Brain wave hearing system helps users focus on one voice in crowds

A brain-wave hearing system helped listeners pick out one voice from overlapping conversations by reading attention signals in the auditory cortex and turning them into real-time changes in volume. The Columbia University’s Zuckerman Institute team said the approach is the first direct human evidence that brain-controlled hearing technology can help people single out a voice in a crowd, a practical advance aimed at the stubborn “cocktail party problem.”

The study, published online May 11, 2026, in Nature Neuroscience under the title “Real-time brain-controlled selective hearing enhances speech perception in multi-talker environments,” used electrodes already implanted in epilepsy patients undergoing brain surgery. Researchers measured activity while four people with typical hearing listened to two simultaneous conversations, then decoded which speaker each person was attending to and adjusted the audio on the fly. That matters because the system did more than amplify sound: it attempted to separate the voice a listener wanted from the one he or she wanted to ignore.

Nima Mesgarani, a Columbia scientist who has spent more than a decade on the problem, said the work moves beyond conventional hearing aids, which can raise speech and reduce some background noise but cannot selectively enhance one specific voice. He described the system as a “neural extension” of the listener, using the brain’s own attention signals to steer what gets heard. First author Vishal Choudhari said the central question was whether brain-controlled hearing could move beyond incremental advances and become a prototype that helps someone hear better in real time.

The research built on a 2012 finding by Mesgarani and UCSF neurosurgeon Eddie Chang showing that the auditory cortex tracks the speaker a person is attending to rather than the one being ignored. That earlier insight helped establish that selective listening is encoded in the brain; the new study tried to use that signal as a control source for assistive technology.

The potential market is enormous. The World Health Organization says more than 430 million people worldwide live with disabling hearing loss, and it estimates that number could top 700 million by 2050. But the system still faces a major hurdle: it was tested only in four people with typical hearing, not in the population it is meant to serve. MIT’s Josh McDermott said it remains an open question whether the approach will work as well for people with hearing loss.