Choosing the Best Sensor Fusion Method: A Machine-Learning Approach

Area of Science:

• Computer Science
• Engineering
• Data Science

Background:

• Multi-sensor fusion combines data from multiple sensors to improve accuracy and reliability.
• Selecting the optimal fusion method for specific sensor sets and applications is challenging.
• Previous work focused on human activity recognition, limiting its general applicability.

Purpose of the Study:

• To extend a machine learning-based sensor fusion method prediction model to new domains.
• To evaluate the generality and effectiveness of the proposed approach across different contexts.
• To provide a data-driven solution for selecting the best sensor fusion strategy.

Main Methods:

• Developed a machine learning model trained on statistical signatures from meta-datasets.
• Extended the model to new application domains: gas detection and facial expression identification.
• Validated the model's predictive performance on diverse datasets.

Main Results:

• The extended model accurately predicts the most suitable sensor fusion method for various datasets.
• Experimental results demonstrate the broad generality of the proposed approach.
• The data-driven method successfully addresses the challenge of choosing optimal fusion strategies.

Conclusions:

• The generalized machine learning approach for predicting sensor fusion methods shows significant promise.
• This method offers a robust solution for optimizing multi-sensor systems across diverse fields.
• The findings support the claim of broad applicability for the proposed sensor fusion prediction model.