Activity recognition using multiplex limited penetrable visibility graph.

Area of Science:

• Computer Science
• Biomedical Engineering
• Data Science

Background:

• Human activity recognition (HAR) faces challenges in distinguishing dynamic activities.
• Existing methods may lack adaptability for complex movement patterns.

Purpose of the Study:

• To propose a novel method for dynamic activity recognition using a multiplex limited penetrable visibility graph (MLPVG).
• To enhance the accuracy and adaptability of human posture and gait recognition.

Main Methods:

• Mapping 21 pressure values to nodes in a first-layer network.
• Constructing a second-layer network using average path length for activity exploration.
• Analyzing characteristic and dynamic parameters, including average clustering coefficient and maximum degree.

Main Results:

• The MLPVG method effectively discriminates between different postures.
• The approach demonstrates good adaptability in recognizing dynamic activities.
• Joint distribution of clustering coefficient and maximum degree parameter improved discrimination.

Conclusions:

• The proposed MLPVG method offers a new approach for human activity and gait recognition.
• This technique has potential applications in medical clinical diagnosis, rehabilitation, and public health.
• The method provides enhanced discrimination capabilities for dynamic human movements.