Feasibility of motion sensor-based human activity recognition for supporting independence in smart homes

Area of Science:

• Geriatric care technology
• Human-computer interaction
• Machine learning for health

Background:

• Assessing independent living in older adults is crucial for timely interventions.
• Current methods may not fully capture real-world functional abilities.
• Technological advancements offer new avenues for continuous monitoring.

Purpose of the Study:

• To determine the feasibility of using Internet of Things (IoT) and artificial intelligence (AI) for recognizing key human activities.
• To evaluate the accuracy of these technologies in identifying activities essential for independent living in older adults.

Main Methods:

• Identified 17 clinically relevant activities for independent living assessment.
• Collected real-world data using 20 IoT wearable and object sensors in a home setting.
• Developed a random forest machine learning algorithm using extracted sensor data features.

Main Results:

• The AI algorithm achieved 87.5% average accuracy in recognizing 17 key activities.
• Achieved 97.95% accuracy in identifying four major functional areas: mobility, hygiene, nutrition/hydration, and medication intake.
• Study involved 10 participants performing activities in a home environment.

Conclusions:

• IoT and AI technologies show significant potential for accurately identifying activities critical for independent living.
• Reliable activity recognition enables effective monitoring of older adults' capabilities.
• This approach can support caregivers and clinicians in providing personalized and timely care.