The Perception and Cognition of Sound Texture

The environments in which we live are filled with noise. Noise makes speech and other sounds of interest harder to hear, but also conveys information about the world around us, telling us how hard it is raining, whether the room next to us is full of people, or how close a kettle may be to boiling. Humans excel both at separating other sounds from background noise and at recognizing the causes of the noise in the world. Understanding these abilities could enable more effective hearing aids, cochlear implants, and speech-based machine systems such as those in our phones. Here the investigators propose to study human abilities to hear in noisy conditions. The proposed work will jointly pursue three goals. First, the investigators will develop a computational framework to understand the properties of noise that underlie human perception. This will involve assembling a large data set of environmental noises and measuring their statistical properties. These properties will then be related to human perception by synthesizing sounds that have the same properties and asking whether they are realistic and recognizable to humans. Second, the investigators will measure the ability of human listeners to perceptually separate noise from other sounds. They will then build models that can predict the conditions in which such separation should succeed or fail. Third, the investigators will measure the ability of human listeners to remember different noises and develop a model to explain why some sounds are better remembered than others. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.