Acoustic, spatial and informational cues used to solve the cocktail party problem
This project examines the many acoustic cues and cognitive processes underlying our ability to understand what someone is saying in the noisy, everyday listening environment.
Speech is the principal mode of human communication and yet we understand little about how we are able to concentrate on one talker against a noisy background. Sorting out the talker of interest requires sophisticated signal processing by the central nervous system and relies on bottom-up (acoustic cues) and top-down (cognitive) processing. Current auditory models are poor predictors of listening performance in such real-life environments. This research explores the acoustic, spatial and informational cues used to segregate concurrent talkers in real-world listening environments. These experiments exploit a new masking paradigm, independent of linguistic or semantic context, and the results will provide deep insights into the basic encoding of human speech. Significant improvements in supporting listening by the hearing impaired, in teleconferencing systems and computer speech interfaces are dependent on a more advanced understanding of how we solve this cocktail party problem. There is significant potential for commercialization of the IP that will arise from this research.
Want to find out more?
Hearing impairment, Spatial hearing, cocktail party problem, auditory scene analysis, auditory localisation, neural encoding of speech, Hearing & vision problems, Neuroscience & psychology, The senses
The opportunity ID for this research opportunity is: 6