csis pattern recognition
TRANSCRIPT
1
© S. ChaFall/2002
CSIS
Pattern Recognition
Fall of 2002
Prof. Sung-Hyuk Cha
School of Computer Science & Information Systems
© S. ChaFall/2002
CSISArtificial Intelligence
2
© S. ChaFall/2002
CSISPerception
© S. ChaFall/2002
CSISLena & Computer vision
3
© S. ChaFall/2002
CSISM achine Vision
© S. ChaFall/2002
CSISPatter n Recognition Applications
4
© S. ChaFall/2002
CSIS
�����
���
� ���
� �� �
������
� �� �
������
����� � � �����
I r is authentication
© S. ChaFall/2002
CSIS
5
© S. ChaFall/2002
CSIS
Each person has different faces.
Face Recognition System
© S. ChaFall/2002
CSIS
?Query
Face DB
Face Recognition System
6
© S. ChaFall/2002
CSIS
Sargur N. Srihari
f6
street namef7
secondarydesignator abbr.
f5
primarynumber
f8
secondarynumber
Lee Entrance STE520 202
f2
state abbr.f3
5-digitZIP Code
f4
4-digitZIP+4 add-on
f1
city name
-Amherst NY 14228 2583
• Delivery point: 142282583
Complex Patter n Recognition Applications
© S. ChaFall/2002
CSISSpeech Recognition System
�������������� ����
7
© S. ChaFall/2002
CSIS
biomouseFingerprint
scanner
DigitalCamera
LCD Pentablet Microphone
),...,,( 21x
dxx fff ),...,,( 21
xd
xx fff ),...,,( 21x
dxx fff ),...,,( 21
xd
xx fff),...,,( 21x
dxx fff ),...,,( 21
xd
xx fff
Vital Sign
monitor
Applications
© S. ChaFall/2002
CSIS
8
© S. ChaFall/2002
CSIS
br ightness, length
Salmon
Bass
Salmon1 = ( 12 , 16 )
Salmon2 = ( 11 , 20 )
Bass1 = ( 7 , 6 )
Bass2 = ( 3 , 4 )
Truth features
M easurements
© S. ChaFall/2002
CSIS
9
© S. ChaFall/2002
CSIS
© S. ChaFall/2002
CSISDecision theory (cost)
10
© S. ChaFall/2002
CSIS
© S. ChaFall/2002
CSISDistr ibutions and Errors
Bass
Salmon
Bass identified as salmon
salmonidentified as
bass
Decisionboundary
11
© S. ChaFall/2002
CSISParametric Univariate Dichotomizer
(a) length (b) lightness (c) width
39 %Type II 27 % 26 %9 %Type I 7 % 5 %(a) (b) (c)
© S. ChaFall/2002
CSISM ultivariate Analysis
12
© S. ChaFall/2002
CSIS
length
br ightness
Salmon
Bass
= salmon?
Nearest Neighbor Classifier
© S. ChaFall/2002
CSIS
• too slow for users to wait for the output.
q = 4, 6
r2 = 3, 8
r1 = 5, 7
r3 = 10, 16
r4 = 12, 14
Salmon
Bass
Salmon
Bass
Salmon
rn = 14, 15
R =
t2 = 2, 5
t1 = 4, 6
t3 = 11, 17
t4 = 14, 12
Salmon
Bass
Salmon
Bass
Salmon
tn = 14, 17
T =
Bass
Bass
Salmon
Bass
Bass
• Performance is evaluated by using a testing set.
reference set testing set
Nearest Neighbor Classifier
13
© S. ChaFall/2002
CSIS
length
br ightness
Salmon
Bass
Y > aX + b
? = salmon
M achine Lear ning (L inear function)
© S. ChaFall/2002
CSIS
synapse
nucleus
axon
dendrites
the biological neuron
Σ
x1(t)
x2(t)
xn(t)
w1
w2
wn
a(t)
w0
y
a
y=f(a)
O(t+1)
the artificial neuron
Artificial Neural Networ k
14
© S. ChaFall/2002
CSIS
• extremely fast.
