Group Decision Making-Based Fusion for Human Activity Recognition in Body Sensor Networks

Area of Science:

• Computer Science
• Artificial Intelligence
• Machine Learning

Background:

• Ensemble learning systems (ELS) are common for human activity recognition (HAR).
• Traditional HAR ensemble methods face challenges with classifier accuracy, diversity, pruning, and fusion efficiency.

Purpose of the Study:

• To propose a novel selective ensemble approach with group decision-making (GDM) for decision-level fusion in HAR.
• To enhance the accuracy and efficiency of HAR systems by addressing limitations of traditional ensemble methods.

Main Methods:

• Constructing diverse local base classifiers tailored to specific sensors.
• Implementing ensemble pruning using mixed diversity and complementarity measures.
• Employing a group decision-making (GDM) fusion strategy for combining classifier decisions.

Main Results:

• The proposed GDM-based selective ensemble approach demonstrated superior performance.
• Experimental results on the OPPORTUNITY and DSAD datasets validated the method's effectiveness.
• The approach outperformed existing fusion techniques and state-of-the-art methods in HAR.

Conclusions:

• The novel GDM-based selective ensemble approach offers a significant improvement for HAR.
• This method effectively addresses the limitations of traditional ensemble techniques in HAR.
• The findings suggest a promising direction for future research in sensor-based activity recognition.