TY - GEN
T1 - "Copy and Scale" method for doing time-localized M.I.R. estimation
T2 - 3rd ACM International Workshop on Machine Learning and Music, MML'10, Co-located with ACM Multimedia 2010
AU - Peeters, Geoffroy
PY - 2010/12/1
Y1 - 2010/12/1
N2 - In this work we propose a "copy and scale"method based on the 1-NN paradigm to estimate time-localized parameters and apply it to the problem of beat-tracking. The 1-NN algorithm consists in assigning the information of the closest item of a pre-annotated database to an unknown target. It can be viewed as a "copy and paste"method. The "copy and scale" method we propose consists in "scaling" this information to adapt it to the properties of the unknown target. For this, we first represent the content of an audio signal using a sampled and tempo-normalized complex DFT. This representation is used as the vectors over which the 1-NN search is performed. Along each vector of the 1-NN space, we store the corresponding annotated beat-marker positions in a normalized form. Once the closest vector is found, its tempo is assigned to the unknown item and the normalized beat-markers are scaled to this tempo in order to provide the estimation of the unknown item beat-markers. We perform a preliminary evaluation of this method and show that, with such a simple method, we can achieve results comparable to the ones obtained with sophisticated approaches.
AB - In this work we propose a "copy and scale"method based on the 1-NN paradigm to estimate time-localized parameters and apply it to the problem of beat-tracking. The 1-NN algorithm consists in assigning the information of the closest item of a pre-annotated database to an unknown target. It can be viewed as a "copy and paste"method. The "copy and scale" method we propose consists in "scaling" this information to adapt it to the properties of the unknown target. For this, we first represent the content of an audio signal using a sampled and tempo-normalized complex DFT. This representation is used as the vectors over which the 1-NN search is performed. Along each vector of the 1-NN space, we store the corresponding annotated beat-marker positions in a normalized form. Once the closest vector is found, its tempo is assigned to the unknown item and the normalized beat-markers are scaled to this tempo in order to provide the estimation of the unknown item beat-markers. We perform a preliminary evaluation of this method and show that, with such a simple method, we can achieve results comparable to the ones obtained with sophisticated approaches.
KW - Algorithms
UR - https://www.scopus.com/pages/publications/78650902362
U2 - 10.1145/1878003.1878005
DO - 10.1145/1878003.1878005
M3 - Conference contribution
AN - SCOPUS:78650902362
SN - 9781450301619
T3 - MML'10 - Proceedings of the 3rd ACM International Workshop on Machine Learning and Music, Co-located with ACM Multimedia 2010
SP - 1
EP - 4
BT - MML'10 - Proceedings of the 3rd ACM International Workshop on Machine Learning and Music, Co-located with ACM Multimedia 2010
Y2 - 25 October 2010 through 25 October 2010
ER -