Computational Modelling Group

Optimisation of Acoustic Systems for Perceived Sound Quality

19th September 2016
Research Team
Daniel Wallace
Jordan Cheer

Project workflow

Computational design optimisation has been used in a variety of applications to improve the performance of complex systems by reaching a trade-off between multiple design objectives within certain constraints. The design of acoustic systems to achieve high sound quality is a challenging problem, in part due to the complexity of human auditory perception, and as such has not exploited the potential of design optimisation. The aim of this project is to investigate the application of design optimisation methods to acoustic systems, where the objective is to maximise the perceived sound quality. This will assist designers by reducing the number of prototypes that need to be built and tested, and could be applied to both devices whose primary function is noise generation, for example loudspeakers, and devices that produce noise as a by-product of their primary function, for example household fans.

In order to exploit computational design optimisation methods in the context of improving the sound quality of acoustic systems, this project will require the development of advanced computational models of acoustic systems, the implementation of models of human auditory perception and sound quality, and their integration through advanced optimisation methods. The optimisation algorithm will use the outputs of the perceptual models in order to optimise parameters in the modelled acoustic system. The particular challenges in this project arise from the complexity of perceptual auditory models, which are often non-linear, and both frequency and time dependent. This means that the outputs of these models may lead to objective functions that are non-linear and possibly also non-convex. Consequently, the optimisation of the acoustic systems will not be achievable using standard gradient-based solvers. It will therefore also be necessary to investigate the application of advanced intelligent optimisation methods, such as population-based algorithms, that are able to seek useful solutions to these complex problems.

The project will require the development of novel approaches to solving computationally intensive, large-scale optimisation problems. As such, the project will exploit the multi-core processing avaiable from the supercomputing facilities at the University of Southampton. The interdisciplinary nature of the project will build links between acoustical engineering, subjective acoustics, computational modelling and optimisation and, therefore, will help to expand the benefits of computational modelling into new application areas.


Life sciences simulation: Neuroscience

Physical Systems and Engineering simulation: Acoustics

Socio-technological System simulation: Human environment interaction, Human population

Algorithms and computational methods: Artificial Neural Networks, Classification, Machine learning, Optimisation

Transdisciplinary tags: NGCM