1. Pattern Recognition 9/23/2008 Instructor: Wen-Hung Liao, Ph.D.

2. Administrative Information E-mail: whliao@cs.nccu.edu.tw Office: 大仁樓三樓 Office hours: TBA Course web page: http://www.cs.nccu.edu.tw/~whliao/pr2008/ Textbook: Pattern Classification, 2 nd Edition by Duda, Hart and Stork. ( 歐亞書局 )

3. Definition Pattern recognition is the study of how machines can - observe the environment, - learn to distinguish patterns of interest, - make decisions about the categories of the patterns.

4. What is a pattern? Watanabe defines a pattern “as opposite of chaos; it is an entity, vaguely defined, that could be given a name”. Examples: - fingerprint, - handwritten cursive word, - a human face, - a speech signal... Examples

5. Types of recognition Supervised classification: input pattern is identified as a member of a pre-defined class. Unsupervised classification: input pattern is assigned to a hitherto unknown class.

6. Applications

7. Applications (2)

8. Components of a PR System Data acquisition and pre-processing Data representation Decision making

9. Pattern Recognition Methods Template matching Statistical approach Syntactic approach Neural networks

10. Template Matching A template (typically a 2D shape) or a prototype of the pattern to be recognized is available. Compute the similarity between the template and the pattern to be matched. Take into account pose(rotation, translation) and scale changes.

11. Issues of concern Choice of template Computational complexity Rigidity assumption (use deformable template models)

12. Statistical Approach Each pattern is represented in terms of d features, and is viewed as a point in a d- dimensional space The goal is to choose those features that allow pattern vectors belonging to different categories to occupy compact and disjoint regions in a d-dimensional feature space.

13. Syntactic Approach Use hierarchical structures to represent complex patterns. The simplest unit is called: primitives Complex pattern is represented in terms of the interrelationships (grammars) between the primitives. Grammatical rules can be learned by training.

14. Issues of concern Can be used in situations where the patterns have a definite structure such as EKG waveforms, shape analysis of contours. However, it’s usually difficult to segment noisy patterns and infer grammar from the training set. May yield a combinatorial explosions of possibilities to be investigated.

15. Neural Networks Massively parallel computing systems consisting of an extremely large number of simple processors with many interconnections. Can learn complex non-linear input-output relationships. Feed-forward networks such as multilayer perceptron and Radial Basis Function network are useful for pattern classification.

16. Pattern Recognition Models

17. Reference A.K. Jain, R.P.W. Duin and J. Mao, “Statistical Pattern Recognition: A Review”, IEEE Transactions on Pattern Analysis and Machine Intelligence (PAMI), Vol. 22, No. 1, pp. 4-37, Jan. 2000.