Audio mixing is the process of applying a series of signal-processing operations to multiple audio sources and then combining them together to create a sound mixture. In this chapter we will explore signal-processing algorithms which aim to automate this process. The first automatic mixing implementations for live sound can be traced to [Dug75, Dug89]. Here the basic principles of automatic level adjustment for automatic mixing were proposed. This system is able to maintain constant gain structure regardless of the number of active microphone inputs. Decisions on time-domain gain compensation were based on RMS measurements extracted from the signal. A similar system, but based on a mechanical principle called direction-sensitive gating for automatic microphone mixing was developed by [JT84]. Other implementations of automatic microphone mixing such as [Pet78, SBI00] have been implemented, but most of these designs concentrate on level microphone management for speech applications only. Such automatic microphone mixing implementations make use of low-level features, in order to take mixing decisions. In recent years, techniques have been described related to non-real-time automatic-mixing processing algorithms. In [TR09b] an automatic monitor mixing method for deriving fader levels using optimisation is described. In [Kol08] and [BR09] a method that reconstructs mixing parameter values of each channel through analysing a target mixture was proposed. In [Ree00] a method which uses nearest-neighbour optimisation techniques to attempt to recreate expert mixing is derived. [TR09a] presented a method for automatically setting noise-gate parameters for drum recording using optimisation. Finally, some work on perception and automatic detection of frequencies which require equalisation compensation has been performed by [BLS08, BJ09].
In this chapter we will look at current automatic-mixing techniques for real-time digital signal processing that are capable of panning, correcting polarity and time offset problems, enhancing channels, compensating gain and fader level, and a simple self-equalisation algorithm. We will concentrate on automatic-mixing designs for live music mixing as opposed to speech-oriented microphone systems. Automatic mixing has applications to live sound and music performance, remote mixing, recording and post-production as well as real-time mixing for interactive scenes and games. We will refer to these automatic-mixing digital effect tools as AM-DAFX.