0 Up votes0 Down votes

13 views34 pagesrandom signals notes for mtech students who ever

Apr 17, 2015

© © All Rights Reserved

PDF, TXT or read online from Scribd

random signals notes for mtech students who ever

© All Rights Reserved

13 views

random signals notes for mtech students who ever

© All Rights Reserved

- Analogue Signals on the arduino
- PAM TDM
- Analog vs Digital Communication
- papershort
- Assignment 3
- A Survey: Feature Extraction of ECG Signals using Haar Wavelet
- Random Processes Ma1254
- v6-14
- ABET PTU Syllabus Digital Control(JAFAR JALLAD)
- Reports Hs
- ece308-1
- 3
- Some New Probability Density Functions
- NI CaseStudy Cs 14031
- comp.pdf
- 6605BE 2071 Bhadra New Regular II Part
- Time Averages and Ergodicity
- 2141003.pdf
- Courant_11_2011
- itc2001-ctl

You are on page 1of 34

Jan Cernock

y

UPGM

FIT VUT Brno, cernocky@fit.vutbr.cz

Random signals

deterministic signals (can be represented by an equation) have one substantial

drawback they carry very little information (for example cosine: amplitude,

frequency, phase shift).

real-world signals are tough to be described as deterministic (for example physical

model of speech production is very complex and anyway simplified).

in signal theory, we will consider these useful signals as random signals (processes)

(for example speech, circulation of letters in agencies of Czech Posts, exchange rate

CZK/EUR . . . ).

According to the character of the time axis, random signals are divided into continuous

time random signals (the time is defined for all t) and discrete time random signals

(only for discrete n).

These signals can not be represented in all time-points (in this case, they would be

deterministic), we will rather look for characteristic properties of random signals such as

mean value, probability density function, etc.

2

continuous time: the system {t } of random variables defined for all t is called

random process, it is denoted (t).

discrete time: the system {n } of random variables defined for all n N is called

random process, it is denoted [n].

Set of realizations of random process

the set of realizations includes infinity of possible runs of random process - its

realizations. We will limite ourselves to finite number and denote each realization as

(t), or [n]. In case any parameters are estimated on this set, we will speak about

ensemble estimates.

Example: the random signal is recording of water flowing through the water tube in my

flat. 1068 realizations, each of 20 ms, were recorded. For the demonstration of continuous

random signals, we will imagine this set as (t), for discrete random signals as [n].

3

0.2

0

0.2

0

0.002

0.004

0.006

0.008

0.01

0.012

0.014

0.016

0.018

0.002

0.004

0.006

0.008

0.01

0.012

0.014

0.016

0.018

0.002

0.004

0.006

0.008

0.01

0.012

0.014

0.016

0.018

0.002

0.004

0.006

0.008

0.01

0.012

0.014

0.016

0.018

0.2

0

0.2

0.2

0

0.2

0.2

0

0.2

0.2

0

0.2

0

50

100

150

200

250

300

50

100

150

200

250

300

50

100

150

200

250

300

50

100

150

200

250

300

0.2

0

0.2

0.2

0

0.2

0.2

0

0.2

Distribution function

is defined for one random variable: the random process for given time t or n is such a

random variable. Definition:

F (x, t) = P{(t) < x},

F (x, n) = P{[n] < x},

where P{(t) < x} or P{[n] < x} is the probability that random variable in given time

will be smaller than x. Note, that x is nothing random, it is an auxiliary variable.

Ensemble estimation of distribution function: we will fix ourselves in given time t or n,

and take realizations. For given x we estimate:

P

=1 1 if (t) < x, 0 else

F (x, t) =

P

=1 1 if [n] < x, 0 else

F (x, n) =

1

F(x,0.1ms)

F(x,3.1ms)

F(x,6.3ms)

F(x,9.4ms)

0.9

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0.3

0.2

0.1

0.1

x

0.2

0.3

0.4

0.5

1

F(x,1)

F(x,50)

F(x,100)

F(x,150)

0.9

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0.3

0.2

0.1

0.1

x

0.2

0.3

0.4

0.5

is again defined for one random variable (random process for a given time t or n is such a

random variable). Definition:

F (x, t)

p(x, t) =

x

F (x, n)

p(x, n) =

x

Ensemble estimation of Probability density function: The function can be obtained by

numeric derivation from estimated F (x, t) or F (x, n) or it can be estimated by a

histogram:

choose given t or n

CHoose L values x from xmin till xmax , with regular step =

xmax xmin

:

