pr_unit-i

7/30/2019 PR_UNIT-I

http://slidepdf.com/reader/full/prunit-i 1/28

12/31/2010 SAM_PR_UNIT-I 1

What is Pattern Recognition? Pattern: a description of an object.

A pattern class (or category) is a set of patternssharing common attributes and usually originatingfrom the same source

Recognition: classifying an object to a patternclass.

PR is the science that concerns the description or classification (recognition) of measurements.

PR techniques are an important component of

intelligent systems and are used for Decision making

Object & pattern classification

Data preprocessing

UNIT-I INTRODUCTION

7/30/2019 PR_UNIT-I


12/31/2010 SAM_PR_UNIT-I 2

PR applications: Image Preprocessing, Segmentation, and

Analysis

Computer Vision

Radar signal classification/analysis

Face recognition

Speech recognition/understanding

Fingerprint identification

Character recognition

Handwriting analysis

Electrocardiography signal

analysis/understanding

Medical diagnosis

INTRODUCTION

7/30/2019 PR_UNIT-I


12/31/2010 SAM_PR_UNIT-I 3

Data mining

Image databases

Seismic analysis

Commercial machines that canrecognize patterns: Automated speech recognition


Optical character recognition DNA sequence identification

Blood cells

Printed text

INTRODUCTION

7/30/2019 PR_UNIT-I


12/31/2010 SAM_PR_UNIT-I 4

Machine Perception During recognition (or classification) given objects are

assigned to prescribed classes.

A classifier is a machine which performs classification

Build a machine that can recognize patterns:

Speech recognition


OCR (Optical Character Recognition)

DNA sequence identification

7/30/2019 PR_UNIT-I


12/31/2010 SAM_PR_UNIT-I 5

PR Example “Sorting incoming Fish on a conveyor according to

species using optical sensing”

Salmon or Sea Bass?

Sea bass

SpeciesSalmon

7/30/2019 PR_UNIT-I


12/31/2010 SAM_PR_UNIT-I 6

Problem Analysis

• Automate the process of sorting incoming fishon a conveyor belt according to species(Salmon or Sea bass). Set up a camera

Take some sample images

Note the physical differences between the twotypes of fish

Length

Lightness

Width

No. & shape of fins

Position of the mouth

7/30/2019 PR_UNIT-I

http://slidepdf.com/reader/full/prunit-i 7/2812/31/2010 SAM_PR_UNIT-I 7

Solution by Stages Preprocess raw data from camera

Segment isolated fish

Extract features from each fish (length, width,brightness, etc.)

Classify each fish

7/30/2019 PR_UNIT-I


Problem Analysis cont’d Preprocessing: Use a segmentation operation to

isolate fishes from one another and from the

background

Information from a single fish is sent to a feature

extractor whose purpose is to reduce the data by

measuring certain features

The features are passed to a classifier

Classification: Select the length of the fish as a

possible feature for discrimination

7/30/2019 PR_UNIT-I


7/30/2019 PR_UNIT-I


Classification

Select the length of the fish as a possible feature for

discrimination

7/30/2019 PR_UNIT-I


The length is a poor feature alone!

Select the lightness as a possible feature

7/30/2019 PR_UNIT-I


Threshold decision boundary and cost relationship

Move our decision boundary toward smaller values of

lightness in order to minimize the cost (reduce the number of sea bass that are classified salmon!)

