Pattern Recognition 1: Introduction - KTH
Pattern Recognition 1: Introduction - KTH
Pattern Recognition 1: Introduction - KTH
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
Classification<br />
Summary<br />
<strong>Pattern</strong> <strong>Recognition</strong> 1: <strong>Introduction</strong><br />
Arne Leijon<br />
<strong>KTH</strong> Sound and Image Processing<br />
Aug 27, 2012<br />
Arne Leijon<br />
<strong>Pattern</strong> <strong>Recognition</strong> 1: <strong>Introduction</strong>
Classification<br />
Summary<br />
Intro<br />
Example<br />
Probability Density<br />
A<strong>Pattern</strong>-ClassificationSystem<br />
Input<br />
signal<br />
Transducer<br />
Feature<br />
Extractor<br />
Classifier<br />
Output<br />
decision<br />
Arne Leijon<br />
<strong>Pattern</strong> <strong>Recognition</strong> 1: <strong>Introduction</strong>
Classification<br />
Summary<br />
Intro<br />
Example<br />
Probability Density<br />
Source Categories<br />
Noise<br />
Signal Source<br />
Sub-source #1<br />
Sub-source #2<br />
Noise<br />
Noise<br />
Noise<br />
Transducer<br />
Sub-source # N s<br />
Noise<br />
S<br />
Random State<br />
Switch<br />
Arne Leijon<br />
<strong>Pattern</strong> <strong>Recognition</strong> 1: <strong>Introduction</strong>
Classification<br />
Summary<br />
Intro<br />
Example<br />
Probability Density<br />
Optimal Classification Mechanism<br />
Classifier<br />
Feature Extraction<br />
#1<br />
Discriminant<br />
Function<br />
g 1 ( x)<br />
x<br />
Select<br />
index<br />
of<br />
largest<br />
#K<br />
Discriminant<br />
Function<br />
g N d<br />
( x)<br />
Arne Leijon<br />
<strong>Pattern</strong> <strong>Recognition</strong> 1: <strong>Introduction</strong>
Classification<br />
Summary<br />
Intro<br />
Example<br />
Probability Density<br />
First Classifier Example: Man or Woman?<br />
100<br />
90<br />
x 2<br />
= Weight / kg<br />
80<br />
70<br />
60<br />
50<br />
150 160 170 180 190 200<br />
x = Height / cm<br />
1<br />
Arne Leijon<br />
<strong>Pattern</strong> <strong>Recognition</strong> 1: <strong>Introduction</strong>
Classification<br />
Summary<br />
Intro<br />
Example<br />
Probability Density<br />
Two Gaussian Feature Probability Density Functions<br />
Women<br />
Men<br />
x 10 −3<br />
5<br />
x 10 −3<br />
5<br />
4<br />
4<br />
3<br />
3<br />
2<br />
2<br />
100 0 1<br />
80<br />
x 2<br />
= Weight / kg<br />
60<br />
150<br />
160<br />
190<br />
180<br />
170<br />
x 1<br />
= Height / cm<br />
200<br />
100 0 1<br />
80<br />
x 2<br />
= Weight / kg<br />
60<br />
150<br />
160<br />
190<br />
180<br />
170<br />
x 1<br />
= Height / cm<br />
200<br />
Arne Leijon<br />
<strong>Pattern</strong> <strong>Recognition</strong> 1: <strong>Introduction</strong>
Classification<br />
Summary<br />
Intro<br />
Example<br />
Probability Density<br />
Contours of Equal Probability Density<br />
Women<br />
100<br />
Men<br />
100<br />
90<br />
90<br />
x 2<br />
= Weight / kg<br />
80<br />
70<br />
x 2<br />
= Weight / kg<br />
80<br />
70<br />
60<br />
60<br />
50<br />
150 160 170 180 190 200<br />
x = Height / cm<br />
1<br />
50<br />
150 160 170 180 190 200<br />
x = Height / cm<br />
1<br />
Arne Leijon<br />
<strong>Pattern</strong> <strong>Recognition</strong> 1: <strong>Introduction</strong>
Classification<br />
Summary<br />
Intro<br />
Example<br />
Probability Density<br />
Decision Boundary and Decision Regions<br />
100<br />
90<br />
x 2<br />
= Weight / kg<br />
80<br />
70<br />
60<br />
50<br />
150 160 170 180 190 200<br />
x 1<br />
= Height / cm<br />
Arne Leijon<br />
<strong>Pattern</strong> <strong>Recognition</strong> 1: <strong>Introduction</strong>
