Intelligent Multitrack Dynamic Range Compression

Abstract

We present an intelligent approach to multitrack dynamic range compression where all parameters are configured automatically based on side-chain feature extraction from the input signals. A method of adjustment experiment to explore how audio engineers set the ratio and threshold is described. We use multiple linear regression to model the relationship between different features and the experimental results. Parameter automations incorporate control assumptions based on this experiment and those derived from mixing literature and analysis. Subjective evaluation of the intelligent system is provided in the form of a multiple stimulus listening test where the system is compared against a no-compression mix, two human mixes, and an alternative approach. Results showed that mixes devised by our system are able to compete with or outperform manual mixes by semi-professionals under a variety of subjective criteria.

A History of Audio Effects

“…Within this field, there has been a large focus on the use of some signal analysis approaches to automate, or directly control, some parameters of a preexisting audio effect. This can be performed by machine learning from data [153,154], performing some curve fitting or mapping to some higher-level parameters [155,156] or using some direct signal analysis to control an audio effect parameter directly [157,158].…”

All About Audio Equalization: Solutions and Frontiers

“…Yet those domains are not fully independent. Previous research [122,123] has shown that it is good practice to set dynamic range compressor parameters based on the frequency content in the signal, and many problems in audio production can be addressed by using combinations of filtering and dynamics processing. Thus variants often address specific functionality such as de-essing or hum removal, and as such have limited configurability beyond their applications.…”

A Machine-Learning Approach to Application of Intelligent Artificial Reverberation

“…There has been a lot of research in adaptive digital audio effects for automatic multitrack mixing but in almost all cases they focus on achieving a pre-specified goal. Parameter automation and intelligent control have been applied to many of the most popular audio effects (e.g., gain and faders [5], equalization [6], panning [7], and dynamic range compression [8]), but to the best of the authors' knowledge it has not been attempted on artificial reverberation. Furthermore, all of the above mentioned approaches except [5], which uses linear dynamical systems to estimate mixing weight coefficients, use fixed rules rather than arbitrary rules that are learned from training data.…”