Enhancing Quality of Life for Dementia Patients through Intelligent Memory Aids in Social Communication

Abstract

The prevalence of dementia in South Korea has shown a marked increase, with the number of affected individuals rising from approximately 630,000 in 2015 to nearly 1 million in 2023. This growth aligns with the aging population, as the proportion of elderly individuals has expanded, driving the prevalence rate from 9.5% in 2015 to 10.3% in 2023. This trend is not isolated to South Korea; in the United States, the number of dementia patients has also risen sharply, from around 4.7 million in 2010 to an estimated 13.8 million by 2050, representing a nearly threefold increase over 40 years. Globally, the number of dementia patients is projected to reach 131.5 million by 2050, nearly doubling every 20 years, highlighting the significant impact of population aging as a major risk factor for dementia onset. Memory aids like notes and reminders are commonly used by dementia patients to support daily activities, but they have limitations. While helpful for routine tasks, these simple tools cannot assist patients in remembering faces or recognizing familiar people. To address this issue, we introduce a machine learning-based memory aid system designed to support social communication for dementia patients. The proposed system processes face images to perform facial identification which provides relevant information about the identified person, such as their relationship to the dementia patient and any scheduled activities with them. To improve the system's accuracy, we proposed a metric-based loss function. Experimental results demonstrated that this approach enhanced accuracy, achieving a 2.96% improvement over the previous method.