Enterprise AI Research Analysis
Towards AI-Enabled Innovative Models in Elderly Care and Domestic Services
Authored by Yujie Jiang and Jing Zhang, published at ICAISD 2025 (November 2025)
Executive Impact Summary
This research introduces an AI-driven "Smart Brooch" model to transform elderly care and domestic services, addressing challenges from an aging population in China. The model leverages large language models and multimodal interaction for service standardization, safety, psychological support, and personalized care, establishing a "perception-decision-execution" framework to empower the workforce and drive digital transformation.
Deep Analysis & Enterprise Applications
Select a topic to dive deeper, then explore the specific findings from the research, rebuilt as interactive, enterprise-focused modules.
Enterprise Process Flow: AI Agent "Perception-Decision-Execution"
The AI-enabled agent is built upon a closed-loop "perception-decision-execution" framework, integrating smart hardware for data capture, a large language model for intelligent processing, and human-AI collaboration for service delivery and continuous improvement. This architecture ensures real-time responsiveness and adaptability across diverse elderly care scenarios.
Smart Brooch: Multimodal Data Capture
The "Smart Brooch" is a compact, wearable device integrating a 1080P wide-angle camera, noise-reducing microphone array, and speaker. It utilizes the YOLOv8 object detection algorithm for real-time visual perception and Voice Activity Detection (VAD) for robust auditory perception in noisy home environments, ensuring clean and relevant data input for the AI system.
LLM-Driven Intelligent Decision Core
The core of the system is a multimodal Large Language Model (LLM) specifically fine-tuned for elderly care and domestic service. It is enhanced by a dedicated knowledge base (service standards, operational protocols) and Retrieval-Augmented Generation (RAG). This enables cross-modal semantic understanding, rule-based reasoning, and reinforcement learning for dynamic decision optimization, adapting to real-time feedback for personalized and effective care.
Automated Service Standardization & Quality
This module automates the quantitative assessment of service quality using computer vision and speech analysis. It processes scene images via U-Net segmentation to detect stain residue, calculates stain area ratio, and analyzes speech data via NLP for communication dynamics, enabling objective worker evaluation against built-in service standards.
| Method | Service Quality Score |
|---|---|
| Proposed AI Model | 92.4 |
| Rule-based System | 78.5 |
| Manual Recording | 65.0 |
Enhanced Safety Assurance & Incident Response
A multi-layered data privacy framework ensures "minimum necessary data" collection and encrypted HD recording during emergencies. A multimodal fusion model identifies safety threats (physical conflict, sensitive speech) autonomously. When a threat is detected or a one-click call is activated, the system triggers alarms, pushing encrypted live video, audio, and location data to relevant parties for immediate intervention.
| Method | Incident Response Time (s) |
|---|---|
| Proposed AI Model | 1.2 |
| Rule-based System | 3.5 |
| Manual Recording | >10 |
Case Study: Enhancing Elderly Well-being with Empathetic AI
By leveraging LLMs and affective computing, the Smart Brooch creates dynamic service profiles for each senior, learning dietary preferences, medication schedules, and item locations. It identifies emotional states like loneliness and anxiety from behavioral data, guiding caregivers in empathetic communication and personalized interaction to alleviate emotional distress and enrich spiritual life.
Personalized Service & Psychological Companionship
The system builds a dynamic service profile for each senior by continuously learning their habits. Based on this profile and real-time context, it generates customized service plans and provides operational guidance to workers, reducing the need for repetitive instructions. Affective computing models analyze behavioral data to identify emotional states, facilitating empathetic communication and emotional support to alleviate loneliness and anxiety.
Calculate Your Potential AI Impact
Estimate the efficiency gains and cost savings AI could bring to your organization based on our research findings.
Projected Annual Savings & Efficiency
Your AI Transformation Roadmap
A typical phased approach to integrate AI solutions into your enterprise, based on best practices and similar deployments.
Phase 1: Smart Brooch Deployment & Data Collection
Deployment of Smart Brooch devices and establishment of secure data pipelines for multimodal data (video, audio, sensor) collection. Focus on infrastructure setup and initial data anonymization protocols.
Phase 2: LLM Fine-tuning & Knowledge Base Integration
Customization and fine-tuning of large language models for specific elderly care scenarios. Integration of enterprise-specific knowledge bases including service standards and operational protocols.
Phase 3: Pilot Program & Human-AI Workflow Training
Conduct a pilot program in a controlled environment. Train caregivers on human-AI collaborative workflows, ensuring smooth adoption and feedback collection for iterative system improvements.
Phase 4: Scaled Rollout & Continuous Optimization
Gradual expansion of the AI-enabled system across broader operations. Establish continuous monitoring, performance evaluation, and an agile feedback loop for ongoing model updates and feature enhancements.
Ready to Transform Your Enterprise with AI?
Connect with our AI specialists to explore how these innovative models can be tailored to your specific business needs and drive measurable impact.