Human Activity Recognition via Score Level Fusion of Wi-Fi CSI Signals.

Area of Science:

• Computer Science
• Signal Processing
• Biomedical Engineering

Background:

• Wi-Fi signals offer a non-invasive method for human activity recognition, crucial for healthcare monitoring.
• Existing methods for Wi-Fi-based activity recognition have limitations in accuracy and processing time.

Purpose of the Study:

• To develop and evaluate a score-level fusion structure for human activity recognition using Wi-Fi Channel State Information (CSI).
• To improve the generalization and reduce the learning processing time of Wi-Fi-based human activity recognition systems.

Main Methods:

• Raw Wi-Fi CSI signals were preprocessed.
• Conventional classifiers were used for initial classification.
• An analytic network was employed for score-level fusion of classifier outputs, avoiding iterative learning.

Main Results:

• The proposed score-level fusion structure demonstrated good generalization capabilities.
• The fusion method achieved a shorter learning processing time compared to state-of-the-art networks.
• The system effectively recognizes human activities using Wi-Fi CSI signals.

Conclusions:

• Score-level fusion of Wi-Fi CSI signals presents an effective approach for human activity recognition.
• This method offers a promising, efficient, and accurate solution for healthcare monitoring applications.