AI-Powered Speech Clock Links Voice Patterns to Brain Aging and Dementia
Researchers created an AI model that estimates a person’s “speech age” from voice recordings, finding larger gaps between estimated and actual age in individuals with brain disorders.
The study combined voice recordings from Argentina, Chile, Colombia, Mexico and Peru with a machine-learning model that extracts more than 700 speech characteristics, such as tempo, pauses, pitch and word choice. Led by Agustín Ibáñez of the Global Brain Health Institute and Trinity College Dublin and Adolfo M. García of the Universidad de San Andrés, the researchers calculated a “speech-age gap” by comparing the model’s one outlet estimate with the participant’s actual age.
On average, the model’s predictions were off by roughly nine years. Participants with diagnosed brain diseases, including Alzheimer’s and frontotemporal dementia, showed a larger gap, being rated older than healthy individuals. The gap also correlated with cognitive test scores, brain-age biomarkers and a blood marker for tau protein. While the authors caution that the findings cannot yet be applied to other languages or individual diagnosis, they suggest that speech analysis could become a low-cost, non-invasive tool for monitoring brain health, especially in regions lacking advanced medical infrastructure.
Why it matters
The research suggests voice analysis could offer an inexpensive way to track brain aging and detect dementia risk.
In this story
