ElderMate: A Privacy-Preserving Offline Edge-AI Assistive Healthcare Monitoring System for Elderly Aging-in-Place on Raspberry Pi 5
International Journal of Drug Delivery Technology · 01.06.2026 · 16 (42s)
DOI: 10.25258/ijddt.16.42s.103 ↗Kurz gefasst
Fachartikel in der Zeitschrift International Journal of Drug Delivery Technology, online erschienen am 01.06.2026. Verfasst von Sidhi Anish Kumar und 2 weiteren Autor:innen.
Abstract (Originalsprache)
Elderly individuals living alone face compounding risks from undetected falls, missed medication schedules, limited daily-care support, and delayed emergency escalation. Most home healthcare monitoring systems depend on cloud connectivity or commercial subscriptions, introducing privacy vulnerabilities and reliability failures during the safetycritical events they are designed to address. This paper presents ElderMate (Mango V2), a fully offline multimodal assistive healthcare monitoring system deployed end-to-end on a Raspberry Pi 5 (8GB RAM, ARM Cortex-A76, no accelerator) at approximately $100 hardware cost, with no cloud services or vendor APIs in the request path. The runtime employs a five-node multiprocessing architecture interconnected over ZeroMQ with static CPU-affinity partitioning. The vision subsystem performs fall detection through a zero-training rule-based finite-state machine over MediaPipe Pose landmarks, achieving 96.67% sensitivity, 90.91% specificity, and 95.12% accuracy across 41 UR Fall Detection Dataset sequences at 12.10±0.23FPS within 296±3MB RSS. The audio pipeline pairs offline Vosk speech-to-text (436.8±77.3ms) with Piper neural TTS, fronted by a deterministic two-layer intent router resolving 19 eldercare tool categories at 96.92% accuracy and 0.102ms mean latency over 3,250 trials. A QLoRA-fine-tuned Qwen 2.5 1.5B 4-bit GGUF model sustains 11.33±0.13tokens/s with 96ms warm-cache time-to-first token on two CPU cores. Whole-system evaluation across 13 scenarios yielded 92.31% response success and 1.8ms median fast-path latency, with all subsystems concurrent within the thermal envelope (peak 59◦ C; throttle threshold 80◦ C). From a healthcare-technology perspective, ElderMate is positioned as an assistive home-care prototype for aging-in-place, where privacy-preserving local intelligence, fall-risk monitoring, daily-care reminders, accessible speech interaction for mobility or vision-impaired elderly users, and emergency caregiver notification via GSM are essential. The system is an engineering prototype; it does not replace professional caregivers or clinical monitoring. ADL-specificity expansion, durability characterisation, and supervised elderly user trials remain necessary before deployment validation.
Zitieren
APA
Kumar, S. A., Pabboju, S., & Kiranmaie, P. (2026). ElderMate: A Privacy-Preserving Offline Edge-AI Assistive Healthcare Monitoring System for Elderly Aging-in-Place on Raspberry Pi 5. International Journal of Drug Delivery Technology, 16(42s). https://doi.org/10.25258/ijddt.16.42s.103
Vancouver
Kumar SA, Pabboju S, Kiranmaie P. ElderMate: A Privacy-Preserving Offline Edge-AI Assistive Healthcare Monitoring System for Elderly Aging-in-Place on Raspberry Pi 5. International Journal of Drug Delivery Technology. 2026;16(42s). doi:10.25258/ijddt.16.42s.103
BibTeX
@article{Kumar2026,
author = {Kumar, Sidhi Anish and Pabboju, Suresh and Kiranmaie, P.},
title = {ElderMate: A Privacy-Preserving Offline Edge-AI Assistive Healthcare Monitoring System for Elderly Aging-in-Place on Raspberry Pi 5},
journal = {International Journal of Drug Delivery Technology},
year = {2026},
volume = {16},
number = {42s},
doi = {10.25258/ijddt.16.42s.103}
}
EUVIMED