Face Image Synthesis Using Nonlinear Manifold Learning

1
Introduction of Pattern Recognition
Prof. SungYang Bang IM Lab., CSE


Intelligent Multimedia Lab.
2
Multimedia

• Multimedia ( Multiple Media )
• A combination of text, graphics, images, video
  and audio to enrich its content and enhance
  communication.
• Physical device enabling the exchange of
  information between the user and the computer.

3
What we do

• Character recognition
• Face recognition
• Facial expression recognition
• Sound classification
• Bioinformatics
• Brain Computer Interface

4
Pattern Recognition by Humans by Computers

• Perception by Humans
• Understanding spoken language
• Recognition of faces
• Distinguishing odors
• Remarkable, but we dont know the secret.
• Pattern Recognition by Computers
• Digitize all data
• Use logic and mathematics
• A part of A.I.

5
Recognition Or Classification

• Pattern
• Something that can be given a name or a class.
• Recognition
• Etymologically, the act of thinking again
• Involves Identifying or acknowledging
• Classification
• Etymologically, the act of separating into groups
• Involves Associating to a group

6
Pattern Classification

Feature Extraction
Classification
Z
X
Y
Pattern space (data)
Decision Space
Feature space
7
The Classification Process
Training
Training data
class label
Testing data
Testing
8
Pre-processing

• Noise removal
• Segmentation
• Space ( i.e., character )
• Time interval ( i.e., speech )
• Normalization
• Size normalization ( to compensate for scaling )
• Baseline drift correction ( centering) etc

9
Feature Extraction (cont.)

• What is a feature ?
• A feature is a distinctive attribute or
  characteristic of a stimulus.
• e.g., T has 2 features ? l
• (E.Gibson, 1969)

10
Feature Extraction (cont.)

• Why are the features so important ?
• One of two major modules in PR systems
• Affecting the final recognition performance
  significantly (eg, variation between 9098
  depending on features)
• eg) classification of Salmon and Sea bass
• Good features length, width, color, mouth size
• Useless features number of eyes

11
Feature Extraction (cont.)

• How to select the features ?
• Removing useless, redundant, and/or less useful
  (less discriminative) features
• What features are salient for the classification?
• Are the features robust?
• Do they vary with parameters such as time,
  frequency, scale, translation, rotation, or
  proximity?
• Do subsets of the features provide classification
  efficacy?

12
Classification

13
Classification (cont.)

14
Classification (cont.)

• What are the classifier design objectives?
• Minimize classification error(s)
• Generalization
• Reduced computational complexity

15
Some sample problems

• Handwritten character/word recognition
• Speech recognition
• Radar target recognition
• Face recognition
• Fingerprint classification