| HAL-Inria | Publications, software ... of Inria's scientists |
Conference papers
Automatic transcription of piano music based on HMM tracking of jointly-estimated pitches
Abstract : This work deals with the automatic transcription of piano recordings into a MIDI symbolic file. The system consists of subsequent stages of onset detection and multipitch estimation and tracking. The latter is based on a Hidden Markov Model framework, embedding a spectral maximum likelihood method for joint pitch estimation. The complexity issue of joint estimation techniques is solved by selecting subsets of simultaneously played notes within a pre-estimated set of candidates. Tests on a large database and comparisons to state-of-the-art methods show promising results.
Cited literature [16 references]
https://hal.inria.fr/inria-00452620
Contributor : Valentin Emiya Connect in order to contact the contributor
Submitted on : Tuesday, February 2, 2010 - 5:04:00 PM
Last modification on : Tuesday, December 8, 2020 - 10:21:47 AM
Long-term archiving on: : Friday, June 18, 2010 - 6:27:05 PM
Contributor : Valentin Emiya Connect in order to contact the contributor
Submitted on : Tuesday, February 2, 2010 - 5:04:00 PM
Last modification on : Tuesday, December 8, 2020 - 10:21:47 AM
Long-term archiving on: : Friday, June 18, 2010 - 6:27:05 PM
File
080825_Eusipco_emiya_final.pdf
Publisher files allowed on an open archive