Modeling variations in onsets and dynamics in music performances by Pablo Alvarado
Presented by : Pablo Alejandro Alvarado
This poster explores the modeling of variations in onset location and dynamics in music performances.
By analyzing audio recordings, this study aims to uncover the underlying patterns that characterize the rhythmic and temporal features of diverse music styles.
The deviations of each music note onset is modeled as a stochastic function, specifically, as a Gaussian process. Using Bayesian inference it is possible to get posterior distribution over this function, accurately capturing and representing the variations in performance.
Additionally, this research aims to integrate generative models to create realistic backing tracks that maintain the stylistic rhythm features of human-like music interpretations, facilitating improvisation in an informed context.