It is the outcome of a three-and-a -half year project ‘Effects of Reverberation on Conversation in Rooms’ due to run until July with funding of just under £350,000 from the Engineering and Physical Sciences Research Council.
Right: The audability map: Courtesy University of Cardiff & EPSRC

The software generates maps showing hotspots where conversations would not be intelligible if the room were busy. Architects can then adjust designs to reduce reverberations, until the hotspots are eliminated and audibility is maximised.
 
Software already exists to help architects predict how a building will perform acoustically for audiences in theatres and concert halls. This software however is specifically designed to improve acoustic design of indoor spaces where large numbers of people meet, chat and interact. It could be used for business as well as social purposes, for example, in designing open-plan offices, cafes and reception areas.
 
“A lot of work has been done to understand acoustics in places used for public performances,” says Professor John Culling (right), project leader. “But little has been done to improve the acoustics of day-to-day meeting-places, even though this would help all of us in our working and social lives.” His podcast explains the development.

The new listening software produces results much more rapidly than other acoustic software. The key to its capabilities is the unprecedented sophistication and computational efficiency of the unique mathematical equation that underpins it.

"It is currently nameless but I'm thinking on that one," he says. "At the moment, it is written in Matlab only, but it could be added to any software platform currently able to generate binaural room impulse responses for the purpose of auralisation, so it could be added to Odeon or Ectoect (now Autodesk).

"But we are working with an architectural acoustician who is building it into a standalone rather than a plugin application. The maps we have so far generated use a separate program in C, to make impulse responses. In this standalone version it will work from a standardised output format for architectural software,which basically just describes the room shape and surface materials.

"We reckon this could be ready in a couple of months. The application will do all the acoustics and the psychoacoustic modelling in a single package. Most importantly for us, we will escape our current limitation of having rectagular rooms all the time!"
 
That equation has been built up using the project team’s  research into how people take in sound through both ears, and as it travels round busy rooms, and how noise sources affect each other, in order to accurately predict acoustic quality at every point of an indoor space where people are likely to gather and talk.
 
The architect will be able to call their proposed design onto their computer screen and run the software, which will ask them to specify the locations of the main sound sources in the room. An audibility map will then automatically be produced and the architect will be able to change the room’s dimensions, its shape and/or the materials to be used, until hotspots are eliminated.

This means that rooms could be tailor-made to suit their purpose. The work will also make a significant difference to areas where audibility is important, such as rail and airport announcement waiting areas. In emergency situations such clarity could be vital in saving lives.  The research will also help in the future development of hearing aids and cochlear implants.
 
“Our objective now is to identify and work with a software company to help us develop the software further and market it,” says Professor Culling. “Hopefully it will be available for architects to use within the next 12 months.” 

Future layered effect could be any colour
"The current limitation is that only the effect of continuous noise interference can be
correctly predicted. It is not ideal at assessing the acoustics of rooms where there are likely to be a small number of interfering voices" (left))

"That's a tough problem, though, " says Professor Cullen. "There are multiple effects of speech interference, some of which help the listener (it's periodicity, it's modulation) and some of which are disruptive (the linguistic content - apparently at multiple levels). We will have to deal with them one at a time.

"I should emphasise, though, that all these additional effects are layered on top of the ones that we have addressed so far (better-ear listening, binaural unmasking, reverberation multiple interferers and room colouration), so the predictions do have some relevance to speech interferers - I anticipate that the inaccuracies will start to arise when you have one or two interfering voices that are quite close.

"A case that we are planning to examine at the moment is the restaurant situation, where you have evenly distributed tables thoughout a room, each with one person speaking at any one time (certainly if they are tables for two, anyway!).

This is the kind of situation we have always had in mind, where people want to communicate but there are multiple parallel conversations causing interference. It will be interesting to see how accurate the model is in such a case - are all these reverberant voices effectively like continuous noise, or are the speech specific mechanisms relevant?

"There is no experimental literature on such a situation."