SPEECH AND AUDIO PROCESSING AND RECOGNITION

Home page

Course outline

Matlab scripts

Problem sets

Projects

Columbia Courseworks

Announcements

2009-01-20

Today's lecture is canceled so we can watch the inauguration. First lecture on Thursday!

2009-01-13

Welcome to the Spring 2009 edition of this class!

General Information

Instructors:	Dan Ellis <dpwe@ee.columbia.edu> Schapiro CEPSR room 718 Michael Mandel <mim@ee.columbia.edu> Schapiro CEPSR room 7LE4
Instructor office hours:	Ellis: Thursdays, 14:00-16:00
Text:	Speech and Audio Signal Processing: Processing and perception of speech and music Ben Gold & Nelson Morgan, Wiley 2000 (ISBN: 0-471-35154-7)
Lectures:	Tuesdays and Thursdays, 11:00-12:15 214 Pupin
Credits:	4.5
Course web site:	http://www.ee.columbia.edu/~dpwe/e6820/

Overview

Audio processing is a huge field, so this course will necessarily be eclectic rather than comprehensive. The first half of the course will cover fundamentals in signal processing, pattern recognition, acoustics and auditory perception. In the second half we will look at several application areas, including psychoacoustic compression schemes (such as MP3), speech recognition, music analysis and retrieval, and sound mixture organization.

Objectives

This course will give students a foundation in current audio and recognition technologies. One objective is to build up a familiarity with the perceptually-salient aspects of the audio signal, and how they can be extracted and manipulated through signal processing. A second related but separate objective is to obtain a thorough understanding of the statistical pattern recognition technology at the core of contemporary speech and audio recognition systems. Thirdly, the course aims to deepen each student's familiarity with the practical application of signal processing in general, through the study of specific instances, and through the experience of the term project.

Prerequisites

The course assumes a familiarity with signals and systems. We will be working in the discrete-time domain, so a basic DSP course such as ELEN E4810 is most suitable. The material on pattern recognition assumes a basic familiarity with probability, including Bayes' theory.

Grade structure

The course consists of lectures each week, weekly problem sets, a midterm event, and a final project. The grade will be broken down as follows:

Problem sets:	25%
Mid-term event:	25%
Project:	50%

The mid-term portion of the grade comes from a collective assessment of the mid-term project proposals i.e. each student will make a brief presentation of their proposed projects, and the other students will assign the grades based on the quality of the project and presentation.

Homework

Problem sets will be announced on the problem sets page of this web site directly after each lecture. They will be due by the time of the following lecture, at which time the solutions will be posted on the website. Late problems sets will not be accepted.

Most problem sets will include some kind of practical that will be run under the numerical computation package Matlab; you will need access to a machine running this software. For more information, see the course Matlab page.

Projects

For details and suggestions, see the separate projects page.

Course outline

See the course outline page.

Acknowledgment

This material is based in part upon work supported by the National Science Foundation under Grant No. IIS-0238301. Any opinions, findings and conclusions or recomendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation (NSF).

Dan Ellis <dpwe@ee.columbia.edu>
Last updated: Thu Jan 22 08:58:58 EST 2009