ALISS: a context-aware intelligent assistance system for accessible smart living

Abstract

Age-related memory decline, encompassing conditions such as dementia and Alzheimer’s disease, make significant challenges to older living. A persistent issue that these individuals encounter daily is the inability to locate misplaced household items. This paper introduces A Context-Aware Intelligent Assistance System for Accessible Smart Living (ALISS), a smart assistive technology designed to facilitate object retrieval within domestic environments for those experiencing memory impairments. ALISS includes advanced human and object tracking and reasoning capabilities powered by LLMs, all grounded in a structured. To ensure effectiveness in varied household contexts, ALISS employs a dynamic environment update mechanism that integrates hybrid person re-identification techniques and threshold-based object tracking. This approach ensures long-term consistency, particularly in multi-user scenarios. Through the experimental results indicate that ALISS enhances contextual captioning quality to, minimizes hallucination occurrences around 0.11 compared with single-model baselines. This work also significantly improves the identity and object tracking of dynamic environment updates accuracy, exceeding 82.8% by applied hybrid method. Furthermore, the system demonstrates high accuracy and safety in responses to object-related queries through history-aware RAG. These findings position ALISS as a promising advance in the development of practical, personalized memory assistance systems aimed at improving autonomy and quality of life for older adults in smart living environments.