Video Experimental Relacionado
Updated: Jan 8, 2026
09:43
Transmission of Multiple Signals through an Optical Fiber Using Wavefront Shaping
Published on: March 20, 2017
10.3K
Ecualizador basado en LSTM con mecanismo de atención de múltiples cabezales y compuerta para sistema inalámbrico THz
Optics express
|December 19, 2025
Resumen
Un nuevo ecualizador MHAG-LSTM mejora la comunicación de terahertz asistida por fotónica al reducir las tasas de error de bits y la complejidad computacional. Este modelo avanzado mejora la compensación no lineal para futuros sistemas 6G.
Área de la Ciencia:
- Optical Communications
- Terahertz (THz) Technology
- Machine Learning for Signal Processing
Sus antecedentes:
- Photonics-assisted THz communication is crucial for 6G, offering high speed and capacity.
- Nonlinear damage severely impacts THz signal transmission quality.
- Existing equalization methods struggle with nonlinear impairments in high-frequency systems.
Objetivo del estudio:
- To propose and evaluate a novel Long Short-Term Memory (LSTM)-based equalizer with Multi-Head Attention and Gating Mechanism (MHAG-LSTM).
- To enhance nonlinear compensation capabilities in photonics-assisted THz communication systems.
- To reduce computational complexity for future 6G high-frequency communication.
Principales métodos:
- Developed an MHAG-LSTM model combining LSTM's temporal modeling with attention's global reach and a gating mechanism for dynamic weight adjustment.
- Conducted experiments on a 200m outdoor wireless transmission of a 50 GBaud polarization-division multiplexing (PDM) QPSK signal at 140 GHz.
- Compared MHAG-LSTM performance against Fully Connected Neural Network (FCNN), Convolutional Neural Network (CNN), LSTM, and Transformer equalizers.
Principales resultados:
- MHAG-LSTM achieved lower bit error rates (BER) than FCNN, CNN, LSTM, and Transformer equalizers under identical input power.
- Achieved BER below the 7% hard decision forward error correction (HD-FEC) threshold (3.8 × 10-3) at 16 dBm optical power.
- Reduced average equalized BER by 57.07% and real multiplications per symbol (RMpS) by 13.5% compared to CNN, LSTM, and Transformer.
Conclusiones:
- The MHAG-LSTM equalizer demonstrates superior nonlinear compensation in photonics-assisted THz communication.
- The model offers a promising solution for intelligent equalization in future 6G high-frequency, high-capacity systems.
- MHAG-LSTM provides a low-complexity implementation reference for advanced communication technologies.
