P{ X (0.5)  0}  P{ X (1)  2}   
2 2 4
1 1 1
P{ X (0.5)  1, X (1)  1}  P{ X (0.5)  1}  P{ X (1)  1}   
2 2 4
1 1 1
P{ X (0.5)  1, X (1)  2}  P{ X (0.5)  1}  P{ X (1)  2}   
2 2 4
{X(0.5),X(1)}

X(1) -1 2
X(0.5)

0 P{X(0.5)=0, X(1)= -1}=1/4 P{X(0.5)=0, X(1)=2}=1/4

1 P{X(0.5)=1, X(1)= -1}=1/4 P{X(0.5)=1, X(1)=2}=1/4

Continuing…
(3)
m X (t )  E{ X (t )}   X (t )  P{ X (t )}
1 1
 cos( t )   2t 
2 2
E{| X (t ) |2 }   | X (t ) |2  P{ X (t )}
1 2 1
| cos( t ) |   | 2t | 
2

2 2
DX (t )  E{| X (t ) |2 } | E{ X (t )} |2
1 2 1 1 1 2
| cos( t ) |   | 2t |   [cos( t )   2t  ]
2

2 2 2 2
1
 [ cos( t )  1]2
2
independent-increment process
Definition
{ X (t ), t  T }is a given process.
t1  t2  t3  t4  T
RV ( X (t2 )  X (t1 )) and RV ( X (t4 )  X (t3 ))
are independent to each other.
 Stationary processes (strict-sense)

 Definition
A stochastic process X(t) is called strict-
sense stationary (abbreviated SSS ). This
means that the processes X(t) and X(t+c)
have the same statics for any c.
from the definition it follows that the nth-
order density of an SSS process must be
such that
 f ( x1 , x2 ,, xn ; t1 , t2 , , tn ) 
f ( x1 , x2 ,, xn ; t1  c, t2  c, , tn  c )
For any c and n .
Stationary processes (wide-sense)
 A stochastic process X(t) is called wide-
sense stationary (abbreviated WSS ) if it
satisfy the following condition:
1. The mean function is constant
m X (t )  E{ X (t )}  constant

2. its autocorrelation depends on only   t2  t1

R(t1 , t2 )  E{ X (t1 )  X (t2 )}  R(  t2  t1 )

 Continuing…
Note in particular that :
E{| X (t ) | }  R(0)
2

Thus the average power of a stationary process is

independent to t and it equal R(0).
 Example
Consider random process X (t )  a cos(t   )
A and  are constant , RV  is uniform
In the interval (-π, π) .
(1) m X (t )  E{ X (t )}  0
a2
(2) RX (t, t   )  E{ X (t )  X (t   )}  cos( )
2
X (t )  a cos(t   ) is WSS
 Example
consider the random process X (t )  A  cos( t ) ,
R.V A follows normal distribution , the mean of
A is 0 , the variance is  2 .
EA  0, DA   2
E{ A2 }  DA | EA |2   2
(1)m X (t )  E{ X (t )}
 E{ A  cos( t )}
 cos( t )  E{ A}  0
(2) RX (t , t   )  E{ X (t )  X (t   )}
 E{ A cos( t )  A cos( t   )}
 cos( t ) cos( t   ) E{ A2 }
1
  [cos(2 t   )  cos( )]   2
2
X (t )  A  cos( t ) is not WSS.
Example:
Consider random series { X (n ), n  0, 1, 2,} ,
For all i,j , X (i ) and X(j) are independent to
each other. And for all i E{ X (i )}  0, D{ X (i )}  1
n
Let Yn = X (n ) , then {Yn ,n  1, 2,} is a
i =1

random series.
(1) mY ( n )  0 ;
(2) if n  0, m  0 RY ( n, n  m )  n .
{Yn ,n  1, 2,} is not WSS.
 Weiner process
W  {W (t ),   t  } is called a Weiner
process if it satisfy the following condition:
(1)W(0)=0;
(2)Independent-increment process;
(3)For all t, s , W (t )  W ( s ) ~ N (0,  2  | t  s |),  2  0
W is also called Brown motion which is a model
of noise in communication and molecule
motion .
 Property of weiner process
(1)For all t,
W (t ) ~ N (0,   | t |)
2

1 w2
fW ( w)   exp{ 2 }
  2 2
(2)The autocorrelation :

0t  s
RW (t , s )  E{W (t )  W ( s )}   2  t
 Example:
W  {W (t ),   t  } is a weiner process
with paremeter  2 ,let X (t )  e  tW (e 2 t ),   0
 is a constant.
Find the mean m X (t ) and the autocorelation
RX (t , t   ),  0, t  0
Solution:

W (t ) ~ N (0,   | t |)
2

E{W (t )}  0, D{W (t )}   2  | t |
m X (t )  E{ X (t )}
 t 2 t  t 2 t
 E{e W (e )}  e  E{W (e )}
0
 Continuing…

RX (t , t   )  E{ X (t )  X (t   )}
 E{e  tW (e 2 t )  e  ( t  )W (e 2 t 2 )}
 t  ( t  ) 2 t 2 t  2
e e  E{W (e )  W (e )}
 e 2 t 2   2  e 2 t
  2  e 2
 Assignment 3: question1
 Random process X (t )  Vt  b , b is a constant ,
R.V V follows normal distribution , and EV  0, DV  1
(1) Find the mean function m X (t ) 、 variance
function DX (t ) and autocorrelation function
RX (t , t   ), t  0,  0 .
(2) Find the first-order probability density function
f X (t ) .
 Question 2
 Random process X (t )  A  cos t  B  sin t , 
is constant , R.V A and B are mutually
independent. EA  EB  0, DA  DB   2 .
Both A and B follows normal distribution.
(1) Find the mean function mX (t ) 、 variance
function DX (t ) and autocorrelation function
RX (t , t   ) .
(2) Find the first-order probability density function
f X (t )

(3) Is the Random process X (t ) WSS ?