|
| EPSRC Reference: |
GR/S17246/01 |
| Title: |
Algorithms for Musical Pattern Recognition and Extraction |
| Principal Investigator: |
Professor C Iliopoulos |
| Other Investigators: |
|
| Researcher Co-investigator: |
|
| Project Partner: |
|
| Department: |
Computer Science |
| Organisation: |
King's College London |
| Scheme: |
Standard Research |
| Starts: |
01 August 2003 |
Ends: |
31 July 2006 |
Value (£): |
176,824
|
| EPSRC Research Topic Classifications: |
| Fundamentals of Computing |
|
|
| EPSRC Industrial Sector Classifications: |
|
| Related Grants: |
|
| Panel History: |
|
|
Summary |
The recent explosive increase in the electronic storage of music as computer files, especially on the Internet, gives rise to the need to design new computer methods to answer questions like: How can we retrieve a specific tune among a large number of musical scores? What is the test way of identifying a melody? How can we find musical works that contain a certain motivic pattern or theme? How can significant, recurring musical patterns be automatically extracted from a score and then used for indexing purposes?
We are aiming to design and implement efficient algorithms for several variants of monophonic and polyphonic motif matching in musical sequences. In particular we will design new methods for multi-dimensional approximate motif matching, in order to deal with the presence of errors in musical sequences, compact motif matching in order to deal with temporal gaps, and in order to improve the "quality" of matches we will develop special methods for rhythmic similarity motif matching, metrical congruity motif matching, salient motif matching and harmonic progression motif matching.
|
| Final Report Summary |
The recent explosive increase in the electronic storage of music as computer files, especially on the Internet, gives rise to the need to design new computer methods to answer questions like: How can we retrieve a specific tune among a large number of musical scores? What is the test way of identifying a melody? How can we find musical works that contain a certain motivic pattern or theme? How can significant, recurring musical patterns be automatically extracted from a score and then used for indexing purposes?
We have designed and implemented several algorithms addressing various aspects of these problems: one interprets the notes in preformance control data (eg MIDI) and decides what is the right way to notate them in score format; one looks at the harmonic relations between different pieces of music and compares them on this basis; the third compares two pieces, note by note, and computes the largest part of one which is quoted in the other; and the last uses ideas from a related grant to detect ornamentation in performed music. All these algorithms are useful for Music Information Retrieval and music analysis (in the musicological sense); the first and last are also useful for music transcription.
|
| Further Information: |
|
| Organisation Website: |
http://www.kcl.ac.uk |
|
|