Perceptually Motivated, Intelligent Audio Mixing Approaches for Hearing Loss
Abstract
The growing population of listeners with hearing loss, along with the limitations of current audio enhancement solutions, have created the need for novel approaches that take into consideration the perceptual aspects of hearing loss, while taking advantage of the benefits produced by intelligent audio mixing. The aim of this thesis is to explore perceptually motivated intelligent approaches to audio mixing for listeners with hearing loss, through the development of a hearing loss simulation and its use as a referencing tool in automatic audio mixing. To achieve this aim, a real-time hearing loss simulation was designed and tested for its accuracy and effectiveness through the conduction of listening studies with participants with real and simulated hearing loss. The simulation was then used by audio engineering students and professionals during mixing, in order to provide information on the techniques and practices used by engineers to combat the effects of hearing loss while mixing content through the simulation. The extracted practices were then used to inform the following automatic mixing approaches: a deep learning approach utilising a differentiable digital signal processing architecture, a knowledge-based approach to gain mixing utilising fuzzy logic, a genetic algorithm approach to equalisation and finally a combined system of the fuzzy mixer and genetic equaliser. The outputs of all four systems were analysed, and each approach’s strengths and weaknesses were discussed in the thesis. The results of this work present the potential of integrating perceptual information into intelligent audio mixing production for hearing loss, paving the way for further exploration of this approach’s capabilities.
Authors
Mourgela, ACollections
- Theses [4223]