r2 = 3, 8
r1 = 5, 7
r3 = 10, 16
r4 = 12, 14
Salmon
Bass
Salmon
Bass
Salmon
rn = 14, 15
R =
t2 = 2, 5
t1 = 4, 6
t3 = 11, 17
t4 = 14, 12
Salmon
Bass
Salmon
Bass
Salmon
tn = 14, 17
T =
Bass
Bass
Salmon
Bass
Bass
• Performance is evaluated by using a testing set.
reference set testing set
Y > aX + b
• Performance is not as good as the NN classifier’s.
training set
• No need to load the training data during the classification.
M achine Lear ning (L inear function)
© S. ChaFall/2002
CSIS
length
br ightness
Salmon
Bass
Y > aX + b
Non-L inear case
15
© S. ChaFall/2002
CSIS
length
br ightness
Salmon
Bass
• NN is better.• will learn artificial neural network which is non-linear function.
Non-L inear case
© S. ChaFall/2002
CSISHuman Brain
16
© S. ChaFall/2002
CSIS
Class
2f
3f
4f
5f
6f
1f
7f
Fully Connected, feed forward, back-propagation
multi-layer Artificial neural network (11-6-1) (ANN).
Artificial Neural Networ k
© S. ChaFall/2002
CSIS
17
© S. ChaFall/2002
CSIS
© S. ChaFall/2002
CSIS
• Predict unseen future instance.
• Generalization.
• Inductive step.
Pur pose of Patter n Recognition
18
© S. ChaFall/2002
CSIS
length
width
Generalizability (statistical inferece)
length
width
length
width
training set
validating set
universe
© S. ChaFall/2002
CSISInferential Statistics
1. Inferential Statistics is inferring a conclusion about population of interest from a sample.- need a procedure for sampling the population.- need a measure of reliability for the inference.
2. If error rate in a random sample set is the same as in universe, then the procedure is a sound inferential statistical procedure.
3. If error rate in one random sample set is the same as in another random sample set, then the procedure is sound.
19
© S. ChaFall/2002
CSISGeneralization
δδδδf2
δδδδf1
Universe
© S. ChaFall/2002
CSISSampling & learning
δδδδf2
δδδδf1
Sample 1
20
© S. ChaFall/2002
CSISTesting on another sample
δδδδf2
δδδδf1
Sample 2
© S. ChaFall/2002
CSISGeneralization
δδδδf2
δδδδf1
Universe
21
© S. ChaFall/2002
CSISM ultiple classification
f1
f2
class 1 class 3
class 2
Classes = {class 1, 2, 3}
© S. ChaFall/2002
CSIS
22
© S. ChaFall/2002
CSIS
© S. ChaFall/2002
CSIS
1. Data acquisition:
Template for PR Applications
2. Feature Extraction:
3. Training a classifier:
4. Classification system:
a. Recruit subjects.b. Modality interface (Scanning, picturing, recording, etc).
a. Raw data to feature vectors.b. Involves image/ voice/ signal processing techniques.
a. Design a classifier (e.g., ANN).b. Enter the training (& validating) feature vector set(s).
a. embed the ANN engine to your actual program (Java/C)b. User interface for the Final Product.
23
© S. ChaFall/2002
CSIS
• Fast Nearest Neighbor Search Algorithms
• Decision Tree
• Statistical Pattern Recognition.
• Artificial Neural Network.
• Clustering
• etc.
http://www.csis.pace.edu/~scha/PR
Fur ther Patter n Recognition
© S. ChaFall/2002
CSIS
outlook temperature humidity windy playsunny hot high false nosunny hot high true noovercast hot high false yesrainy mild high false yesrainy cool normal false yesrainy cool normal true noovercast cool normal true yessunny mild high false nosunny cool normal false yesrainy mild normal false yessunny mild normal true yesovercast mild high true yesovercast hot normal false yesrainy mild high true no
Decision Tree
24
© S. ChaFall/2002
CSIS
(b)
ad
kj
h
gif
ec
b
ad
kj
h
g if
ec
b
(a)
(c) 1 2 3
abc
0.4 0.1 0.50.1 0.8 0.10.3 0.3 0.4
...