L1

. . . xL1 = xmin + (L 2), xL = xmin + (L 1) = xmax

In such a way, well obtain L cages with width , for xi , given cage chi is from

9

xi

2 till xi + 2 . The left edge of the

left-most cage (1) will be stretched till , the right edge of the right-most (L) till +.

p(xi , t) =

p(xi , n) =

=1

=1

10

3.5

3

2.5

2.5

p(x,3.1ms)

p(x,0.1ms)

2

1.5

1.5

1

0.5

0.5

0

0.2

0.2

0.4

0.6

0.2

0.2

0.4

0.6

0.2

0.4

0.6

3.5

3

p(x,9.4ms)

p(x,6.3ms)

2.5

2

1.5

2.5

2

1.5

0.5

0.5

0.2

0.2

0.4

0.6

0.2

0

x

11

3.5

2.5

p(x,50)

p(x,1)

2.5

2

1.5

1.5

1

0.5

0.5

0

0.2

0.2

0.4

0.6

0.2

0.2

0.4

0.6

0.2

0.4

0.6

3.5

3

3

2.5

p(x,150)

p(x,100)

2.5

2

1.5

2

1.5

0.5

0.5

0.2

0.2

0.4

0.6

0.2

0

x

12

our task is to determine the probability, that the value of random process in time t or n is

in the interval [a, b]. We can estimate this in two ways (shown only for continuous time):

from distribution function, having in mid its definition: F (x, t) = P{(t) < x}, then

P{a < (t) < b} = F (b, t) F (a, t)

from probability density function. Densities can be integrated:

Z b

P{a < (t) < b} =

p(x, t)dx

a

13

the values of random process will hardly be smaller than , therefore

F (, t) = P{(t) < } = 0.

the values of random process will all be smaller than , therefore

F (+, t) = P{(t) < +} = 1.

probability density function is derivation of distribution function, the inverse sense is

given by an integral:

Z x

F (x, t) =

p(g, t)dg

14

Z +

p(x, t)dx = 1

the value of probability densifty function for given x is not a probability !!!

15

beer keg is being drunk from x = 18.00 till 22.00. Well define function p(x) as immediate

beer consumption (drinking function) and F (x) as function of drunk beer (attention, x is

time in this example):

The behavior is simnilar to PDF and distribution functions, just replace 1 with 1 keg:

F (x) is zero in time (it is zero till 18.00), as there was no beer.

F (x) is 1 keg in time + (actually already at 22.00), as all beer was drunk and

16

therell be no more.

Rx

R +

total amount of drunk beer: F (+) = p(x)dx = 1 keg.

points on F (x) or by integration of p(x).

one value of p(x) (for example. p(19.00)) can NOT be called amount of drunk beer

noone can drink anything in infinitely short time.

17

Moments

on contrary to functions, moments are values, characterizing random process in time t or

n:

mean value or Expectation, is the 1st moment:

Z +

a(t) = E{(t)} =

xp(x, t)dx

a[n] = E{[n]} =

xp(x, n)dx

ensemble estimate of mean value for each time t or n the estimate is simply the mean

value of samples over all realizations:

1 X

(t)

a

(t) =

=1

1 X

[n]

a

[n] =

=1

18

3

x 10

5

0

5

10

15

0

0.002

0.004

0.006

0.008

0.01

t

0.012

0.014

0.016

0.018

x 10

5

0

5

10

15

0

50

100

150

200

n

19

250

300

D(t) = E{[(t) a(t)]2 } =

p

p

(t) = D(t)

[n] = D[n]

ensemble estimate of variance and std: for each t or n:

X

1

D(t)

=

[ (t) a

(t)]2 ,

=1

X

1

[ [n] a

[n]]2 ,

D[n]

=

=1

20

(t) = D(t)

[n] =

D[n]

0.14

0.135

0.13

0.125

0

0.002

0.004

0.006

0.008

0.01

t

0.012

0.014

0.016

0.018

0.14

0.135

0.13

0.125

0

50

100

150

200

n

21

250

300

Correlation function

is quantifying the similarity between values of random process in times t1 (or n1 ) and t2

(or n2 ):

Z + Z +

R(t1 , t2 ) =

x1 x2 p(x1 , x2 , t1 , t2 )dx1 dx2 ,

R(n1 , n2 ) =

function between times t1 and t2 , resp. n1 and n2 . Theoretically, it can be computed from

two-dimensional distribution function:

