MAT 240F: Digital Audio Programming: Audio Analysis and Music
Information Retrieval
UCSB, Spring Quarter, 2006
Downloads
Course Description
The
focus of the MAT 240F course is on
audio analysis and signal processing techniques applied to sound/music
databases and music information retrieval systems. We will work with
various C++ libraries for signal analysis and feature extraction to
develop skills in time-domain processes such as beat following, tempo
analysis, and song segmentation, and in spectral-domain analysis
techniques such as pitch estimation, spectral peak analysis and
tracking, and instrument signature identification. The topic of feature
vector design will play an important role in the development tasks.
Applications will include music segmentation, finger-printing,
thumb-nailing, clustering and genre classification, and user preference
matching.
Prerequisites
Students are expected to know the
basics of digital audio signal representation and processing, and to be
proficient in C, C++, or Java (Smalltalk and/or SuperCollider also
useful). Grading will be on the basis of in-class participation and
programming projects.
Course Materials
Readings will be distributed in class
and via the course web site. See the downloads section below.
Instructor
Stephen T. Pope (stp@create.ucsb.edu)
Meeting time and place
Time: Tuesday/Thursday, 12:00-2:00 PM
First meeting: April 4, 2006
MAT Graduate Program's multimedia class room (South Hall 3431)
Topical Outline: Audio Analysis and Music Information Retrieval
- Introduction
- Kinds of Audio Data and Metadata
- Dimensions of Music Information Retrieval Applications
- APIs for MIR Tools: Marsyas, D2K/M2K, LibOFA, FMAK, 8S
- Multimedia Databases
- Feature Vectors and Indexing
- Feature Extraction and Signal Analysis
- Numerical Processing: Clustering and Classification
- Basics of Audio Signal Processing
- Time Sequences, Windowing
- Analysis Domains and Transformations
- Multi-level Analysis and High-level Features
- Data Smoothing and Reduction Techniques
- Time-domain Audio Analysis and Applications
- Windowed RMS Envelope Extraction
- Beat Detection and Tempo Analysis
- Time-based signal segmentation
- Frequency-domain Analysis
- Pitch Detection Techniques
- Spectral Analysis and Interpretation
- Spectral Peaks and Tracking
- Other Spectral Measures
- Other Kinds of Analysis: Wavelets
- Cross-domain analysis
- Numerical and Psychological Processing
- Data Reduction and Smoothing
- Psychoacoustical Mapping
- Audio Segmentation and Musical Form
- Clustering and Classification
- Databases and Handling of Large or Dynamic Feature Vectors
- Application Requirements and Design
- Searching, Indexing, and Players
- Audio Summarization and Thumb-nailing
- Content Matching and Finger-printing
- Data Clustering and Genre Classification
- Other Applications: Mapping Systems
For more details, down-load the Reader Table of Contents (PDF file)
Web Links
Software used in MAT 240F
Music Information Retrieval Conference Proceedings
