Applying Machine Learning Technologies Based on Historical Activity Features for Multi-Resident Activity Recognition.

Area of Science:

• * Gerontechnology and Ambient Assisted Living (AAL)
• * Human-Computer Interaction (HCI)
• * Machine Learning and Artificial Intelligence (AI)

Background:

• * Growing aging population necessitates advanced home care solutions.
• * Existing activity recognition systems often fail in multi-resident environments, hindering personalized care.
• * Accurate identification of individual activities is crucial for ensuring safety and well-being of the elderly.

Purpose of the Study:

• * To develop a robust method for recognizing home activities in multi-resident environments.
• * To accurately distinguish the association between residents and their specific activities.
• * To enhance the safety and personalized care for the elderly living with others.

Main Methods:

• * Utilizing historical activity characteristics: interaction, frequency, duration, and behavioral patterns.
• * Applying a suite of machine learning models, including five supervised learning and two deep learning methods.
• * Training and testing models on real-world datasets to evaluate performance.

Main Results:

• * Proposed methods achieved higher precision, recall, and accuracy compared to traditional approaches.
• * Reduced training time was observed with the implemented machine learning techniques.
• * Achieved a maximum accuracy of 91% with J48 Decision Tree (DT) and 95% with Long Short-Term Memory (LSTM) networks.

Conclusions:

• * The developed approach effectively recognizes individual activities in multi-resident homes.
• * Machine learning, particularly deep learning models like LSTM, shows significant promise for advanced elderly care.
• * This research paves the way for more sophisticated and personalized home monitoring systems.