Detection of abnormal behaviour for dementia sufferers using Convolutional Neural Networks.

Area of Science:

• Gerontology
• Artificial Intelligence
• Cognitive Science

Background:

• The global elderly population is rapidly increasing.
• Cognitive decline, a significant consequence of aging, leads to dementia and impacts daily living.
• Early detection of dementia is crucial for timely intervention and support.

Purpose of the Study:

• To investigate activity recognition and abnormal behavior detection in elderly individuals with dementia.
• To develop a methodology for generating synthetic data for dementia research.
• To evaluate the efficacy of Convolutional Neural Networks (CNNs) for identifying dementia-related behavioral patterns.

Main Methods:

• A synthetic data generation methodology was created to address the scarcity of real-world data.
• Convolutional Neural Networks (CNNs) were employed to model activity sequences and detect anomalies.
• Activity recognition was framed as a sequence labeling task, with abnormal behavior identified by deviations from normal patterns.

Main Results:

• CNNs demonstrated effectiveness in modeling activity patterns and detecting abnormal behaviors associated with dementia.
• The performance of CNNs was benchmarked against established methods like Naïve Bayes (NB), Hidden Markov Models (HMMs), Hidden Semi-Markov Models (HSMM), and Conditional Random Fields (CRFs).
• CNNs proved to be competitive with state-of-the-art methods in dementia-related behavior analysis.

Conclusions:

• The proposed CNN-based approach offers a promising solution for early dementia detection through activity recognition.
• Synthetic data generation is a viable strategy for overcoming data limitations in dementia research.
• This work contributes to developing assistive technologies for elderly care and dementia management.