F (x1 , x2 , t1 , t2 ) = P{(t1 ) < x1 a (t2 ) < x2 },

F (x1 , x2 , n1 , n2 ) = P{[n1 ] < x1 a [n2 ] < x2 }

22

2 F (x1 , x2 , t1 , t2 )

p(x1 , x2 , t1 , t2 ) =

x1 x2

2 F (x1 , x2 , n1 , n2 )

p(x1 , x2 , n1 , n2 ) =

x1 x2

Well rather be interested in ensemble estimation using 2D-histogram:

similarly as for 1D histogram, define cages, in forms of squares: chij is for the first

till

x

and

for

the

second

dimension

from

x

dimension from xi

i

j

2

2

2 till

xj +

2.

23

24

the value of 2D-histogramu for cage chij with values xi and xj will be:

P

=1 1 if (t1 ), (t2 ) chij , 0 else

p(xi , xj , t1 , t2 ) =

2

P

1 if [n1 ], [n2 ] chij , 0 else

p(xi , xj , n1 , n2 ) = =1

2

25

0.4

0.4

100

20

0.2

0.2

80

60

x1

x1

15

0

10

40

0.2

0.2

5

20

0.4

0.4

0.2

0

x2

0.2

0.4

0.4

0.4

0.4

0.2

0

x2

0.2

0.4

0.4

14

12

20

0.2

0.2

10

15

x1

x1

8

0

10

0.2

0.2

5

0.4

0.4

0.2

0

x2

0.2

0.4

0.4

0.4

26

0.2

0

x2

0.2

0.4

R(0, 0) = 0.0188: the process is very similar to itself (the same point). similar. . .

R(0, 1) = 0.0151: if shifted to neighboring sample, still very similar to time n1 = 0.

R(0, 5) = 0.0030: times n1 = 0 and n2 = 5 are not similar at all.

R(0, 11) = 0.0133: in times n1 = 0 and n2 = 11 the process is similar to itself, but

with inverse sign ! It is probable, that if the value [n1 ] is positive, [n2 ] will be

negative and vice versa.

27

n1 = 0, n2 = n1 +k for k = 0 . . . 40, comparison with n1 = 100, n2 = n1 +k for k = 0 . . . 40:

0.02

0.015

R(0,k)

0.01

0.005

0

0.005

0.01

0.015

10

15

20

k

25

30

35

40

10

15

20

k

25

30

35

40

0.02

R(100,100+k)

0.015

0.01

0.005

0

0.005

0.01

0.015

28

simply said, the behavior of random process is not changing with the time. The values do

not depend on time t or n. Correlation function does not depend on the precise values of

t1 , t2 or n1 , n2 , but only on their difference: = t2 t1 , k = n2 n1 . For stationary

continuous-time signal:

F (x, t) F (x)

p(x, t) p(x)

a(t) a D(t) D

(t)

Similarly for discrete time:

F (x, n) F (x)

p(x, n) p(x)

a[n] a D[n] D

p(x1 , x2 , n1 , n2 ) p(x1 , x2 , k)

29

[n]

R(n1 , n2 ) R(k)

the mean value was similar for all times. In case we had more realizations, it would be

even more similar.

dtto for standard deviation.

Correlation function for n1 = 0, n2 = n1 + k and for n1 = 100, n2 = n1 + k was

similar.

30

2

1.5

1

x (t)

0.5

0

0.5

1

1.5

2

0

0.5

1.5

t

2.5

0.5

1.5

t

2.5

1

0.8

0.6

0.4

xk(t)

0.2

0

0.2

0.4

0.6

0.8

1

0

31

all parameters can be estimated from 1 realization:

32

Example of stationary and ergodic and of stationary but non-ergodic random process.

2

1.5

1

x (t)

0.5

0

0.5

1

1.5

2

0

0.5

1.5

t

2.5

0.5

1.5

t

2.5

2

1.5

1

x (t)

0.5

0

0.5

1

1.5

2

0

33

of lenght T (continuous time) or of N samples (discrete time). The only realization we

have will be simply denoted x(t), resp. x[n]:

histograms can be applied to estimate distribution function and PDF:

mean value, variance, std:

Z

1 T

a

=

x(t)dt

T 0

N 1

1 X

a

=

x[n]

N n=0

D=

T

[x(t) a

]2 dt

N

1

X

1

=

D

[x[n] a

]2

N n=0

correlation function:

)= 1

R(

T

x(t)x(t + )dt

N

1

X

1

x[n]x[n + k]

