Piano Timbre Development Analysis using Machine Learning
This addresses the challenge of objectively measuring subtle timbre changes in musical instruments for acousticians and musicians, but it is incremental as it builds on prior work with specific feature analysis.
The study tackled the problem of analyzing timbre development in a concert grand piano over time by applying machine learning to psychoacoustic features, finding that spectral flux perfectly clusters recordings from two stages (new and after one year of use) and that other features like SPL and fractal dimension reveal sub-clusters indicating homogenization and ordering of key regions.
A data set of recorded single played tones of a concert grand piano is investigated using Machine Learning (ML) on psychoacoustic timbre features. The examined instrument has been recorded at two stages: firstly right after manufacture and secondly after being played in a concert hall for one year. A previous study [Plath2019] revealed that listeners clearly distinguished both stages but no clear correlation with acoustics, signal processing tools or verbalizations of perceived differences could be found. Using a Self-Organizing Map (SOM), training single as well as double feature sets, it can be shown that spectral flux is able to perfectly cluster the two stages. Sound Pressure Level (SPL), roughness, and fractal correlation dimension (as a measure for initial transient chaoticity) are furthermore able to order the keys with respect to high and low notes. Combining spectral flux with the three other features in double-feature training sets maintains stage clustering only for SPL and fractal dimension, showing sub-clusters for both stages. These sub-clusters point to a homogenization of SPL for stage 2 with respect to stage 1 and a pronounced ordering and sub-clustering of key regions with respect to initial transient chaoticity.