(d)
g a c i e d k b j f h
Cluster ing
© S. ChaFall/2002
CSISTerminology
Classification:
The process of assigning one of a limited set of alternative interpretations to (the generator of) a set of data. Often requires the steps of the computation of relative probabilities (or a quantity related to them) followed by the application of a decision rule. All classification processes can be evaluated in terms of "detection" and "mis-classification" rates. See "receiver operator characterstics”
25
© S. ChaFall/2002
CSISTerminology
• Computer Vision: Compter Vision is the subject area which deals with the automatic analysis of images for the purposes of quantification or system control (often mimicking tasks which humans find trivial). It is to be distinguised from "Image Processing" which deals only with the computational processes applied to images, including enhancement and compression, but does not deal with abstract representation for the purposes of reasoning and interpretation. Compter Vision can be seen as the inverse of Computer Graphics, though generally the representations and methods of this area are not of use in Computer Vision due to the incomplete and therefore ambiguous nature of images. This requires prior knowledge to be used in order to obtain robust scene interpretation.
© S. ChaFall/2002
CSISTerminology
• Machine Vision:Like "computer vision" but generally more closely associated
with its use in robotics.
• Pattern Recognition Pattern recognition is the process of assigning a pattern classification to a particular set of measurements, normally represented as a high dimensional vector. This is normally done within the context of "probability theory", whereby a particular set of assumptions regarding the expected statisticaldistribution of measurements is used to compute classification probabilities which can be used as the basis for a decision such as the "Bayes decision rule". There are several popular forms of classifier including "k-nearest neighbour", "parzenwindows", "mixture methods" and more recently "artificial neural networks".
26
© S. ChaFall/2002
CSISTerminology
• Images:An image is two dimensional spatial representation of a group
of "objects" (or "scene") which exists in two or more dimensions. It is an intuitive way of presenting data for computer interfaces in the area of graphics, but in machine vision it may be defined as a continuous function of two variables defined within a bounded (generally rectangular) region.
• Histograms A histogram is an array of non negative integer counts from a set of data, which represents the frequency of occurance of values within a set of non-overlapping regions.
© S. ChaFall/2002
CSIS
.95 .49 .70 .71 .50 .10 .51 .92 .13 .47 .32 .21
.94 .49 .75 .70 .50 .11 .53 .84 .26 .54 .35 .18
.94 .49 .67 .74 .50 .10 .45 .85 .23 .48 .32 .22
.93 .72 .33 .47 .50 .21 .28 .30 .66 .60 .42 .10
.93 .74 .33 .48 .50 .22 .26 .30 .60 .59 .45 .10
.93 .79 .36 .54 .50 .18 .27 .32 .60 .59 .52 .09
.92 .30 .61 .66 .60 .11 .35 .49 .70 .71 .57 .10
.94 .42 .72 .66 .60 .11 .32 .49 .67 .74 .53 .10
.94 .40 .75 .67 .60 .12 .34 .49 .75 .70 .54 .11
.96 .30 .60 .59 .50 .10 .21 .30 .66 .60 .36 .10
.95 .32 .60 .59 .50 .09 .22 .30 .60 .59 .39 .10
.95 .30 .66 .60 .50 .10 .21 .32 .60 .59 .34 .09
dark blob hole slant width skew ht pixel hslope nslope pslope vslope
int int int real int real int int int int int int
Features
SSSSSS
BBBBBB
class
Features & Class
27
© S. ChaFall/2002
CSIS
length4035302520151050-5
-10
lightness
width
10 1214 16
18 20 2224
56
78
910
a
b
c
a = (12,6,-5)
b = (16,9,10)
c = (19,7,-10)
Representation
© S. ChaFall/2002
CSIS
The EndSee U all next week.
Patter n Recognition