Syllabus

Southern Oregon University

Department of Computer Science

CS 455 Speech Recognition

crn 3853 (undergraduate), crn 3854 (graduate)

Prerequisite: cs258

 

Instructor Information

 

           Instructor:   Dan Harvey

           Room:         Computer Science Building #CS218

           Phone:        552-6149

           E-mail:       harveyd@sou.edu

 

           Office Hours: Monday, Wednesday    10:00-12:00, 1:00-2:00

 

           Web Site:     http://cs.sou.edu/~harveyd    

 

The web site is available for quiz results, lab assignments, weekly handouts, current grade status, and contact with class members. Click on the appropriate class, and then select the desired option.

 

Class Times

 

Monday, Wednesday (CS224)   3:00 to 4:50

 

Course Text

 

Spoken Language Processing:
A guide to theory, algorithm, and system development

Huang, Acero, Hon
Prentice Hall, 2001, ISBN 0-13-022616-5

 

Course Objectives

 

This class introduces students to speech recognition techniques. We will discuss why speech recognition is difficult. For example,how is is possible for a computer to recognize multilingual speech. The class will focus on the back end speech recognition processes after algorithms execute to extract a feature vectors from the raw sound signal. Among the algorithms we will consider are Hidden Markov, Viterbi, pattern recognition, vocabulary search, and phonetic analysis. We will describe the acoustic and language models that are popular for modern speech recognition systems.

 

Tentative Chapter Coverage

 

Week 1   Chapter 1 + handout  Introduction and Introduction to DSP

Week 2   Chapter 2            Spoken Language Structure

Week 3   Chapter 3            Probability, Statistics Background

Week 4   Chapter 4            Pattern Recognition

Week 5   Chapter 8            Hidden Markov Models

Week 6   Chapter 9            Acoustic Modeling

Week 7   Chapter 10           Environment Robustness

Week 8   Chapter 11           Language Models

Week 9   Chapter 12, 13       Search Algorithms

Week 10  Chapter 17           Spoken Language Understanding

 

 

Course Grading

 

There will be three twenty minute student presentations. Each of these presentations will require students to read and present the contents of a published scholarly paper. Grades measure quality of the presentation and understanding of the subject matter. The presentations count for 20 percent of the final grade.

 

A quarter long project requires students to implement an algorithm related to speech recognition. The project is due during finals week. A final technical paper will describe the project in detail. Students will prepare a technical paper, power point slides, and keep a research journal during the quarter. The project grade bill measure the quality of the design specification, the final technical paper, the research journal, how well the software works, and the presentation. The project counts for 35 percent of the final grade.

 

Students will take a midterm examination in week nine of the quarter. This examination contributes thirty-five percent towards the final grade.

Additional homework assignments and attendance will count for ten percent of the final grade. There will also be opportunities for extra credit.

 

Grade Breakdown:                      93-100% A      90-92% A-

                            88-89%   B+     82-87   B      80-81% B-

                            78-79%   C+     72-77   C      70-71% C-

                            68-69%   D+     62-67   D      60-61% D-

                            Under 60 F

 

Disabilities

 

If you are in need of academic support because of a documented disability (whether it be learning, mobility, psychiatric, health-related, or sensory) you may be eligible for academic or other accommodations through Disability Services for Students. Contact Disability Services for Students; Director DSS 552-6213, or schedule an appointment in person at the ACCESS Center, Stevenson Union, lower-level.