STOCHASTIC PROCESSES - ANALYSIS AND PROCESSING

advertisement

7. ANALYSIS AND PROCESSING OF RANDOM

PROCESSES

Basic Calculus of Random processes (Mean Square Calculus)

Stochastic continuity

A random process

X(t )

is said to be continuous in mean square or mean square continuous if:

 lim

0

E

X

 t

  

X

  

2

0

We can write the expression under the limes as:

E

E

X

R

X

X

 t

 t

2

 t

, t

X

 

 

2

X

2

2

 t

R

X

 t

  

 

, t

X

2

R

X

  

 

The random process

X(t )

is continuous if, and only if, its autocorrelation function

R

X

(t

1

,t

2

) is continuous.

Stochastic Processes – Analysis and Processing

7-1

Note: The mean square continuity of does not imply that the sample functions of are continuous, e.g. Poisson process is mean square continuous, but sample functions of the Poisson process have a countably infinite number of discontinuities.

For the Poisson process X(t ) with the rate

> 0 , we have:

E

X

 t

  

X

  

2

   

    

2

0

0

Stochastic Processes – Analysis and Processing

7-2

If

X (t )

is WSS process, then it is mean square continuous only, and only if, its autocorrelation function

R

X

( continuous for

= 0

.

) is

For WSS process we can write the previous expression as:

E

X

R

X

2 R

X

 t

 t

 

2

( t

R

X

X

 

 

)

2

2 R

X

 t

   t

R

X

 

 lim

0

E

X

 t

  

X

  

2

 lim

0

R

X

 

R x

  

0

Stochastic Processes – Analysis and Processing

7-3

If

X (t )

is a mean square continuous process, then its mean is also continuous (but not mean square continuous):

 lim

0

X

( t

 

)

 

X

( t ) which can be written as

 lim

0

E

X

 t

 

 

E

 lim

)

X

 t

 

Stochastic Processes – Analysis and Processing

7-4

Stochastic derivatives

The random process

X (t )

has a mean square derivative X

(t )

if l

.

i .

m

0

X

 t

  

X

 

X

( t ) where l.i.m

denotes limit in the mean (square), that is

 lim

0

E

X

 t

  

X

 

X

( t )

 2

0

Stochastic Processes – Analysis and Processing

7-5

The mean and the autocorrelation function of

X

(t )

are given by:

E

X

   

 d dt

E

X

  

  

X

 

R

X

 t

1 t

2

E

X

    

1 2

 2

R

X

 

1

 t

1

 t

2

2

Stochastic Processes – Analysis and Processing

7-6

Stochastic integrals

A mean square integral of a random process

X (t )

is defined by:

Y t

1  

X

  d

 t

0

  l .

 t i i .

m

0 where t

0

 t

1

Stochastic Processes – Analysis and Processing

  t and

 t i

7-7

 i

X

  i

 t i

 t i

1

 t i

 t

 t

0

X

  d

R

Y

 t

1

, t

2

E

 t

 t

1

0

X

  d

 t

2

X

  d

 t

0

0 t t

1   2 t t

0

E

X

     d

 d

7-8 Stochastic Processes – Analysis and Processing

0 t t

1   2 t t

0

R

X

 

,

  d

 d

Power spectral densities

The power spectral density (or just power spectrum

)

S

X

(

) of a continuous time random process

X(t )

is defined as the

Fourier transform of

R

X

(

)

:

S

X



 

R

X

  e

 j

 d

R

X

1

2

S

X

  e

 j

 d

These two relations are known as the Wiener-Khinchin relations .

Stochastic Processes – Analysis and Processing

7-9

White Noise

A continuous time white noise process

W(t )

is a WSS zeromean random process, with the autocorrelation function given by:

R

W

 

 

2

( t )

The power spectrum of the white noise is obtained to be:

S

W

 

2



 

   e

 j

 d

  

2

Stochastic Processes – Analysis and Processing

7-10

Response of a Continuous Time Linear Time-Invariant (CT LTI)

System to Random Inputs

X(t )

CT LTI

System

Y(t )

If input to the system

X(t )

is a random process,, then the output

Y(t )

will also be a random process, given by:

Y

 

 h

   t

   d

 

Stochastic Processes – Analysis and Processing

7-11

Also:

R

Y t

1

,

2

E

Y

 

 h

 

E

X

 t

    d

 

 

 

   h

    

X t

1

 

, t

2

   d

 d

If the input

X(t )

is WSS process, then:

E Y t

 

X



 h

  d

  

X

H

   

 h

  e

 j

 t dt

R

Y





 

   h

     

X

     d

 d

Stochastic Processes – Analysis and Processing

7-12

The power spectrum of

Y(t )

is given by:

S

Y

 



R

Y

  e

 

  j

 d

 

H

 

2

S

X

 

When the autocorrelation function of the output

R

Y

(

)

is required, than it is easier to obtain it from the power spectrum

S

Y

(

)

:

R

Y

2

1

S

Y

  e

 j

 d

H

 

2

S

X

  e

 j

 d

Stochastic Processes – Analysis and Processing

7-13

30

Stochastic Periodicity

A continuous time random process

X(t )

is mean square periodic , with period

T

, if

E

X

 t

T

X

  

2

0

If

X(t )

is WSS, then

X(t )

is mean square periodic if, and only if, its autocorrelation function is periodic with period

T

:

R

X

  

T

R

X

 

Stochastic Processes – Analysis and Processing

7-14

Download