[REVISION HISTORY]

Machine learning for age estimation via voice analysis

Updated 1 time since CLSTR started tracking revisions of this situation.

What changed

2026-09-30 21:05 UTC → 2026-10-01 22:41 UTC · added ~~removed~~

Researchers are developing machine learning models to estimate age through vocal analysis. An initial study published in npj Aging introduced the ‘Voice Age’ model, which uses 30 seconds of recorded speech to estimate chronological age. Testing on Hebrew-speaking adults showed an average deviation of approximately four years for women and 4.4 years for men, utilizing physiological changes in the respiratory and nervous systems as indicators. Subsequent research published in Science Advances expanded this concept into a ‘speech clock’ that analyzes acoustic and linguistic features, such as pitch, vocabulary, and speech rate. This model identifies a ‘speech age gap’ between a person’s actual age and their predicted speech age. In a study of Spanish-speaking participants across several South and Central American countries, researchers found that a speech age older than chronological age correlated with accelerated biological aging and cognitive decline. These discrepancies were notably significant in individuals with Alzheimer’s disease and various forms of dementia. While offering a non-invasive alternative to expensive medical scans, researchers note the tool requires further validation across different cultures and is not yet a formal diagnostic instrument. Building on this, researchers at Trinity College Dublin utilized data from 2,928 Spanish-speaking participants across Argentina, Chile, Colombia, Mexico, and Peru to refine the ‘speech clock’. The model analyzes hundreds of features, including pauses, semantic precision, and emotional expression. The study found that participants with a speech age older than their chronological age showed higher associations with Alzheimer’s disease, frontotemporal dementia, and mild cognitive impairment. Additionally, the findings linked speech aging to social adversity, such as economic hardship and limited healthcare access, suggesting the tool could serve as an accessible method for monitoring cognitive health in low-resource regions.

Versions

Only revisions since CLSTR began indexing content versions appear here. Select a version to see what changed compared to the one before it.