Using sequence alignment and voting to improve optical music recognition from multiple recognizers

TY - GEN

T1 - Using sequence alignment and voting to improve optical music recognition from multiple recognizers

AU - Bugge, Esben Paul

AU - Juncher, Kim Lundsteen

AU - Mathiesen, Brian Søborg

AU - Simonsen, Jakob Grue

N1 - Conference code: 12

PY - 2011

Y1 - 2011

N2 - Digitalizing sheet music using Optical Music Recognition (OMR) is error-prone, especially when using noisy images created from scanned prints. Inspired by DNA-sequence alignment, we devise a method to use multiple sequence alignment to automatically compare output from multiple third party OMR tools and perform automatic error-correction of pitch and duration of notes. We perform tests on a corpus of 49 one-page scores of varying quality. Our method on average reduces the amount of errors from an ensemble of 4 commercial OMR tools. The method achieves, on average, fewer errors than each recognizer by itself, but statistical tests show that it is significantly better than only 2 of the 4 commercial recognizers. The results suggest that recognizers may be improved somewhat by sequence alignment and voting, but that more elaborate methods may be needed to obtain substantial improvements. All software, scanned music data used for testing, and experiment protocols are open source and available at: http://code.google.com/p/omr-errorcorrection/.

AB - Digitalizing sheet music using Optical Music Recognition (OMR) is error-prone, especially when using noisy images created from scanned prints. Inspired by DNA-sequence alignment, we devise a method to use multiple sequence alignment to automatically compare output from multiple third party OMR tools and perform automatic error-correction of pitch and duration of notes. We perform tests on a corpus of 49 one-page scores of varying quality. Our method on average reduces the amount of errors from an ensemble of 4 commercial OMR tools. The method achieves, on average, fewer errors than each recognizer by itself, but statistical tests show that it is significantly better than only 2 of the 4 commercial recognizers. The results suggest that recognizers may be improved somewhat by sequence alignment and voting, but that more elaborate methods may be needed to obtain substantial improvements. All software, scanned music data used for testing, and experiment protocols are open source and available at: http://code.google.com/p/omr-errorcorrection/.

M3 - Article in proceedings

SN - 978-0-615-54865-4

SP - 405

EP - 410

BT - Proceedings of the 12th International Society for Music Information Retrieval Conference

A2 - Klapuri, Anssi

A2 - Leider, Colby

PB - University of Miami

T2 - 12th International Society for Music Information Retrieval Conference

Y2 - 24 October 2011 through 28 October 2011

ER -