ABSTRACT
The most common form of multitrack automatic mixing system is based around simple level adjustments on each track. An offline system described by D. Ward et al. attempted to control faders with auditory models of loudness and partial loudness. Of all the standard audio effects found on a mixing console or as built-in algorithms in a digital audio workstation, there has perhaps been the least effort on furthering understanding of reverberation. Mixing decisions are based solely on the extracted, low-level features of the signals and no high-level semantic information, such as which instruments the incoming tracks accommodate or the genre of the song, is provided by the user or extracted by the system. In one of the earliest automatic mixing papers, E. Perez Gonzalez and J. D. Reiss described an autonomous panning system based on several best practices in music production.
