Terahertz tag identifiable through shielding materials using machine learning.

Area of Science:

• Physics
• Electrical Engineering
• Materials Science

Background:

• Chipless radio-frequency identification (RFID) technology is gaining interest for terahertz (THz) applications.
• Practical THz RFID implementation faces challenges with cost, detection accuracy, and shielding materials.

Purpose of the Study:

• To propose novel, low-cost THz-tags for enhanced identification.
• To leverage machine learning for high-precision tag recognition in challenging environments.

Main Methods:

• Development of two types of low-cost THz-tags: one based on coated polyethylene thickness variation, another on reagent fingerprint spectra.
• Application of deep learning (convolutional neural networks) for tag identification from spectral data.
• Integration with a multiwavelength injection-seeded THz parametric generator for real-time analysis.

Main Results:

• Achieved nearly 100% identification accuracy for inexpensive THz-tags.
• Demonstrated successful tag identification through shielding materials with -50 dB attenuation.
• Validated real-time tag identification capabilities.

Conclusions:

• The proposed low-cost THz-tags combined with deep learning offer a viable solution for practical THz RFID systems.
• This approach overcomes significant limitations posed by shielding materials and weak signals.
• High-precision, real-time identification is achievable even with cost-effective tag designs.