Objective evaluation of spectral sparsity introduced by professional audio mixing practices in multi-track mixes
* Presenting author
Abstract:
Multitrack mixing is a cardinal facet of modern day music production. Although professional mixing engineers are concerned with the aesthetic aspects of the mix, the mathematical and psychoacoustical transformations implied by the practice are seldom studied. Considering the former, this study will attempt at testing the hypothesis of whether mixing increases sparsity in the spectral domain. Spectral sparsity could be a key feature to allow individual instruments to remain audible despite commonly large numbers of instruments that can be part of an overall mix. Using a constant-Q spectral transformation, entropy-based and probabilistic measures of sparsity will be applied to stems from around 100 different mixes from various musical genres to test this hypothesis. Specifically, this study will consider the Shannon entropy for its inherent numerical stability as well as the Gini index. An objective comparison of the aforementioned sparsity measures between un-mixed and mixed stems will be made to underpin the assertion. This will serve as a first step to better characterize the acoustical signal transformations of multitrack mixing.