Early detection of cognitive disorders such as dementia on the basis of speech analysis - a cross-linguistic comparison of speech features

Alexandra König, Frank Rudzicz, Kathleen Fraser, Liam Kaufman, Jan Alexandersson, Nicklas Linz, Johannes Tröger, Maria Wolters, Francois Bremond, Philippe Robert

In: Proceedings of the Alzheimer's Association International Conference (AAIC). Alzheimer's Association International Conference (AAIC-17) July 16-20 London United Kingdom Alzheimer's Association 2017.


The people best placed to spot early cognitive decline are carers, social workers, and family. But there is a clear lack of affordable, usable screening apps that people without medical training can use to validate these concerns and to provide actionable data for medical professionals. The study aims to validate a new tool for fully-automated, reliable, unobtrusive, self-managed screening for cognitive decline, in particular dementia, and other cognitive disorders based on automatic speech analysis. It will allow earlier detection and, through that, more effective interventions resulting in the reduction of overall costs associated with treatment and reha- bilitation. For users it will offer the comfort of flexible usage without visiting professional physicians. Methods: At the moment there is an American English corpus of speech data used for training the algorithms of the system used by the tool for automatic detection of dementia in the USA and Canada. The main objective of the study will be the first experience in producing such corpus for another language, namely French. The corpus will contain ca. 250 samples of speech of patients with various levels of the syndrome, as well as other cognitive and behavioral disorders and ca. 50 samples of healthy people as a control group. All participants will be asked to perform a set of vocal tasks such as describing a series of images or perform verbal fluencies. Then, all speech samples will be tran- scribed and annotated by professional clinicians to make the corpus suitable for machine learning and identify which features transfer between languages. The processes, tools and experi- ences will be then presented in the blueprint for transferring the system into a new language. Results: First results of the analysis and cross linguistic comparison of the speech fea- tures will be presented at the conference. Conclusions: The proposed solution may supplement neuropsychological assess- ment with sophisticated and unobtrusive natural biomarkers extracted from speech data that can be collected outside of medi- cal consultations. It can provide rich information about cognitive and emotional characteristics and can be used to inform clinical judgment during consultations, saving time and money.

Deutsches Forschungszentrum für Künstliche Intelligenz
German Research Center for Artificial Intelligence