- Published on 24 October 2013
Richard Turner, an engineer with the Computational Perception Group in the Department of Engineering at the University of Cambridge in the UK, is working on technology that aims to identify corrupting ambient noise and to remove these sounds selectively using a statistical approach.
Turner explains how background sounds, known as audio textures, are made up of small single events that combine to produce the sound. These textures can be modeled mathematically and thus be distinguished from other sounds. Although no two rain sounds are identical, because the exact arrangement of the falling water drops is never repeated in the same way, there are statistical similarities in the sounds compared with other noises, such as wind or a crackling fire. Once modeled, the unwanted sounds can be removed by the system, providing the user directly with cleaned up speech audio. “Wouldn’t it be great if a hearing aid could detect environmental noises and automatically remove them? Well, the statistical description of sounds that we’re developing allows us to do just that,” he says.
The technology is in its early stages of development but future devices may function using different modes depending on the context, a car or train, or a restaurant environment, for instance. Further down the line, it may be possible to develop a more sophisticated device able to automatically select the most appropriate mode by analyzing the sound textures it receives.Source: Wired.co.uk/Science