UHF RFID Sensing for Dynamic Tag Detection and Behavior Recognition: A Multi-Feature Analysis and Dual-Path Residual...

Area of Science:

• * Computer Science
• * Electrical Engineering
• * Artificial Intelligence

Background:

• * Current Ultra-High-Frequency Radio-Frequency Identification (UHF RFID) behavior recognition methods face challenges with dynamic coupling interference and time-frequency feature degradation.
• * Signal interference and loss of detailed temporal and frequency information limit the accuracy and robustness of existing recognition systems.

Purpose of the Study:

• * To develop a novel behavior recognition method for UHF RFID systems that effectively handles interference and preserves time-frequency features.
• * To improve the accuracy and robustness of recognizing interaction behaviors like 'taking away' and 'putting back' in complex environments.

Main Methods:

• * Mitigation of signal interference using phase difference methods, cross-correlation, and adaptive equalization algorithms.
• * Identification of active target tags via a 3D feature space and an improved weighted isolated forest algorithm.
• * Extraction of behavioral features using Doppler shift analysis and multiscale wavelet-packet decomposition for time-frequency representation.
• * Fusion of global and local features using a dual-path residual network for behavioral classification.

Main Results:

• * The proposed method successfully mitigates dynamic coupling interference and addresses time-frequency feature degradation.
• * Active tags participating in interactions are accurately detected.
• * The dual-path residual network effectively fuses multi-feature information for classification.
• * Achieved a behavioral recognition accuracy of 94% in complex scenarios.

Conclusions:

• * The integrated multi-feature analysis and dual-path residual network approach significantly enhances UHF RFID behavior recognition.
• * The proposed method demonstrates superior robustness and accuracy in complex environments, outperforming existing techniques.
• * This work provides a robust framework for advanced human-object interaction recognition using UHF RFID technology.