Background road traffic noise synthesis

Licentiatavhandling, 2017

As urban planning evolves along with societal life, new requirements that should be met are rising. One of those, is the inclusion of acoustic quality in urban environments, before, during, or after a planning process. To address this for a yet to be built area, metrics are needed that try to quantify acoustic quality. Currently, due to limited models of auditory perception, it is essential that sound samples for the said environments are produced, in order to be used either for further perception based research, or to provide audio examples that can be judged on a case by case basis. The products from this procedure, auralisations, are usually realised with physically valid models, which simulate physical processes to create realistic sounds. While there are several methods aiming for this, fast and computationally expensive simulations of extended in area urban environments, is still a challenge. This report suggests a method for auralising background traffic noise, produced by cars travelling on roads distant from a listener, usually obscured by buildings or other barriers. To achieve computationally efficient auralisation of these scenarios, some parts of the technique are modelled using simulations of physical processes, while others are more simplified methods. For the former, ground reflection and air turbulence are modelled in detail. Individual pass-by events and the Doppler effect, are not modelled explicitly for each vehicle, but instead an approach that considers the traffic as cumulated noise is used. Based on the long distance limitation, shifts in the frequency domain, combined with modulation transfer functions and controlled coherence between the two channels of a listener, attempt to simulate the Doppler effect, fluctuations due to traffic flow inhomogeneities, and the spatial audio image. The modelling of the source output power of the vehicles, uses recorded data as a basis. For validation, auralisations are tested against mixed output from a previously developed and validated demonstrator.