[go: up one dir, main page]
Scribd
Upload
136 views62 pages

2 Classical Detection and Estimation Theory

Classical Detection and Estimation Theory discusses key concepts including: 1) Detection theory involves developing decision rules based on probabilistic observations to choose between hypotheses about an unknown source. 2) Simple binary hypothesis tests involve choosing between two hypotheses H1 and H0 based on a set of observations using likelihood ratios. 3) The Neyman-Pearson criterion aims to maximize detection probability subject to a constraint on false alarm probability, while the Bayes criterion minimizes expected risk. 4) Estimation theory involves using observations to estimate unknown parameters, with goals including minimizing mean squared error. Maximum likelihood estimation chooses parameters that make the observations most likely.

Uploaded by

thesunbk25
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
136 views62 pages

2 Classical Detection and Estimation Theory

Classical Detection and Estimation Theory discusses key concepts including: 1) Detection theory involves developing decision rules based on probabilistic observations to choose between hypotheses about an unknown source. 2) Simple binary hypothesis tests involve choosing between two hypotheses H1 and H0 based on a set of observations using likelihood ratios. 3) The Neyman-Pearson criterion aims to maximize detection probability subject to a constraint on false alarm probability, while the Bayes criterion minimizes expected risk. 4) Estimation theory involves using observations to estimate unknown parameters, with goals including minimizing mean squared error. Maximum likelihood estimation chooses parameters that make the observations most likely.

Uploaded by

thesunbk25
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 62

Classical Detection and Estimation Theory

Pukyong University, EE
Park, Jaehyun
Detection theory

 Component of a decision theory problem


 Source
 Hypothesis (가설)
• P. 20 of textbook
 Probabilistic transition mechanism
 Observation space

2
3
4
Simple binary hypothesis tests

 A set of N observations
 r1 
r 
r  2
 
 
 rN 

 Probabilistic transition mechanism

pr|H1  R | H1  pr|H 0  R | H 0 

 The object
 Development of a suitable decision rule based on this information

5
Decision criteria

 Bayes Criterion
 Neyman-Pearson Criterion

6
Bayes Criterion

A priori probability
P1 : Source generates H1 with a probability P1
P0 : Source generates H 0 with a probability P0

 Expected risk

7
8
 Binary test

9 Positive
Assign R to Z1

10
Likelihood ratio test (LRT)

 Likelihood ratio

 Threshold

 LRT

Log-Likelihood ratio test (LLRT)

11
Example 1

12
13
14
15
Special case of Bayes tests

 ML symbol detection in wireless communications

16
Probabilities of False alarm, detection, miss

 Conditional probabilities

17
Neyman-Pearson test

 It is difficult to assign realistic costs or a priori probabilities

 Neyman-Pearson Criterion

F  PM    PF   

18
To satisfy the constraint, PF   .
The density of  when H 0 is true as p|H 0   | H 0 

   PF   pr|H  R | H 0  dR
0
Z1

19
Receiver Operating Characteristic (ROC)

 Example 1

20
 False alarm prob.

 Detection prob.

21
Error function

22
23
24
Approximation of error function

25
Estimation theory

26
Four components for estimation problem model

27
28
Bayes Estimation

29
30
31
32
33
34
35
36
37
The invariance to the choice of a cost function is a useful feature because of
the subjective judgments that are frequently involved in choosing C  a  .

Some conditions under which this invariance holds are developed


in the next two properties

38
39
40
Real (Nonrandom) parameter estimation

 MSE criterion

 Minimizing MSE Unknown parameter

41
Real (Nonrandom) parameter estimation

42
Maximum likelihood estimation

43
Cramer Rao bound

44
Proof

45
46
47
Example 2 (p.35 of this lecture note)

48
Properties of the ML estimate

49
Composite Hypotheses

50
Composite Hypotheses (cont’d)

51
Example 1

52
Example 2

53
54
Likelihood ratio test

55
56
Generalized LRT (GLRT)

57
58
59
60
61
62

You might also like