Classification<br />
Summary<br />
Intro<br />
Example<br />
Probability Density<br />
Gaussian Probability Density (2-dim)<br />
5<br />
4<br />
3<br />
x 2<br />
2<br />
1<br />
0<br />
−1<br />
−2<br />
−3<br />
−4<br />
−5<br />
−5 −4 −3 −2 −1 0 1 2 3 4 5<br />
x 1<br />
✓ ◆ 3<br />
2<br />
0<br />
cov [X ]=<br />
0 1 2<br />
✓ ◆ 9 0<br />
=<br />
0 1<br />
=<br />
Arne Leijon<br />
<strong>Pattern</strong> <strong>Recognition</strong> 1: <strong>Introduction</strong>
Classification<br />
Summary<br />
Intro<br />
Example<br />
Probability Density<br />
Gaussian Probability Density (2-dim)<br />
5<br />
4<br />
3<br />
x 2<br />
2<br />
1<br />
0<br />
−1<br />
−2<br />
−3<br />
−4<br />
−5<br />
−5 −4 −3 −2 −1 0 1 2 3 4 5<br />
x 1<br />
✓ ◆ 1<br />
2<br />
0<br />
cov [X ]=<br />
0 3 2<br />
✓ ◆ 1 0<br />
=<br />
0 9<br />
=<br />
Arne Leijon<br />
<strong>Pattern</strong> <strong>Recognition</strong> 1: <strong>Introduction</strong>
Classification<br />
Summary<br />
Intro<br />
Example<br />
Probability Density<br />
Gaussian Probability Density (2-dim)<br />
5<br />
x 2<br />
4<br />
3<br />
2<br />
1<br />
0<br />
−1<br />
−2<br />
−3<br />
−4<br />
−5<br />
−5 −4 −3 −2 −1 0 1 2 3 4 5<br />
x 1<br />
✓ ◆ 3<br />
2<br />
0<br />
cov [X ]=P<br />
0 1 2 P T =<br />
✓ ◆ 5 4<br />
=<br />
4 5<br />
where<br />
P = p 1 ✓ ◆ 1 1<br />
2 1 1<br />
Arne Leijon<br />
<strong>Pattern</strong> <strong>Recognition</strong> 1: <strong>Introduction</strong>
Classification<br />
Summary<br />
Intro<br />
Example<br />
Probability Density<br />
Gaussian Probability Density (K-dim)<br />
f X (x) =<br />
1<br />
(2⇡) K/2p det C e 1 2 (x µ)T C 1 (x µ)<br />
Arne Leijon<br />
<strong>Pattern</strong> <strong>Recognition</strong> 1: <strong>Introduction</strong>
Classification<br />
Summary<br />
Intro<br />
Example<br />
Probability Density<br />
General Probability Density: Gaussian Mixture (1-dim)<br />
0.4<br />
0.35<br />
0.3<br />
Prob. Density<br />
0.25<br />
0.2<br />
0.15<br />
0.1<br />
0.05<br />
0<br />
−2 0 2 4 6<br />
x<br />
Arne Leijon<br />
<strong>Pattern</strong> <strong>Recognition</strong> 1: <strong>Introduction</strong>
Classification<br />
Summary<br />
Intro<br />
Example<br />
Probability Density<br />
General Probability Density: Gaussian Mixture (2-dim)<br />
Arne Leijon<br />
<strong>Pattern</strong> <strong>Recognition</strong> 1: <strong>Introduction</strong>
Classification<br />
Summary<br />
Intro<br />
Example<br />
Probability Density<br />
General Probability Density: Gaussian Mixture (K-dim)<br />
f X (x) =<br />
MX<br />
m=1<br />
w m<br />
1<br />
(2⇡) K/2p det C m<br />
e 1 2 (x µ m )T C 1<br />
m (x µ m )<br />
Arne Leijon<br />
<strong>Pattern</strong> <strong>Recognition</strong> 1: <strong>Introduction</strong>
Classification<br />
Summary<br />
Summary: Important Concepts<br />
Feature Vector: vector x containing all input data to the classifier<br />
Observation Space: set of all possible feature-vector values<br />
Decision Function: d(x) mapsfeaturevectorx into discrete<br />
decision output, coded as integer<br />
Decision Region: set of feature-vector values giving same decision:<br />
⌦ i = {x : d(x) =i}<br />
Discriminant Function: g i (x) mapsanyfeaturevectorx into real<br />
number, used for classification<br />
General Optimal Classifier: d(x) = argmax k g k (x)<br />
Two-category Classifier: can be implemented using a single<br />
Discriminant Function: d(x) = sgn g(x)<br />
Gaussian Mixture Model: can approximate any distribution<br />
Arne Leijon<br />
<strong>Pattern</strong> <strong>Recognition</strong> 1: <strong>Introduction</strong>