Ai Driven Multitrack Audio Mixing Using Fuzzy Logic Tools

The art of audio mixing is an inherently subjective and nuanced process that requires years of experience to master. This thesis explores a novel use of AI to overcome some of the challenges of multitrack audio mixing for the non-expert. This thesis discusses data representation, common audio engineering practices, modeling of sounds into the stereo field, and proposes a solution for multitrack audio mixing using an artificial intelligent system incorporating fuzzy logic tools in Matlab. The artificial intelligent system is comprised of three modules that are each independently responsible for one of three dimensions in the sonic space. It is shown that by using fuzzy logic for the inherently fuzzy problem of audio mixing, a non-expert can achieve an improved mix of multiple audio tracks. For the purpose of testing, 8 multitrack songs consisting of varying instrumentation and styles were recorded and processed through the AI system with improvements in the resultant mixes.