Home page

SPEECH AND AUDIO PROCESSING AND RECOGNITION

ELEN E6820 - Spring 2001

News

2001-04-25: Congratulations to everybody on yesterday's project presentations, which I thought went very well.
Remember to hand in your project reports by next Monday (April 30th) and please also remember to fill in the Web-based Course Evaluation for this class -- thanks!
2001-04-17: This week's slides are here. Next week (Tuesday April 24th) will be the project presentations class. All on-campus students are expected to make a 12 minute oral presentation of their project. I have slightly elaborated the guidelines on the projects page. You should complete your project reports by the last day of classes, Monday April 30th.
2001-04-10: This week's slides are here. No homework this week to give you time for your projects.
2001-04-03: This week's lecture slides are here; we will continue with the HMM training material next week. This week's homework is here.
2001-03-28: This week's lecture slides are here, although note that we didn't actually get onto the acoustic classifier part. I've also put all the linked sound examples in that same directory, so they should work if you view the slides from a web browser. This week's homework is up on the problem sets page; after our discussion in class, I've tried to make it a little less burdensome: this week's paper is just 4 pages, I simplified the practical, and I dropped the book questions.

General Information

Instructor: Dan Ellis
<[email protected]>
Schapiro Research building room 718

Instructor office hours: Mon 2-4pm

Required 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, 4:10-6:40

Lecture room: 535 Mudd

Credits: 4.5

Course web site: http://www.ee.columbia.edu/~dpwe/courses/e6820-2001-01

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, the largest of which will be automatic speech recognition, with the goal of understanding all the components of a modern large-vocabulary continuous speech recognizer. We will also consider topics such as high-quality psychoacoustic-based compression schemes (exemplified by the ubiquitous 'MP3'), and various issues in content-based retrieval for audio data.

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 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, midterm and final exams, and a term project. The grade will be broken down as follows:

Problem sets: 20%

Mid-term exam: 20%

Final exam: 20%

Project: 40%

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.

Projects

For details and suggestions, see the separate projects page.

Course outline

See the course outline page.

Dan Ellis <[email protected]>
Last updated: Wed Apr 25 18:33:22 EDT 2001