Particle.news
Download on the App Store
5 ARTICLES
·
4w ago

Columbia Team Demonstrates Brain‑Controlled Hearing That Isolates a Chosen Voice

The approach targets a long‑standing weakness in hearing aids in crowded spaces.

After more than a decade of research, scientists evaluate implanted brain-decoding technology that helps people hear one voice among many. (Credit: Matteo Farinella/Columbia's Zuckerman Institute)
By monitoring the synchronization of brain waves with the rhythms of specific voices, this real-time system acts as a neural extension that amplifies a listener's intended conversation while silencing competing background noise. Credit: Neuroscience News
Participants with intracranial electrodes listened to two competing, spatially separated conversations. Their neural signals were recorded and fed into a real-time processing system. The system uses a linear regression model to reconstruct the temporal envelope of the attended speech from low-frequency (LF) and high-gamma (HF) neural features. The reconstructed envelope is then compared to the envelopes of the two conversations to determine the listener’s focus, which in turn drives the selective amplification of the attended speaker. Image credit: Choudhari et al., doi: 10.1038/s41593-026-02281-5.

Overview

  • The Nature Neuroscience paper, which went online Monday, reports the first human test of a system that uses brain signals to turn up the speaker a listener is focusing on.
  • The device reads neural activity and reconstructs the rhythm of the attended speech, then matches it to the competing audio streams to pick the right talker and boost that voice in real time.
  • Epilepsy patients with existing intracranial electrodes listened to two overlapping conversations as the decoder identified their chosen speaker and adjusted the volumes automatically.
  • Listeners understood the target speech more clearly and with less effort, and some described the experience as startling because the volume seemed to shift with pure attention.
  • Researchers call the study a proof of concept and now aim to move from invasive brain recordings to minimally invasive or wearable sensors that work in everyday noisy settings.