Modelling of distributed activity recognition in the home environment

Area of Science:

• Computer Science
• Artificial Intelligence
• Ubiquitous Computing

Background:

• Distributed sensor systems, including ambient and on-body sensors, are effective for recognizing complex human activities in daily life.
• Distributed activity recognition systems offer advantages over centralized solutions in managing processing and communication loads.
• A significant challenge lies in developing distributed systems that efficiently utilize available resources, ensuring scalability and adaptability.

Purpose of the Study:

• To present a novel approach for distributed activity recognition using an activity-event-detector (AED) concept.
• To demonstrate the formal construction and application of AEDs within distributed recognition systems.
• To analyze the scalability and efficiency of AED-based systems using directed acyclic graphs.

Main Methods:

• Introduction of the activity-event-detector (AED) concept for distributed activity recognition.
• Formal methods for constructing and utilizing AEDs based on directed acyclic graphs.
• Evaluation through a home monitoring study focused on daily life activities.

Main Results:

• The proposed AED approach provides a structured method for building distributed activity recognition systems.
• AED graphs formally illustrate key properties contributing to system scalability and efficiency.
• The home monitoring study demonstrates the practical applicability and reconfiguration capabilities of the AED model.

Conclusions:

• The activity-event-detector (AED) offers a robust framework for distributed human activity recognition.
• The AED-based model enhances resource utilization, scalability, and dynamic reconfiguration in sensor networks.
• This approach is well-suited for real-world applications such as home monitoring and daily life activity analysis.