Role of Voice Signal Analysis in Screening for Cognitive Impairment: A Study on Community-dwelling Older Adults

This study investigated the potential of voice signal analysis for the early detection of cognitive impairment in community-dwelling older adults. The “Voice Sensor” technology, integrated into the “Malg-eun Village” platform, was evaluated by comparing its predictions with traditional cognitive assessments, including the Korean Mini-Mental State Examination-2. Between January 22, 2024, and May 31, 2024, voice data were collected from 200 participants in Jecheon City, comprising 100 individuals with cognitive impairment and 100 cognitively normal individuals. Voice recordings were gathered through bi-weekly caregiver calls and weekly AI-driven calls. The cognitive status was assessed based on the cumulative voice scores. Statistical analysis included sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV). A t-test was conducted to compare the voice scores between groups, and correlation analysis validated the relationship between the voice sensor results and cognitive assessment scores. The voice sensor showed moderate sensitivity (72.41%) and specificity (63.46%) in detecting cognitive impairment. In particular, it showed a high NPV (89.19%), highlighting its ability to rule out cognitive decline. Nevertheless, its PPV (35.59%) and specificity reveal areas for refinement. These findings suggest that voice signal analysis is a promising, non-invasive, and scalable tool for early cognitive screening in community settings.