Using machine-learning methods for musical style modeling

Abstract

Building on the work of Leonard B. Meyer, 3 researchers commonly agree that expectations based on recent past context guide musical perception. In music applications, exactly how we make musical Research using statistical and information-theoretic tools provides inference and prediction models that, to a certain extent, can generate new musical works imitating the style of the great masters.

Algorithmic Clustering of Music Based on String Compression

“…Another very interesting line of music research using compression-based techniques may be found in the survey by Dubnov et al (2003) and the references therein. In that survey, the aim is not to cluster similar musical pieces together, but to model the musical style of a given MIDI file.…”

Predictive models for music

“…While a few generative models have already been proposed for music in general [7,18], we are not aware of proper quantitative comparisons between generative models of music, as it is done in Sections 3 and 5.…”

“…What we really want to measure is how much gain we observe in terms of out-ofsample prediction accuracy when using an arbitrary model if we impose additional constraints based on distance patterns. That being said, it would be interesting to measure the effect of appending distance constraints to more complex music prediction models [7,18]. Results in Table 1 for the jazz standards database show that considering distance patterns significantly improves the HMM model.…”

Melodic analysis with segment classes

“…The representation of segments by the set of properties of component events has similarities to work on statistical models of harmony and polyphonic musical structures (Pickens & Crawford, 2002;Dubnov et al, 2003;Temperley, 2004). In these works, voicing or temporal information in a simultaneity is often discarded in favor of more abstract representations in terms of pitch class profiles.…”