An Evolutionary Computation Approach to Intelligent Music Production Informed by Experimentally Gathered Domain Knowledge

A novel methodology for intelligent music production has been developed using evolutionary computation. Mixes are generated by exploration of a “mix-space”, which consists of a series of inter-channel volume ratios, allowing efficient generation of random mixes. An interactive genetic algorithm was used, allowing the user to rate mixes and guide the system towards their ideal mix. Currently, fitness evaluation is subjective but can be aided by specific domain knowledge obtained from a large-scale study of real mixes. 1. BACKGROUND Intelligent music production (IMP) has been an active research topic for over a decade. One aim is the development of systems which perform common tasks: level-balancing, equalisation, panning, dynamic range compression and application of artificial reverberation. Many previous IMP systems developed were modelled as expert systems wherein a music production task is solved by optimisation, and domain knowledge, obtained by examining industry “bestpractice” methods, is used to determine the optimisation target [1]. Drawbacks to this method include the fallibility of this type of domain knowledge and the fundamental assumption that there is a global optimum, i.e. one mix which all users would agree is best. Subjective evaluation suggested that existing systems struggled to compete with human-made mixes [2], perhaps due to a lack of what we would perceive as creativity. Additionally it has been suggested that mix engineers prefer their own mix to those of their peers [2]. Consequently, IMP tools would benefit from increased interactivity and subjectivity, to determine userspecific “personal” global optima in the solution space, instead of a single “universal” global optimum. 2. CONCEPT We propose to use interactive evolutionary computation (IEC) to solve this problem, being well-suited to aesthetic design problems which are non-linear and non-deterministic [3]. The flowchart in Fig. 1 demonstrates the method, with an interactive genetic algorithm (IGA). The solution space we explored is a “mix-space” which theoretically represents all the mixes that it is possible to create with a finite set of tools [4]. For level-balancing, the gains g of all n tracks are selected from a unit hypersphere in Rn. This hypersurface has n−1 dimensions, representing a series of inter-channel Genetic operators Set-up Clustering and fitness evaluation Create mix START Import Audio Normalise loudness of files Initialise population Evaluate fitness of sub-population Is stop condition met?