Decision Fusion-Based Deep Learning for Channel State Information Channel-Aware Human Action Recognition.

Area of Science:

• Computer Science
• Signal Processing
• Machine Learning

Background:

• WiFi channel state information (CSI) offers a non-invasive method for human action recognition.
• Current methods often process CSI channels collectively, potentially missing channel-specific insights.

Purpose of the Study:

• To propose a novel architecture, DF-CNN, for human action recognition that processes CSI channels individually.
• To evaluate the effectiveness of a decision fusion (DF) strategy for integrating channel-specific information.

Main Methods:

• Developed DF-CNN architecture for separate CSI channel processing.
• Implemented a decision fusion strategy to combine outputs from individual channels.
• Conducted extensive experiments to validate the proposed method.

Main Results:

• DF-CNN significantly outperformed traditional collective processing approaches.
• Achieved state-of-the-art performance in human action recognition using CSI.
• Demonstrated the effectiveness of separate channel processing and decision fusion.

Conclusions:

• Separate processing of CSI channels is crucial for enhancing human action recognition accuracy.
• The proposed DF-CNN architecture establishes a new benchmark for CSI-based action recognition.
• This approach highlights the importance of leveraging channel-specific information in signal processing applications.