MIT neuroscientists used a computational model of the auditory system to explain how the brain solves the 'cocktail party problem' — the ability to focus on one voice among many. The key mechanism is multiplicative gain: neurons tuned to features of a target voice (like pitch) amplify their activity, effectively boosting that voice above background noise. The model, trained to apply these gains based on a cue voice, reproduced a wide range of human attentional behaviors and errors. It also revealed that horizontal spatial separation between voices aids attention more than vertical separation — a finding later confirmed in human subjects. Researchers plan to use the model to simulate cochlear implant listening and potentially improve implant design.

7m read timeFrom news.mit.edu
Post cover image

Sort: