Automatic Speech Analysis for the Detection of Emotional Disturbances in Persons with Dementia

Publication

Alexandra König, Nicklas Linz, Johannes Tröger, Aurore Rainouard, Auriane Gros, Jan Alexandersson, Philippe Robert

In: Proceedings of the Alzheimer's Association International Conference (AAIC). Alzheimer's Association International Conference (AAIC-2018) July 21-25 Chicago Illinois United States Alzheimer's Association 2018.

Abstract

Emotional disturbances found in dementia patients such as depression, anxiety or apathy have an important impact on quality of life of both patients and their caregivers and represent a strong predictor for illness progression. However, they are often underdiagnosed since current assessment tools rely primarily only on subjective measures. New computational approaches may allow a more objective evaluation of these behaviors that humans struggle to quantify. Therefore the study aims to investigate whether automated analysis of linguistic and paralinguistic biomarkers derived from audio recordings could be useful to determine emotional disturbances in dementia patients and thus demonstrate potential to improve early diagnosis. Methods: 150 participants including both dementia patients and healthy control subjects were recorded while answering several mood--related open questions (i.e. ‘can you tell me something about a pleasant event coming to your mind ?’) along with a cognitive assessment Apathy Inventory. Speech signal processing techniques were applied to extract features to compare to these baseline scores. Results: Vocal features such as speech rate or sound duration correlate significantly with severity levels of certain emotional disturbances like apathy. Classification between pathological and non-pathological groups based on the extracted features obtain 88% accuracy. Voice--based emotion detection has many potential applications in healthcare for screening and diagnosis, even remotely, but also in helping identify new behavioral markers of emotional disturbances, as well as to measure intervention response, or test clinical theories about underlying mechanism.

Automatic Speech Analysis for the Detection of Emotional Disturbances in Persons with Dementia

Abstract

Follow us on:

Editor

Newsletter

Locations