Elder Depression Detection by Multimodal Means

Abstract

Depression in the elder group is a widespread but silent issue. This article presents a solution for detecting depression in the elderly group via multi-modal approaches. The conceptual framework for the solution is demonstrated by the Multimodal Transformer Elder Depression Detection model: the model processes the input with multiple forms of data, such as text, audio, and video, and gathers information for data fusion. After data fusion, the model can produce a score to reflect the mental state of the patient. In the model, DAIC and extended DAIC databases are used for training and testing. Also, we have established additional testing based on the information gathered from the elder members from local hospital. The testing results demonstrate that the model can successfully detect depression in the elderly group.

Recommended citation: Tianruo Xu. Elder Depression Detection by Multimodal Means. BCP Education & Psychology, November 7, 2022.