Task of decision theory

“Customers do not want sea bass in their

cans of salmon”

7/30/2019 PR_UNIT-I


Adopt the lightness and add the width of the fish

Fish x = [x1, x2]

Lightness Width

7/30/2019 PR_UNIT-I


12/31/2010 SAM_PR_UNIT-I 14

7/30/2019 PR_UNIT-I


12/31/2010 SAM_PR_UNIT-I 15

We might add other features that are not correlated

with the ones we already have. A precaution should

be taken not to reduce the performance by addingsuch “noisy features”

Ideally, the best decision boundary should be the one

which provides an optimal performance such as inthe following figure:

7/30/2019 PR_UNIT-I


12/31/2010 SAM_PR_UNIT-I 16

7/30/2019 PR_UNIT-I


12/31/2010 SAM_PR_UNIT-I 17

However, our satisfaction is premature because the

central aim of designing a classifier is to correctly

classify novel input

7/30/2019 PR_UNIT-I


12/31/2010 SAM_PR_UNIT-I 18

Decision Boundaries

Observe: Can do much better with two features

7/30/2019 PR_UNIT-I


12/31/2010 SAM_PR_UNIT-I 19

Pattern Recognition Systems /

A Complete PR System Sensing

Use of a transducer (camera or microphone)

PR system depends of the bandwidth, theresolution, sensitivity, distortion of the transducer

Segmentation and grouping

Patterns should be well separated and should notoverlap

7/30/2019 PR_UNIT-I


12/31/2010 SAM_PR_UNIT-I 20

Pattern Recognition Systems /

A Complete PR System cont’d

7/30/2019 PR_UNIT-I


12/31/2010 SAM_PR_UNIT-I 21

Pattern Recognition Systems cont’d Feature extraction

Discriminative features

Invariant features with respect to translation,rotation and scale.

Classification

Use a feature vector provided by a featureextractor to assign the object to a category

Post Processing Exploit context dependent information other than

from the target pattern itself to improveperformance

7/30/2019 PR_UNIT-I


12/31/2010 SAM_PR_UNIT-I 22

The Design Cycle

Data collection

Feature Choice Model Choice

Training

Evaluation

Computational Complexity

7/30/2019 PR_UNIT-I


12/31/2010 SAM_PR_UNIT-I 23

The Design Cycle cont’d

7/30/2019 PR_UNIT-I


12/31/2010 SAM_PR_UNIT-I 24

The Design Cycle cont’d Data Collection

How do we know when we have collected an

adequately large and representative set of examples

for training and testing the system?

Feature Choice

Depends on the characteristics of the problem

domain. Simple to extract, invariant to irrelevant

transformation insensitive to noise.

Model Choice

Unsatisfied with the performance of our fish classifier

and want to jump to another class of model

7/30/2019 PR_UNIT-I


12/31/2010 SAM_PR_UNIT-I 25

The Design Cycle cont’d Training

Use data to determine the classifier. Many differentprocedures for training classifiers and choosing

models Computational Complexity

What is the trade-off between computational easeand performance?(How an algorithm scales as a function of the number

of features, patterns or categories?) Evaluation

Measure the error rate (or performance and switchfrom one set of features to another one

7/30/2019 PR_UNIT-I


12/31/2010 SAM_PR_UNIT-I 26

LEARNING & ADAPTATION

Any method that incorporates information from trainingsamples in the design of a classifier employs learning.

We use learning because all practical or interesting PRproblems are so hard that we cannot guess

classification decision ahead of time. Approach: Assume some general form of model

Use training patterns to learn or estimate theunknown parameters.

7/30/2019 PR_UNIT-I


12/31/2010 SAM_PR_UNIT-I 27


Supervised Learning

Teacher provides a label or cost for each pattern ina training set.

Objective: Reduce the sum of the costs for thesepatterns

Issues: How to make sure that the learningalgorithm

Can learn the solution.

Will be stable to parameter variation.

Will converge in finite time.Scale with # of training patterns & # of inputfeatures.

Favors "simple" solutions.

7/30/2019 PR_UNIT-I


12/31/2010 SAM PR UNIT I 28


Unsupervised Learning (Clustering)

There is no explicit teacher.

System forms clusters or "natural grouping" of theinput patterns.

Reinforcement Learning (Learning with a critic)

No desired category is given. Instead, the onlyteaching feedback is that the tentative categoryis right or wrong.

Typical way to train a classifier:

Present an inputCompute its tentative label

Use the known target category label toimprove the classifier.

pr_unit-i

Documents