Multi-Task Estimation of Age and Cognitive Decline from Speech

Abstract

Speech is a common physiological signal that can be affected by both ageing and cognitive decline. Often the effect can be confounding, as would be the case for people at, e.g., very early stages of cognitive decline due to dementia. Despite this, the automatic predictions of age and cognitive decline based on cues found in the speech signal are generally treated as two separate tasks. In this paper, multi-task learning is applied for the joint estimation of age and the Mini-Mental Status Evaluation criteria (MMSE) commonly used to assess cognitive decline. To explore the relationship between age and MMSE, two neural network architectures are evaluated: a SincNet-based end-to-end architecture, and a system comprising of a feature extractor followed by a shallow neural network. Both are trained with single-task or multi-task targets. To compare, an SVM-based regressor is trained in a single-task setup. i-vector, xvector and ComParE features are explored. Results are obtained on systems trained on the DementiaBank dataset and tested on an inhouse dataset as well as the ADReSS dataset. The results show that both the age and MMSE estimation is improved by applying multitask learning, with state-of-the-art results achieved on the ADReSS dataset acoustic-only task.

Explainable Identification of Dementia From Transcripts Using Transformer Networks

“…Specifically, after proposing three deep learning architectures based on CNNs and RNNs, the authors applied visualization techniques and showed which linguistic characteristics are indicative of dementia, i.e., short answers, repeated requests for clarification, and interjections at the start of each utterance. Authors in [25] proposed a multi-task learning framework (Sinc-CLA), so as to predict age and MMSE scores (both considered as regression tasks) and used only speech as input for their proposed network. Concurrently, they introduced shallow networks with input i-vectors and x-vectors both in single and multi-task learning frameworks.…”

Harnessing Speech-Derived Digital Biomarkers to Detect and Quantify Cognitive Decline Severity in Older Adults

“…Language functioning can provide valuable insights into cognition and behavior, serving as a window into an individual's cognitive functioning [7][8][9][10][11][12]. Speech and language offer an opportunity to use ecologically valid methods to assess key symptoms related to cognitive function, a unique advantage in the assessment and monitoring of cognitive decline.…”

“…The growing prevalence of cognitive impairments, like dementia, coupled with the current limitations in diagnostic approaches, underscores the urgent need for a robust Remote Patient Monitoring (RPM) system. Such a system would be instrumental in the early identification of cognitive decline, enabling ongoing monitoring of its progression [11]. In this context, the development of a wearable device, functioning as an RPM tool, could significantly bridge this gap.…”

Towards Speaker Age Estimation with Label Distribution Learning

“…Recently, there is a line of work that utilize multi-task learning to estimate age and another speaker attribute, such as gender [13], and emotion [14]. [15] applied a multi-task learning for the joint estimation of age and the Mini-Mental Status Evaluation criteria, showing improved performance on input features including i-vector and x-vector.…”