Department of Electrical and Computer Engineering

ECE 568 Pattern Classification

Classification models, discriminant functions, decision surfaces, generalized linear discriminant functions, parameter estimation, problems of dimensionality, component analysis, Fisher discriminant analysis, hidden Markov models, nearest neighbor rules, classification trees, string matching, resampling for classifier design and evaluation, clustering algorithms, projects.

Credit Hours: 3 Lecture

Prerequisites: Consent of instructor

Course Coordinator: Lalit Gupta

Textbooks:

1. R.O. Duda, P.E. Hart, & D.G. Stork, Pattern Classification, John Wiley & Sons, 2001.

References:

1. Journal Papers.

Goals:

1. To introduce graduate students to the major topics in pattern classification and applications in automatic target recognition, human-computer interfacing, biometric identification, bioinformatics, biomedical signal classification, and industrial inspection.

Projects:

1. Multidimensional data generation, distance measures, decision surfaces.

2. Principal component analysis and Fisher linear discriminant analysis.

3. Nearest neighbor and nearest mean classification.

4. Hidden Markov models and dynamic time warping.

5. Resampling for classifier design and performance evaluation.

6. Clustering algorithms.

Computer Tools: Matlab

Last Review: Spring Semester 2004