Binaural Mixing Using a Recurrent Neural Network
Sciences and Mathematics, College of
Chemistry and Physics, Department of
SURS Faculty Advisor
Dr. Scott Hawley
Surround sound mixing for music is an area of heavy interest in music production. Over the past few years, the rise in popularity of the Dolby Atmos format resulted in an increased demand for surround sound mixing for musical projects. A song that has already been mixed in stereophonic format can be remixed in the Dolby Atmos format, placing audio sources in various locations in a three-dimensional space. Because of the limited availability of Dolby Atmos mixing services, most independent artists cannot afford to pay a professional mixer for their services. The goal of this research was to develop an artificial intelligence mixing assistant that could recreate a surround sound mix so that independent artists could hear what their music could sound like when mixed in the Dolby Atmos format. This mixing assistant was developed in tandem with the Blender application using the Python programming language in Google Colab. This mixing assistant was able to create static, binaural mixes based on positional data similar to that of Dolby Atmos mixes. While these mixes are not truly reflective of a Dolby Atmos mix performed by a professional surround sound mixer, they replicate a surround sound mix closely enough to help independent artists determine if it is worth it to hire a surround sound mixer to mix their records in the Dolby Atmos format.
Morgan, Grant, "Binaural Mixing Using a Recurrent Neural Network" (2022). Science University Research Symposium (SURS). 64.