Related Experiment Video
Updated: Mar 30, 2026
12:43
A Protocol for Comprehensive Assessment of Bulbar Dysfunction in Amyotrophic Lateral Sclerosis ALS
Published on: February 21, 2011
36.2K
Automatic Prosodic Analysis to Identify Mild Dementia.
Eduardo Gonzalez-Moreira1, Diana Torres-Boza1, Héctor Arturo Kairuz1
1Center for Studies on Electronics and Information Technologies, Universidad Central "Marta Abreu" de Las Villas, 54830 Santa Clara, Cuba.
Biomed Research International
|November 12, 2015
Summary
This study introduces a novel speech analysis technique to detect mild dementia. The method accurately identifies dementia features in elderly adults by analyzing speech prosody, achieving 85% classification accuracy.
Area of Science:
- Neurology
- Computational Linguistics
- Gerontology
Background:
- Mild dementia diagnosis presents challenges, often relying on subjective assessments.
- Objective, quantifiable biomarkers for early dementia detection are needed.
- Speech deficits are increasingly recognized as potential indicators of cognitive decline.
Purpose of the Study:
- To develop and evaluate an exploratory technique for identifying mild dementia using speech analysis.
- To assess the efficacy of computational prosodic analysis in distinguishing between individuals with mild dementia and healthy controls.
- To identify key prosodic features indicative of early-stage dementia.
Main Methods:
- Recorded audio sessions of twenty participants (10 mild dementia, 10 healthy controls) during a reading task.
- Employed automatic prosodic analysis to measure twelve distinct prosodic features from speech samples.
- Utilized a novel computational method for feature extraction and classification.
Main Results:
- The study successfully measured and analyzed prosodic features in both groups.
- The best classification model achieved 85% accuracy in identifying mild dementia.
- Four specific prosodic features were found to be most effective for classification.
Conclusions:
- Computational speech analysis, focusing on prosodic features, offers a viable tool for the automatic identification of dementia.
- This technique provides a promising, objective, and non-invasive method for early dementia detection in elderly adults.
- Further research can refine this approach for broader clinical application.
