ISA5571: Audio Signal Processing and Recognition (音訊處理與辨識)

Roger Jang (張智星)

Objectives

To introduce fundamentals of audio processing that is commonly used for modern applications, such as speech recognition, speaker identification/verification, query by singing/humming, audio classification, and so on.
To cover essential hands-on techniques in MATLAB programming for audio signal processing and recognition.

本課程之目的在提供同學對音訊（Audio Signals）處理的基本概念，並學習如何使用 MATLAB 電腦程式來實做各種音訊處理的方法，以方便進行對音訊的分析、辨識、合成等，符合音訊在多媒體方面的應用需求。

Prerequisites

Elementary calculus
Linear algebra
Probability

Instructor and TAs

Instructor: Roger Jang (張智星), , Room 444, Ext: 42796, mobile: 0966-646464
TA: Ken Yeh (葉子雋), , mobile: 0955-880642

Course material

Textbooks (mostly available online)
- Audio Signal Processing and Recognition (音訊處理與辨識)
- Data Clustering and Pattern Recognition (資料分群與樣式辨認)
- MATLAB tutorials:
  1. MATLAB Primer (in English) by Sigmon
  2. 02-初探MATLAB (in Chinese) and examples
  3. 20-音訊讀寫、錄製與播放 (in Chinese) and examples
Reference material
- "Spoken Language Processing" by X. Huang, A. Acero, and H.-W. Hon
- "MATLAB程式設計【入門篇】" and "MATLAB程式設計【進階篇】" by Roger Jang (Both in Chinese)

Weekly Schedule and lecture formats

	Monday
15:20~17:00	Lecturing (EECS 224)
17:00~22:00(?)	Lab section (EECS 328)

Every Monday starting from 3:20 PM, we have about 2-hour lecture on fundamentals of audio signal processing/recognition.
(每週上課約 1~2 小時，說明音訊處理的原理。)
For the following 4~6 hours, we have the lab session in the computer room. You are expected to finish the programming assignments and demo your result to TA during the lab session. TA will stay in the computer room and help you with the assignments throughout the lab session.
(上課後，其餘約有 4~6 小時讓同學在電腦教室進行作業的實做，並展示給助教看。助教會在電腦教室幫忙，並進行評分。)
Basically you are expect to spend about 7 hours per week on this course on Monday. The bad news is that it could be more than 7 hours per week. The good news is that you don't need to spend any other time on this course for the rest of the week (unless for midterm/final and term project).
(基本上希望每位同學每週花在本課程的時間至少是 7 小時。壞處是每個星期一下午三點到十點（或更晚！），你一定要花時間在這門課；好處是回家後，你可以不用再花時間在這門課。當然，碰到考試或期末報告，你還是多少要花些時間。)
For the term project, we have two tracks:
- Paper based: You need to find a paper of your interests and present the paper in the class. In particular, you need to implement the experiments and compare your results to those in the paper.
- Project based: You can propose a mini project of audio/speech application, and carry it out on Android phones. We have 5 HTC Heros for this part.

Grading Policy

40% or so: 10~15 Programming assignments and several other recording tasks
40% or so: Midterm or final (programming-based)
15% or so: Final project
5% or so: Course interactions (including mandatory two questions from each students during the semester, which accounts for 2%)

Other requirements of the course (選修本課程的其他需求)

You need to have a USB flash drive (or the like) to store your programs. This is absolutely necessary since each computer in the computer room is set to erase everything you stored in the local drives (c: or d:) once it is rebooted after accidental crashes.
You need to have a fast computer, the faster, the better. You can bring your own notebook if necessary.
(需要有一台夠快的 PC，越快越好！)
You computer should be equipped with MATLAB, CoolEdit, speakers, and microphone.
(電腦必須配備 MATLAB、CoolEdit、喇叭、麥克風。)
Your microphone should be uni-directional instead of omni-directional.
(麥克風必須是指向式，而不是全向式。)
You need to get hands-on experiences by using MATLAB for the programming assignments.
(一定需要實際動手寫 MATLAB 程式。)

Back to ISA5571 Home