TY - GEN
T1 - Algorithmic composition parameter as intercultural and cross-level MIR feature
T2 - 42nd International Computer Music Conference, ICMC 2016
AU - Lin, Hsin Ming
AU - Dubnov, Shlomo
N1 - Publisher Copyright:
© 2016 Hsin-Ming Lin et al.
PY - 2016/1/1
Y1 - 2016/1/1
N2 - Algorithmic composition (AC) and music information retrieval (MIR) can benefit each other. By compositional algorithms, scientists generate vast materials for MIR experiment; through MIR tools, composers instantly analyze abundant pieces to comprehend gross aspects. Although there are manifold musicologically valid MIR features, most of them are merely applicable to Western music. Besides, most high-level and low-level features are not interchangeable to retrieve from both symbolic and audio samples. We investigate the susceptibility of melodic pitch contour, a parameter from an AC model. It was created to regulate a generative monophonic melody's sensitivity to make a return after consecutive pitch intervals. It takes audio frequency values rather than symbolic pitch numbers into consideration. Hence we expect its intercultural and cross-level capabilities. To validate, we modify the original model from compositional to analytical functions. Our experimental results unveil a clear trend of mean susceptibilities from vocal to instrumental styles in 16522 samples from 81 datasets across numerous composers, genres, eras, and regions. We demonstrate the mutual benefits between AC and MIR. The parameter operates as an intercultural and crosslevel feature. The relationship between susceptibility and register width is surprising in several comparisons. Further investigation is ongoing to answer more questions.
AB - Algorithmic composition (AC) and music information retrieval (MIR) can benefit each other. By compositional algorithms, scientists generate vast materials for MIR experiment; through MIR tools, composers instantly analyze abundant pieces to comprehend gross aspects. Although there are manifold musicologically valid MIR features, most of them are merely applicable to Western music. Besides, most high-level and low-level features are not interchangeable to retrieve from both symbolic and audio samples. We investigate the susceptibility of melodic pitch contour, a parameter from an AC model. It was created to regulate a generative monophonic melody's sensitivity to make a return after consecutive pitch intervals. It takes audio frequency values rather than symbolic pitch numbers into consideration. Hence we expect its intercultural and cross-level capabilities. To validate, we modify the original model from compositional to analytical functions. Our experimental results unveil a clear trend of mean susceptibilities from vocal to instrumental styles in 16522 samples from 81 datasets across numerous composers, genres, eras, and regions. We demonstrate the mutual benefits between AC and MIR. The parameter operates as an intercultural and crosslevel feature. The relationship between susceptibility and register width is surprising in several comparisons. Further investigation is ongoing to answer more questions.
UR - http://www.scopus.com/inward/record.url?scp=85003456401&partnerID=8YFLogxK
M3 - Conference contribution
AN - SCOPUS:85003456401
T3 - ICMC 2016 - 42nd International Computer Music Conference, Proceedings
SP - 357
EP - 362
BT - ICMC 2016 - 42nd International Computer Music Conference, Proceedings
A2 - Timmermans, Hans
PB - HKU University of the Arts Utrecht, HKU Music and Technology
Y2 - 12 September 2016 through 16 September 2016
ER -