Improving the Intelligibility of Underground Stations using Horizontal Line Arrays
An Audio-Based Vehicle Classifier Using Convolutional Neural Network
Using Non-wave-based Modelling to Explore How much Acoustic Diffusion is too much
Presentations by Diego Cordes, Ekim Bakirci and Michael Fort.
The students on the London South Bank University Masters course in Environmental and Architectural Acoustics each must undertake a dissertation. Presented are three of the best projects.
The first dissertation investigates the how to improve speech intelligibility on London Underground by applying modern technology in the form of Line Arrays. There has been a struggle to achieve the minimal life safety intelligibility requirements on station platforms making millions of daily passengers’ life difficult. Through computer simulations the PA/VA system was changed to a planar source to determine how much this could improve the intelligibility of the system. By creating a homogeneous sound distribution close to the public, with minimal room excitation thus minimising reverberations destructive interference to intelligibility could be mitigated resulting in significantly improvements to communication.
The second presentation focuses on audio-based event and scene classification. Many examples of environmental noise detection, vehicle classification, and soundscape analysis are developed using state of art deep learning techniques. The major noise source in urban and rural areas is road traffic noise. To develop an audio-based vehicle classifier a convolutional neural network-based algorithm was proposed using Mel spectrogram of audio signals as an input feature. Different variations of the network were generated by changing the parameters of the convolutional layers and the length of the network. filter size, number of filters were tested with a dataset prepared with various real-life traffic records and audio extracts from traffic videos. The precision of the networks was evaluated with common performance metrics. Further assessments were conducted with longer audio files and predictions of the system compared with actual traffic flow.
The final project explores the relationship between acoustic surface diffusion and acoustical character in concert halls by using non-waved based modelling. Measurements were undertaken in a small concert hall to use as a baseline, which is then transferred into ODEON to experiment with the acoustic parameters further. The relationship between amount of diffusion vs scattering coefficient to achieve optimal objective parameters in a concert hall were generated. These trends were then tested on existing famous concert hall models, e.g. Musikvereinssaal. The experiments showed that when considering how much diffusion is too much in a concert hall, the statistical amount of diffusion is less of a factor, when compared to retaining strong early reflections from side walls and ceilings and the amount of scattering provided by diffuse surfaces.
All three projects and many more will be presented by London South Bank University students in the upcoming Inter-Noise conference to be held in Glasgow, August 2022.
For more details of IOA regional branch activities please refer to the IOA website: www.ioa.org.uk
Topics and speakers for the evening meetings are generally identified and organised by the London Branch Committee, but we always welcome new ideas and suggestions for future presentations. If you have any ideas or suggestions, or may even like to give a presentation yourself, please contact the London Branch Committee Roslyn Andrews. Roslyn.Andrews@aecom.com
This meeting is open to Members and Non Members
The presentations will be recorded if the speakers give permission.
The Zoom details have already been emailed to London Branch members but can be requested by email at firstname.lastname@example.org