R(k)

=

N n=0

34

- Analogue Signals on the arduinoUploaded byWingston Sharon
- PAM TDMUploaded bydubstepo
- Analog vs Digital CommunicationUploaded byYOGESH
- papershortUploaded bypostscript
- Assignment 3Uploaded bySeema P Diwan
- A Survey: Feature Extraction of ECG Signals using Haar WaveletUploaded byInternational Journal for Scientific Research and Development - IJSRD
- Random Processes Ma1254Uploaded bySurya Kumar Maran
- v6-14Uploaded byAdegbayibi Adesanmi Timothy
- ABET PTU Syllabus Digital Control(JAFAR JALLAD)Uploaded byGreen Heart
- Reports HsUploaded byDuarte Jose Antunes
- ece308-1Uploaded byfreeTerp
- 3Uploaded byNegin Maleki
- Some New Probability Density FunctionsUploaded byGeorge Zafeiriou
- NI CaseStudy Cs 14031Uploaded bytyc0square
- comp.pdfUploaded byjennybunnyomg
- 6605BE 2071 Bhadra New Regular II PartUploaded byGaurab Rai Lama
- Time Averages and ErgodicityUploaded byShubham
- 2141003.pdfUploaded bykapiljoshi007
- Courant_11_2011Uploaded byAbdullah Kaymakcı
- itc2001-ctlUploaded byumesh
- DIGLAB5 - Control of Analog to Digital Converters Aug 28Uploaded byMarlon Boucaud
- Slide 03Uploaded byMohit Gupta
- Seminar ReportUploaded byNk Kushal
- ECE 6602 - Week1 Lec1.pptUploaded byAnonymous 4kM3hcpy
- Pulse Direction (1)Uploaded byAnonymous TThmYKFp
- 18 Earthquake PredictionUploaded byjafeelnm
- ABB SupplementUploaded bysuhailfarhaan
- W2 - Playground Accessibility - Scott A. Bennet.pdfUploaded bycem demirci
- Chaos Theory FinanceUploaded byElmeDino
- MIT15_075JF11_chpt05.docxUploaded byRobert Manea

- Ysmael v. Deputy ExecUploaded byDom Umandap
- Why do a PhDUploaded byAdeniji Olusegun
- 52745654-Her-drearyUploaded byNanin Rao
- Fraleigh - Consciousness MattersUploaded bygollyjeepers
- Module 4 Mechanical Properties of Engineering MaterialsUploaded byChristopher Nazareno
- Vision Industrial ExpoUploaded byMurli Iyer
- BB Chapter 5 Gaseous State 2010Uploaded bykn01zg
- OR PositionsUploaded byEduard
- Moving From Time Based Maintenance to Condition Based Maintenance (Webinar)Uploaded byJor Billacis
- Amanishah Nallah Presentation Part 2 - R5Uploaded byAnonymous ciKyr0t
- Public Awareness of and Opinion on the Reintroduction of Elk in Western MarylandUploaded byAmmoLand Shooting Sports News
- Detecting Heartbeats in the Ballistocardiogram With Clustering-paalasmaaUploaded bymac2022
- Final CSRUploaded byrajubansal
- Chapter 19 VirusesUploaded byjpayne655
- Steering FundamentalUploaded byLUCIFERBREO
- Business Systems Analyst Manager In Boston MA Resume Aaron StarrUploaded byAaron Starr1
- Adaptogenic 2Uploaded byAnggra Retno
- Classification of Electrical Installations in Healthcare Jul10 EnUploaded byAndres Zuñiga
- meningitis.pdfUploaded byM Faddhil Wasi Pradipta
- Stengler+Intranasal+Light+02-14Uploaded byCiugudean Mircea
- UFO Contact From Planet UmmoUploaded bylectoraavida
- manual MMX 6Uploaded bysheyla
- Lecture 1Uploaded byMahrukh Fida
- Design EquationsUploaded bymurakami27
- Minerals and Rocks Lecture Notes(3)Uploaded byTundeOyedotun
- s6_btechUploaded byHarshal Sp
- J-05Uploaded byRochdi Sahli
- DynamoUploaded byMina Farag
- Continuous and Discrete Time Signals and Systems (Mandal & Asif) Solutions - Chap03Uploaded byrdugas
- chloroform for the kingUploaded byapi-184836296

## Much more than documents.

Discover everything Scribd has to offer, including books and audiobooks from major publishers.

Cancel anytime.